La complejidad en el fútbol: dinámica exploratoria y emergencia de

comportamiento táctico

Ángel Ric Diez

http://hdl.handle.net/10803/404150

L'accés als continguts d'aquesta tesi queda condicionat a l'acceptació de les

condicions d'ús establertes per la següent llicència Creative Commons:
http://creativecommons.org/licenses/by-nc-nd/4.0/

(c) 2017, Ángel Ric Diez

 TESIS DOCTORAL

La complejidad en el fútbol: dinámica

exploratoria y emergencia de comportamiento
táctico

Ángel Ric Diez

Memoria presentada para optar al grado de Doctor por la Universitat de Lleida

Programa de Doctorado en Actividad Física y Deporte

Directores
Robert Hristovski
Carlota Torrents Martín

2017
A Emili. Por tu generosidad, humildad y bondad eternas.
Simplemente gracias
 Agradecimientos

A Fátima, por estar siempre ahí, en los momentos de mayor dificultad para animarme a
continuar, y en los de mayor euforia y alegría tras la publicación de cada uno de los
artículos. Gracias por tu infinita paciencia durante estos “tres años de sufrimiento”, como tú
dices; pero que también han servido para sentir tu admiración por mi trabajo y mi persona.
Empezamos juntos el que hasta ahora ha sido el proyecto más importante de mi vida para
en unos meses tener entre nuestros brazos al que pasará a ser el más importante de las
nuestras.

A mi familia, en especial a mi madre y a mi padre. A ti mamá por acompañarme siempre a

todos los lados. Me acuerdo de los viajes en autobús en los que nos acompañabas cuando
comenzaba a darle patadas al balón, aquella devoción por el fútbol es la que me ha
permitido investigarlo con la misma pasión con la que jugaba. Y a ti papá, porque con tu
esfuerzo para proporcionarme la formación, no solo académica, que tú no pudiste tener, has
conseguido hacerme llegar más lejos de lo que yo nunca hubiera imaginado, y seguramente
más allá de lo que para ti ya era un orgullo. ¡Os quiero!

A Carlota Torrents, por hacerme el mejor regalo después de aquella clase de expresión
corporal que nunca olvidaré, decir que sí a dirigir mí trabajo final de máster, en fútbol…
que se ha acabado convirtiendo en una tesis doctoral. Gracias por la oportunidad que me
diste y por no dejar de constreñirme, siempre con única intención de aumentar mis grados
de libertad.

To Robert Hristovski, for gifting myself with your infinite knowledge. For always being
available to solve my doubts and ready to find a solution to any problem. Thank you for
supervising this thesis and for hosting me for the three months which I was in Skopje, one
of the best academic and personal experiences of my life.

A Bruno Gonçalves, porque tú has sido mucho más que un compañero, un verdadero
amigo. Te considero un director más de esta tesis, no sólo has ayudado a mejorar el trabajo
que aquí se presenta, sino que también me has ayudado a mejorar como investigador y

2
como persona. Gracias por todo lo que me has enseñado y por lo que todavía te queda por
enseñarme.

A Lorena Torres, porque aunque solo hayamos compartido cafés y conversaciones durante
la primera parte del partido, sin duda tu apoyo y compañía me sirvió para superar quizás los
momentos más complicados de este proceso. Gracias por hacer de directora cuando la jefa
no pudo estar presente y por compartir conmigo algunos de tus múltiples proyectos.

A Jaime Sampaio, por tu profesionalidad y por dar luz verde a todos los proyectos
propuestos. Gracias por tratar de buscar siempre la innovación científica e intentar
encontrar las aplicaciones prácticas a cada una de las investigaciones.

Al Reus Deportiu, Lleida Esportiu y C.D. Ebro de Zaragoza, a todos sus jugadores y
cuerpos técnicos por colaborar en el diseño del protocolo y recogida de datos para la
investigación. En especial a Emili Vicente, Josep Maria Martínez “Joma” y Jose Luis Arjol
por facilitar e intermediar con cada club para hacer posible la realización de los estudios
con jugadores de fútbol masculino profesionales.

A Àlex Gordillo, Pere Lavega y Cristòfol Salas por ofrecerme la oportunidad de trabajar
como profesor en las asignaturas de aprendizaje motor, teoría y práctica del juego motor, y
deportes colectivos I. Además a los dos últimos por ayudarme desde la subdirección y
dirección del INEFC de Lleida.

A Montse Sanmartí, por su complicidad, sus consejos y su ayuda durante los tres años de
beca con toda la parte burocrática y en el desarrollo de mis funciones como secretario de la
Comisión Académica de Doctorado.

A la Agencia de Gestión de Ayudas Universitarias y de Investigación (AGAUR), a la

Universidad de Lleida (UdL) y al Instituto Nacional de Educación Física de Cataluña
(INEFC) por el apoyo institucional y la financiación para formarme como investigador y
docente.

A todos los miembros del grupo de investigación Sistemas Complejos y Deporte, por hacer
crecer y mejorar la investigación desde esta perspectiva. En especial a mis compañeros
Daniel Aragonés, Pablo Vázquez, Sergi Garcia y Agne Slapsinskaite quienes me han
ayudado desde el inicio a profundizar en el conocimiento de los sistemas complejos y
3
enseñarme la posibilidades que nos da la utilización de esta teoría en la investigación
científica. Gracias también, a Natalia Balagué, la coordinadora del grupo, por toda su
energía, su forma de transmitir conocimiento y su predisposición para iniciar siempre
nuevos proyectos que nos permiten seguir creciendo personal y académicamente.

To Alexander Aceski, for welcoming me into your home, introducing me to your friends
and giving me your friendship during my time in Skopje. I will never forget those three
months. Thank you Tejonazo! You will be a friend for life.

4
 Resumen

Las ciencias de la complejidad y en concreto la teoría de sistemas dinámicos no

lineales ofrecen una serie de principios, conceptos y herramientas metodológicas que han
permitido hacer grandes avances en el estudio del fútbol. Recientemente, se han aportado
evidencias empíricas que explican la emergencia de patrones de coordinación a diferentes
niveles de análisis y la influencia de diferentes constreñimientos propios del juego o
relacionados con las tareas de entrenamiento. A pesar de estos avances, resulta
imprescindible profundizar en la comprensión de la dinámica o estructura temporal
inherente a estos procesos. Por este motivo el objetivo de esta tesis fue identificar las
dinámicas de comportamiento táctico que emergen bajo la influencia de ciertos
constreñimientos en diferentes situaciones de juego en futbol masculino. Dos tipos de
variables generales (los componentes principales y el promedio del coseno de similitud a
distancias de tiempo incrementales) y cuatro específicas (las acciones tácticas y zonas de
campo así como la distancia a la zona de marca y al oponente más cercano) fueron
consideradas como parámetros de orden para caracterizar la estructura y la dinámica del
comportamiento táctico. Bajo la influencia del conjunto de constreñimientos derivados de
la competición, de los desequilibrios numéricos en juegos reducidos o de la restricción del
espacio de interacción durante una situación de entrenamiento se identificó una estructura
jerárquica del comportamiento táctico en la que el jugador o equipo explora diferentes
soluciones metaestables. De los resultados de los diferentes estudios presentados se puede
concluir que la manipulación adecuada de los constreñimientos permite moldear una
estructura y dinámica de comportamiento más regular y/o estable, o más variada y/o
flexible dependiendo de los objetivos que se persigan.

Palabras clave: constreñimientos, comportamiento exploratorio, parámetro de orden,

estado coordinativo, metaestabilidad, jerarquía, escalas temporales

5
 Resum

Les ciències de la complexitat i en concret la teoria de sistemes dinàmics i no lineals

ofereixen una sèrie de principis, conceptes i eines metodològiques que han permès fer grans
avenços per a l'estudi del futbol. Recentment, s'han aportat evidències empíriques que
expliquen l'emergència de patrons de coordinació a diferents nivells d'anàlisi i la influència
de diferents constrenyiments propis del joc o relacionats amb les tasques d'entrenament.
Malgrat aquests avenços, resulta imprescindible aprofundir en la comprensió de la dinàmica
o estructura temporal inherent a aquests processos. Per aquest motiu l'objectiu d'aquesta tesi
fou identificar les dinàmiques de comportament tàctic que emergeixen sota la influència de
certs constrenyiments en diferents situacions de joc en futbol masculí. Dos tipus de
variables generals (els components principals i la mitjana del cosinus de similitud a
distàncies de temps incrementals) i quatre específiques (les accions tàctiques i zones de
camp així com la distància a la zona de marca i a l'oponent més proper) van ser
considerades com a paràmetres d'ordre per caracteritzar l'estructura i la dinàmica del
comportament tàctic. Sota la influència del conjunt de constrenyiments derivats de la
competició, dels desequilibris numèrics en jocs reduïts o de la restricció de l'espai
d'interacció durant una situació d'entrenament es va identificar una estructura jeràrquica del
comportament tàctic en la qual el jugador o equip explora diferents solucions metastables.
Dels resultats dels diferents estudis presentats es pot concloure que la manipulació
adequada dels constrenyiments permet modelar una estructura i dinàmica de comportament
regular i/o estable o més variada i/o flexible depenent dels objectius que es persegueixin.

Paraules clau: constrenyiments, comportament exploratori, paràmetre d'ordre, estat

coordinativo, metaestabilitat, jerarquía, escales temporals

6
 Abstract

Complexity sciences and specifically non-linear dynamic system theory offer a set of
principles, concepts and methodological tools that have provided grateful advances for the
study of football. Recently, empirical evidence has shown the emergence of coordinative
patterns at different levels of analysis and the influence of different game or training task
constraints. Nevertheless, it is essential to go further in the understanding of the inherent
dynamics or temporal structure of these processes. For this reason, the aim of this thesis
was to identify the dynamics of tactical behavior emerging under the influence of
constraints in different game scenarios of male football. Two general (the principal
components and the lagged average cosine similarity) and four specific (the tactical actions
and field zones as well as the distance to the target and to the nearest opponent) variables
were treated as order parameters in order to characterize the structure and dynamics of
tactical behavior. Under the influence of the set of constraints derived from the
competition, the numerical imbalances in small-sided games or the restriction of the
interaction space during a training game situation, a hierarchical structure of behavior was
identified in which the player or team explores many different metastable tactical solutions.
From the results of the different studies, it can be concluded that the adequate manipulation
of the constraints allows shaping a structure and dynamics of behavior more regular and/or
stable or more variate and/or flexible depending on the goals.

Key words: constraints, exploratory behavior, order parameter, coordinative state,

metastability, hierarchy, timescales

7
 Índice de contenidos
Agradecimientos ..................................................................................................................... 2
Resumen ................................................................................................................................. 5
Resum ..................................................................................................................................... 6
Abstract................................................................................................................................... 7
1. Introducción .................................................................................................................. 10
1.1. Bases teóricas para estudiar la complejidad en el fútbol ....................................... 10
1.2. Dinámicas de comportamiento táctico en deportes colectivos .............................. 15
1.2.1. Dinámicas interpersonales (jugador-jugador) ................................................ 16
1.2.2. Dinámicas intergrupales (equipo-equipo) ...................................................... 17
1.2.3. Dinámicas multinivel (jugador-equipo y/o equipo-jugador) .......................... 18
1.3. Rol de los constreñimientos en el entrenamiento .................................................. 19
1.4. Objetivo y estructura de la tesis ............................................................................. 21
2. Métodos ......................................................................................................................... 23
2.1. Participantes ........................................................................................................... 23
2.2. Adquisición y preparación de los datos ................................................................. 23
2.3. Análisis de los datos .............................................................................................. 26
2.3.1. Análisis Jerárquico de Componente Principales (hierarchical Principal
Component Analysis) .................................................................................................... 26
2.3.2. Análisis de superposición dinámica qd(t) (Dynamic overlap) ........................ 27
2.3.3. Análisis complementarios .............................................................................. 28
3. Estudio I - Soft-assembled Multilevel Dynamics of Tactical Behaviors in Soccer ..... 31
4. Estudio II - Timescales for exploratory tactical behaviour in football small-sided
games ................................................................................................................................... 44
5. Estudio III - Emergence of Exploratory, Technical and Tactical Behavior in Small-
Sided Soccer Games when Manipulating the Number of Teammates and Opponents ....... 53
6. Estudio IV - Dynamics of tactical behavior in football when manipulating players'
space of interaction ............................................................................................................... 69
7. Discusión ....................................................................................................................... 86
7.1. Síntesis de los principales hallazgos ...................................................................... 86
7.2. Aportaciones teóricas y metodológicas ............................................................... ..93
7.3. Aplicaciones prácticas ......................................................................................... ..95

8
 7.4. Limitaciones y perspectivas de investigación futuras.......................................... 100
8. Conclusiones ............................................................................................................... 104
9. Referencias .................................................................................................................. 107

9
 1.Introducción
Las recientes aportaciones de la teoría de sistemas dinámicos no lineales (TSDN) al
estudio de fenómenos relacionados con el deporte (Vázquez, 2017) han inspirado también
la investigación reciente en el fútbol. Concretamente, la propuesta teórica denominada
Coordination Dynamics (Kelso, 1995) define, explica y predice cómo se forman, adaptan,
persisten y cambian los patrones coordinativos en los seres vivos. Desde su aparición en la
década de los 80’ ha proporcionado conceptos y herramientas para verificar empíricamente
el comportamiento emergente en diferentes contextos relacionados con el aprendizaje
motor (Kelso, 1984) y el deporte (Araujo, Davids y Hristovski, 2006), y, más
recientemente, para el estudio del fútbol (Duarte, 2012). Este enfoque, junto con algunas
ideas de la psicología ecológica, permite abordar la comprensión de este deporte colectivo
examinando cómo múltiples variables pueden constreñir el comportamiento de los sistemas
neurobiológicos (deportistas) y de éstos actuando en conjunto (equipo/s), modificando los
diferentes estados estables de organización (patrones de comportamiento). En este primer
capítulo se introducen detalladamente los principios y conceptos clave que nos ayudan a
comprender el fútbol como un fenómeno complejo, dinámico y no lineal. En segundo lugar,
se revisa la literatura centrada en cómo estos principios y conceptos han servido para
estudiar la emergencia de comportamiento táctico y/o patrones coordinativos en deportes
colectivos a diferentes niveles de análisis. En tercer lugar, se recogen algunos trabajos que
proponen modificar estos comportamientos a partir de la manipulación de constreñimientos
relacionados con las tareas de entrenamiento. Y finalmente, se detalla la estructura y el
principal objetivo de la tesis.

1.1. Bases teóricas para estudiar la complejidad en el fútbol

Para tratar de responder a la pregunta de cómo jugadores o jugadoras y equipos son

capaces de resolver los problemas derivados de la práctica (ej. partido o tareas de
entrenamiento), la psicología ecológica enfatiza en la interacción entre el organismo y su
entorno (Gibson, 1979). Este enfoque sugiere que las personas perciben información del
entorno mediante la acción. Por lo tanto, la acción no sólo es guiada por la percepción, sino
que también es fundamental para ella. Este intercambio continuo de flujos de información
10
procedentes de la percepción y de los flujos cinéticos que se producen a partir de la acción
define la interfaz perceptivo-motor (Newell, McDonald y Kugler, 1991) o acoplamiento
(ciclo) percepción-acción (Warren, 1990). La propiedad fundamental de esta interacción
deportista-entorno son las affordances, que han sido definidas como oportunidades
(Gibson, 1979), invitaciones (Withagen, Poel, Araújo y Pepping, 2012) o solicitaciones
para la acción (Bruineberg & Rietveld, 2014). Por lo tanto, canalizan la acción del o la
deportista dentro del espectro total disponible (Chemero, 2003; Turvey, 1992). Por
ejemplo, la oportunidad de marcar gol emergerá de la interacción entre la habilidad de cada
individuo (la capacidad que tenga para golpear el balón con la tensión justa y en el lugar
adecuado), de la distancia entre la portería y el jugador o jugadora con balón, y entre éste y
su oponente directo, así como del ángulo que se deriva de estas dos (Vilar et al., 2014;
Vilar, Araújo, Davids, Correia y Esteves, 2013). Por otro lado, en deportes colectivos,
también debemos considerar el concepto de shared affordances, es decir, posibilidades de
acción colectivas (Silva, Garganta, Araújo, Davids y Aguiar, 2013). A través de la práctica,
las jugadoras o los jugadores son capaces de percibir situaciones concretas que pueden
generar un comportamiento colectivo armonizado y de adaptarse funcionalmente a las
acciones individuales de compañeros o compañeras del mismo equipo o del oponente
(Sampaio & Maçãs, 2012; Silva, Chung, et al., 2016). Por ejemplo, un pase hacia atrás de
un defensor oponente sobre el portero puede provocar un comportamiento sincronizado de
todo el grupo reduciendo los espacios en campo rival. Este mismo comportamiento
colectivo puede estar inducido por la presión inicial de un jugador sobre el oponente con
balón; al igual que el avance de la línea defensiva en su conjunto puede obligar al/ los
delantero/s a iniciar una acción de presión. Toda esta información que se desprende de las
acciones individuales o colectivas, fruto del enfrentamiento entre los miembros de dos
equipos, del reglamento y de la estrategia del entrenador o del estilo de juego, entre otros,
será relevante para la acción.

Cualquier variable que actúe como fuente de información durante el juego podrá
influir en las dinámicas de comportamiento del sistema (Fajen & Warren, 2003). Desde la
perspectiva que se aborda en esta tesis, estas variables reciben el nombre de parámetros de
control (Haken, 1983). Los parámetros de control son representaciones matemáticas de los
constreñimientos que actúan sobre cualquier sistema, entre ellos los biológicos,

11
presionándolo hacia diferentes estados de organización o patrones coordinativos concretos.
La manipulación de los constreñimientos, especialmente aquellos relacionados con las
tareas de entrenamiento, condicionará la emergencia de patrones de comportamiento
determinados (Davids, Button, Araújo, Renshaw y Hristovski, 2006). La modificación o
variaciones de este tipo de variables o parámetros de control en la dirección adecuada
pueden producir cambios funcionales en el comportamiento en contextos competitivos o de
rendimiento.

En este punto aparece una propiedad fundamental de los sistemas dinámicos no

lineales. A diferencia de cualquier sistema lineal, los comportamientos de sistemas
neurobiológicos pueden verse afectados de forma no proporcional ante cambios menores en
valores del parámetro de control. Es decir, pequeños cambios en el parámetro de control
pueden provocar cambios coordinativos abruptos. Por ejemplo, Passos y colaboradores
(2009) evidenciaron que los pequeños cambios en la distancia interpersonal y la velocidad
relativa atacante-defensor (parámetros de control) provocaron cambios cualitativos
sustanciales en el ángulo entre el atacante y el defensor (parámetro de orden), indicando el
éxito ofensivo en situaciones de 1 contra 1 en rugby masculino. Los parámetros de orden
son pues variables relacionales ya que representan la relación entre los componentes o
configuraciones del sistema. Así, muestran el tipo y el grado de orden (estabilidad) del
sistema a nivel macroscópico. Sin embargo, los parámetros de control no solo son variables
que no contienen la misma información que los parámetros de orden, sino que además
permiten alterar la relación entre los componentes, es decir, los valores del parámetro de
orden, y por lo tanto, modificar las propiedades de estabilidad del sistema.

Estos cambios cualitativos en los estados coordinativos del sistema están

caracterizados por la pérdida de estabilidad de un estado previo sobre el actual. En este
sentido, las dinámicas de comportamiento corresponden al cambio de trayectorias que se
producen en el espacio de estados, haciendo referencia a la hipotética totalidad de estados
coordinativos posibles de un sistema, y que puede estar definido por una o varias variables
coordinativas o parámetros de orden (Araujo et al., 2006; Haken, 1983). Estos parámetros
capturan las propiedades de estabilidad del sistema a nivel macroscópico dentro del espacio
de estados definido. Los estados más estables del sistema, que a su vez serán los más

12
probables o los que muestren mayor correlación entre sí, reciben el nombre de atractores,
por la atracción que generan dichos estados sobre las trayectorias que sigue el sistema
próximas al atractor. Por contra, los estados más inestables indican la baja probabilidad de
que emerja ese comportamiento o estado coordinativo. El paisaje potencial que se muestra
en la figura 1 representa el espacio de estados donde se puede apreciar el total de
configuraciones que realiza un jugador. En este ejemplo, la combinación de la distancia del
balón respecto a su portería (variable y) y la trayectoria de movimiento del jugador respecto
al balón (variable x) se considera como parámetro de orden. En rojo se pueden apreciar los
estados que atraen al sistema mientras que en azul se representan aquellos comportamientos
que son menos probables que realice el jugador. A modo de ejemplo se puede decir que
cuando el balón está cerca de la portería propia, el jugador tiende a moverse hacia el balón,
pero nunca en una trayectoria frontal (0º). Además se puede ver que cuando el balón está a
la altura del área, nunca se alejará del balón (180º), sin embargo este movimiento es
probable que se dé cuando el balón está en campo contrario, aunque la tendencia (atractor)
será a ofrecerse al portador del balón.

13
Figura 1. Espacio de estados (configuraciones) de un jugador y la dinámica asociada. El potencial
(Vi) indica un paisaje con muchos mínimos metastables (atractores) que atrapan temporalmente el
comportamiento del sistema.

Como cualquier sistema complejo, deportistas y equipos están sometidos a

múltiples constreñimientos, como son los propios de la competición (instrucciones del
entrenador o de la entrenadora, estrategia del equipo rival, importancia del partido,
marcador, etc.) o los impuestos en las tareas de entrenamiento. Debido a la influencia de
éstos el sistema deportista/equipo no podrá explorar de forma arbitraria extensas áreas del
espacio de estados disponible, dando lugar a las asociaciones entre configuraciones
(estados) de movimiento o comportamientos, haciendo que unas sean más estables (mayor
correlación/probabilidad) respecto a otras. Esto dará lugar a la formación de una estructura
jerárquica de comportamiento caracterizada por una dinámica metaestable (Hristovski,
Davids, Araújo y Passos, 2011). La metaestabilidad está relacionada con la existencia de
múltiples escalas temporales. Escalas cortas, donde el sistema parece estar en un estado de
equilibrio, es decir, temporalmente estable; sin embargo, únicamente explora una sección

14
concreta del espacio de estados disponibles, mientras que en escalas de tiempo mayores,
transita entre todos los estados metaestables (Bovier, 2006).

Por lo tanto, es de particular interés para las ciencias del movimiento y el deporte
descubrir cómo influyen ciertos constreñimientos (parámetros de control) en la dinámica
exploratoria y el comportamiento táctico emergente (parámetros de orden). Especialmente
aquellos de alta validez ecológica, es decir, aquellos ampliamente utilizados durante el
entrenamiento o que puedan aparecer durante la competición. Estos constreñimientos
desempeñarán un papel esencial para provocar cambios en el comportamiento individual y
colectivo, así como, para determinar los contextos favorables en los cuales deportistas y
equipos sean capaces de coordinarse funcionalmente con el fin de alcanzar un objetivo.

1.2. Dinámicas de comportamiento táctico en deportes colectivos

Tradicionalmente, el análisis de comportamiento en el fútbol se ha centrado en la

anotación de acciones técnicas relacionadas por el jugador con balón, como pueden ser los
pases, las interceptaciones, los regates o chutes, entre otras (Liu, Gómez, Gonçalves y
Sampaio, 2016; Rampinini, Impellizzeri, Castagna, Coutts y Wisløff, 2009). Resulta
complicado comprender el comportamiento aislando la acción de su contexto, así como
obviar el resto de acciones que puedan realizar los oponentes o compañeros sin balón. Por
lo tanto, para contextualizar las acciones técnicas individuales dentro del comportamiento
colectivo e identificar el comportamiento táctico de jugadores y equipos se requiere tener
en cuenta la interacción espacio-temporal entre los jugadores y el entorno. Por ejemplo,
algunos autores han estudiado el posicionamiento de los jugadores sobre el terreno de juego
y/o sus trayectorias de movimiento junto con la localización del balón y la distancia a las
porterías (McGarry, 2009). La interacción entre la intención del deportista para actuar y los
rápidos e impredecibles cambios del entorno provoca la exploración de soluciones tácticas
individuales o colectivas. Así, la dinámica exploratoria se puede definir como la
emergencia y disolución de patrones de comportamiento táctico o reconfiguraciones de
movimiento ante la aparición consecutiva de nuevos contextos de juego. (Hristovski et al.,
2013, 2011; Hristovski, Davids, Passos y Araújo, 2012).

15
 En el caso de los deportes colectivos de invasión masculinos se pueden identificar y
estudiar dinámicas de comportamiento táctico a tres niveles distintos1: A nivel
intrapersonal, donde el jugador trata de desarrollar patrones de comportamiento o
configuraciones de movimiento que le permitan dar solución a la heterogeneidad de las
tareas-problema que debe solucionar durante el desarrollo del juego. A nivel interpersonal,
en el que los jugadores del equipo tratan de coordinar sus acciones o movimientos para
conseguir objetivos compartidos. Finalmente, a nivel intergrupal, donde se analizan y
estudian los patrones de comportamiento o coordinación entre dos equipos enfrentados con
el fin de puntuar y evitar que lo haga el rival, para finalmente ganar el partido o no
perderlo. A continuación se describen los niveles interpersonales e intergrupales, ya que no
se han encontrado estudios específicos de fútbol a nivel intrapersonal.

Durante los últimos años ha habido un incremento en el uso de medidas no lineales

para evaluar la estructura en evolución temporal que subyace a la dinámica de
comportamiento. Herramientas y métodos tales como el análisis de fase relativa, análisis de
correlaciones o medidas de entropía se han utilizado con el fin de conocer la estabilidad,
regularidad y previsibilidad de tales comportamientos (Harbourne & Stergiou, 2009;
Stergiou, Buzzi, Kurz y Heidel, 2004).

1.2.1. Dinámicas interpersonales (jugador-jugador)

Utilizando como referencia los avances en otras disciplinas como la vela, el criquet
o el boxeo (Araujo et al., 2006), el rugby fue uno de los primeros deportes colectivos en los
que se estudió la dinámica de comportamiento identificando parámetros de orden y control
relevantes que permitieran describir con precisión el comportamiento del sistema en sub-
fases específicas del juego (ej. 1 contra 1) (Passos et al., 2008). Bourbousson y
colaboradores (2010) estudiaron las dinámicas de comportamiento entre diadas de
jugadores de baloncesto utilizando la fase relativa (ver para revisión completa Kelso, 1995;
McGarry, 2006; Palut & Zanone, 2005) como parámetro de orden para identificar fases
(estados) estables e inestables así como transiciones en las relaciones de fase. Ellos

1 No se han encontrado estudios que sigan esta línea de investigación en deportes colectivos de
equipos femeninos. Por este motivo, todos los ejemplos que se describen en este apartado y los
siguientes hacen referencia al comportamiento de jugadores o equipos masculinos.
16
evidenciaron que los jugadores (jugador-compañero, jugador-oponente) tendieron a
moverse coordinados en fase, es decir, moviéndose en la misma dirección a la vez en el eje
longitudinal. Por otro lado, identificaron estrategias de marcaje al hombre así como los
momentos en los que alguna perturbación (finta, dribling, etc) provocó la pérdida de
estabilidad en el estado de coordinación estable obligando al sistema jugador-oponente a
coordinarse de otro modo en caso de no poder recuperar la estabilidad en el estado anterior.
El análisis de fase relativa también se utilizó para analizar la coordinación de movimiento
entre jugadores en situaciones específicas de 5 contra 4 más el portero permitidas por el
reglamento de fútbol sala, donde tras ser notificado, un jugador de campo puede sustituir al
portero para generar superioridad ofensiva (Travassos, Araújo, Vilar y McGarry, 2011).
Estos autores incluyeron el balón en el análisis y demostraron que los defensores están
mucho más acoplados al movimiento del balón mostrando una menor distribución de
frecuencias en torno a los 0 grados (coordinación en fase) y por lo tanto mayor estabilidad o
rigidez en su comportamiento en relación al balón. También se ha demostrado como, tras
un periodo aproximado de tres meses (trece semanas), jugadores de fútbol sala
desarrollaron comportamientos mucho más regulares (estables) que en el test inicial
(Sampaio & Maçãs, 2012). Tras el mismo periodo de tiempo, Silva y colaboradores (2016)
demostraron que el número de acoplamientos entre los jugadores aumentó
significativamente disminuyendo el tiempo necesario para actuar de forma sincronizada.
Otros factores que parecen resultar clave para competir de forma sincronizada son el nivel
del oponente y la densidad de partidos (Folgado, Duarte, Fernandes y Sampaio, 2014;
Folgado, Duarte, Marques y Sampaio, 2015). Estos autores demostraron que cuanto mayor
es el nivel del oponente más sincronizado actúa el equipo, así como que el hecho de jugar
partidos entre semana disminuye los niveles de sincronización de los equipos. También se
observó que la sincronización es mayor entre los jugadores que ocupan posiciones más
retrasadas y que disminuye a medida que la posición de los jugadores es más adelantada.

1.2.2. Dinámicas intergrupales (equipo-equipo)

En relación al equipo se han identificado variables derivadas del posicionamiento

del equipo como son el centro geométrico del equipo (centroid), el índice de dispersión
(stretch index), el área (surface area) o la amplitud y profundidad, que pueden aportar

17
información relevante para conocer el comportamiento colectivo (Duarte et al., 2012). Las
dos primeras variables fueron procesadas también mediante la fase relativa en baloncesto
(Bourbousson, Seve y McGarry, 2010) demostrando que los equipos opuestos están
fuertemente atraídos hacia modos de coordinación en fase. Sin embargo, perturbaciones en
el eje latitudinal, es decir, periodos en los que se perdía este modo de coordinación,
anticipaban lanzamientos a canasta. Por otro lado, los estados estables en modos de
coordinación en fase relativos al índice de dispersión mostraban una clara relación con los
intercambios de posesión del balón, es decir, cuando un equipo tiene la posesión del balón
tiende a expandirse mientras el equipo oponente se comprime. En situaciones de juego
reducido en el fútbol diferentes análisis de correlación mostraron resultados similares
(Duarte, Araújo, Freire, et al., 2012; Frencken, Lemmink, Delleman y Visscher, 2011). Sin
embargo, un estudio reciente demostró que esto no pasaba en situaciones de 11 contra 11,
ya que cuando un equipo tendía a expandirse el otro también, sin que este índice de
dispersión pudiera asociarse a los momentos de posesión en partidos de fútbol (Moura et
al., 2016). Además, estos autores identificaron cuatro modos de coordinación. Las
frecuencias de modos de coordinación anti-fase (expansión del equipo atacante y
contracción del defensor) en el primer tercio de las secuencias ofensivas fueron
significativamente superiores en aquellas secuencias que finalizaron con lanzamiento a
portería. Teniendo en cuenta la velocidad promedio de subgrupos de jugadores se demostró
que durante la confrontación (defensores-atacantes, medios-medios, atacantes-defensores)
estos subgrupos tendían a competir acoplados (Memmert, Lemmink y Sampaio, 2017), es
decir que si los atacantes de un equipo aumentaban la velocidad de desplazamiento también
lo hacían los defensores oponentes. Destacar que esta sincronización disminuyó
significativamente en las tres confrontaciones durante la segunda parte. Estos mismos
autores identificaron una perturbación significativa en la diferencia de velocidad promedio
entre los diferentes subgrupos enfrentados en instantes previos al gol.

1.2.3. Dinámicas multinivel (jugador-equipo y/o equipo-jugador)

Recientemente se ha profundizado en la importancia del vínculo existente entre las

dinámicas de comportamiento individual y colectivo en deportes de equipo (Araújo et al.,
2015). Esta relación multinivel puede ser explicada por la relación circular resultante de la

18
asociación (correlación) entre comportamientos identificados a diferentes niveles de
organización jerárquica del sistema y que se caracterizan por su evolución en escalas
espacio-temporales distintas (Hristovski et al., 2011). Así, en un nivel superior ciertos
comportamientos evolucionan más lentamente cambiando a escalas de tiempo más largas,
mientras que en un nivel inferior los cambios se producen más rápido por lo que esos
comportamientos permanecen estables en escalas temporales más cortas. Bourbousson y
col (2014) encontraron esta asociación entre las dinámicas colectivas e individuales
identificando que una pérdida de estabilidad de un estado de coordinación en fase
(sincronía) en el eje latitudinal entre los centros geométricos y el índice de dispersión de los
equipos inducía a la penetración del jugador con balón. Otro estudio a destacar es el de
Duarte y colaboradores (2013) que demostró como la baja irregularidad (estabilidad) a
nivel colectivo puede ser compensada y asociada a los valores más altos a nivel individual
como un tipo de compensación recíproca (Riley, Richardson, Shockley y Ramenzoni, 2011)
o sinergia multinivel. A pesar de estos ejemplos aislados todavía no ha sido demostrado
empíricamente esta relación en los diferentes niveles de organización del sistema.

1.3. Rol de los constreñimientos en el entrenamiento

Podemos encontrar en la literatura numerosas investigaciones que han estudiado la

influencia de los constreñimientos y su manipulación en el fútbol masculino. Sin embargo,
la mayoría de éstos están centrados en la respuesta física y fisiológica del jugador (sin
prestar atención la dinámica de esa respuesta) durante la práctica de juegos reducidos (Hill-
Haas, Dawson, Impellizzeri y Coutts, 2011). También se ha estudiado su influencia en el
número de acciones técnicos realizados (Abrantes, Nunes, Maçãs, Leite y Sampaio, 2012;
Kelly & Drust, 2009). Incluso existen investigaciones sobre el efecto que tienen en el
comportamiento táctico la manipulación del número de porterías (Travassos, Gonçalves,
Marcelino, Monteiro y Sampaio, 2014), de la disposición de éstas o su no utilización
(Almeida, Duarte, Volossovitch y Ferreira, 2016), del espacio de juego (Vilar, Duarte,
Silva, Chow y Davids, 2014), del número de jugadores, del nivel del equipo (Silva, Duarte,
et al., 2014) o de la densidad de jugadores (Silva et al., 2015).

También existen investigaciones sobre la manipulación de la relación numérica

entre equipos, demostrando que cuando los jugadores se enfrentan a desventajas numéricas
19
los acoplamientos más fuertes se produjeron entre los defensa y el balón (Travassos, Vilar,
Araújo y McGarry, 2014). Estos hallazgos demuestran que los jugadores de equipos en
inferioridad priorizan la protección de la portería, es decir, evitar encajar un gol (Silva,
Travassos, et al., 2014). En un estudio más reciente se demostró que el aumento en el
número de oponentes aumentaba la rigidez del sistema expresado por la disminución en los
valores de entropía del espacio efectivo de juego, el índice de dispersión y la distancia a los
compañeros y oponentes más cercanos (Gonçalves, Marcelino, Torres-Ronda, Torrents y
Sampaio, 2016). Sin embargo, el aumento en el número de compañeros supuso un aumento
en la regularidad de las medidas anteriores. Esta misma tendencia se encontró en un estudio
anterior pero en el que los equipos compitieron en equilibrio numérico (Aguiar, Gonçalves,
Botelho, Lemmink y Sampaio, 2015). Resultados similares se obtuvieron utilizando la
distribución del porcentaje de tiempo en la serie de valores del índice de dispersión y del
radio de acción libre de oponentes, en los que el aumento de jugadores incrementó la
rigidez ya que se desarrollaron menos variedad de comportamientos y durante un mayor
porcentaje de tiempo (Silva, Vilar, Davids, Araújo y Garganta, 2016).

La literatura ha avanzado en la identificación de ciertos estados coordinativos

durante la competición y el entrenamiento. A pesar de ello sigue siendo escaso el
conocimiento generado por la investigación sobre la dinámica y la estructura de
comportamiento táctico en fútbol. Además, el alto grado de incertidumbre que caracteriza a
los deportes colectivos genera la necesidad de conocer los efectos de la modificación de
parámetros de control clave para desarrollar comportamientos variados y flexibles, así
como para estabilizar aquellos que puedan resultar más eficaces durante la competición. A
partir de la conceptualización teórica introducida y el modelo de jerarquía soft-assembled
(Hristovski et al., 2013) se ha analizado y verificado la dinámica y estructura del
comportamiento táctico y exploratorio en deportes o prácticas motrices individuales
(Hristovski et al., 2011, 2012; Torrents, Hristovski, Coterón y Ric, 2016; Torrents, Ric y
Hristovski, 2015). Los hallazgos encontrados por estos autores sugieren la hipótesis de que
en deportes colectivos, concretamente en el fútbol, también puede existir una estructura
jerárquica de comportamiento táctico. La hipótesis de partida de esta tesis es que tanto esta
estructura como la dinámica exploratoria pueden ser modificadas mediante la manipulación
de parámetros de control clave.

20
 1.4. Objetivo y estructura de la tesis

El objetivo principal de esta tesis es identificar la dinámica exploratoria y la

estructura de comportamiento táctico que emergen a partir de la manipulación de los
constreñimientos en diferentes situaciones de entrenamiento y aquellos propios de la
competición en el fútbol masculino.

Un total de 4 artículos originales (3 publicados y 1 en revisión) en revistas

indexadas en la ISI web of knowledge constituyen el cuerpo principal de esta tesis, de los
cuales se derivan cuatro objetivos específicos. Cada uno de estos artículos se presenta como
capítulos individuales siguiendo el formato de publicación o el solicitado por la revista para
su admisión.

En este primer capítulo se han presentado las bases teóricas que permiten estudiar
las dinámicas de comportamiento apoyándonos en la TSDN, con un resumen de los
resultados derivados de investigaciones que se han centrado en el estudio de la
coordinación a diferentes niveles de análisis en el deporte colectivo, para finalizar con una
breve exposición del rol de los constreñimientos en la emergencia de comportamiento.

En el capítulo 2 se hace una breve exposición de la metodología utilizada para la

recogida y el análisis de datos.

En los siguientes 4 capítulos se presentan los objetivos específicos que

corresponden a los cuatro artículos que conforman el cuerpo de esta tesis:

- Estudio I: Identificar los patrones tácticos y las escalas temporales de las

variables que los definen durante un partido de fútbol y determinar la similitud
entre los patrones identificados en la primera y la segunda parte correspondiente
al artículo titulado: “Soft-assembled Multilevel Dynamics of Tactical Behaviors
in Soccer”, publicado en Movement Science and Sport Psychology, una sección
de la revista Frontiers in Psychology.
- Estudio II: Identificar la dinámica de comportamiento táctico emergente en
diferentes escalas temporales y cuantificar el comportamiento exploratorio a
corto y largo plazo en función del número de oponentes en juegos reducidos,

21
 que pertenece al artículo titulado: “Timescales for exploratory tactical behaviour
in football small-sided games”, publicado en la revista Journal of Sport Science.
- Estudio III: Determinar cómo el número de compañeros y oponentes afecta al
comportamiento exploratorio de equipos profesionales y amateur en juegos
reducidos, correspondiente al artículo titulado: “Emergence of Exploratory,
Technical and Tactical Behavior in Small-Sided Soccer Games when
Manipulating the Number of Teammates and Opponents”, publicado en la
revista Plos One.
- Estudio IV: Idenficar cómo las restricciones espaciales afectan tanto al
comportamiento táctico exploratorio como a la interacción del jugador con
balón con su entorno y compañeros, dentro del artículo titulado: “Dynamics of
tactical behavior in football when manipulating players' space of interaction”,
enviado a la revista Plos One, actualmente en revisión.

En el capítulo 7 se presenta la discusión general de los resultados obtenidos de los

cuatro estudios, algunas aportaciones teóricas y metodológicas a considerar, un apartado de
aplicaciones prácticas y otro sobre las limitaciones y futuras líneas de investigación.

En el capítulo 8 se presentan las conclusiones de la tesis.

Finalmente, en el capítulo 9 se incluye el listado de referencia incluidas en los

capítulos 1, 2, 7 y 8.

22
 2.Métodos
2.1. Participantes

Dos equipos profesionales de fútbol masculino (2ª división B) y alumnos de la

asignatura de fútbol del grado en ciencias de actividad física y deporte (CCAFD)
participaron en los estudios de esta tesis. Por un lado, veintiún jugadores de un equipo
profesional participaron en los estudios I y IV. Por otro lado, veintidós jugadores del otro
equipo participaron en el estudio II, y, junto a los veintidós alumnos de CCAFD,
constituyeron la muestra del estudio III (los detalles de los jugadores, en relación a edad,
años de experiencia, altura y peso se detallan en cada uno de los estudios). Todos los
participantes fueron informados sobre los procedimientos, requisitos, beneficios y riesgos
de cada una de las investigaciones en las que participaron, dando su consentimiento por
escrito. Las investigaciones fueron aprobadas por el Comité de Ética de la institución local,
ajustándose a las recomendaciones de la Declaración de Helsinki.

2.2. Adquisición y preparación de los datos

Los datos se obtuvieron mediante observación sistemática, el uso de dispositivos

GPS (Global Positioning systems) (SPI ProX, GPSports, Canberra, ACT, Australia) y la
combinación de ambos. Todas las situaciones de juego fueron grabadas utilizando una o
varias cámaras de vídeo digitales. Se construyeron varios instrumentos de observación ad-
hoc con el fin de registrar las acciones técnico-tácticas utilizando para ello el software
“Lince” (Gabin, Camerino, Anguera y Castañer, 2012). Los análisis observacionales fueron
realizados por dos entrenadores de fútbol experimentados independientes (licenciados en
Ciencias del Deporte, en posesión de la licencia UEFA PRO y con 8,5±2,1 años
experiencia como entrenadores). El análisis Kappa arrojó coeficientes satisfactorios intra- e
inter-observadores (ver detalles en los artículos), garantizando la fiabilidad de los datos.

Los datos de posición de los jugadores se registraron utilizando dispositivos. Los

valores de latitud y longitud obtenidos de cada jugador fueron transformados en metros
alineando la longitud del campo de juego con el eje X y el ancho con el eje Y (para más

23
detalles consultar Folgado, Duarte, Fernandes y Sampaio, 2014). Los datos recogidos a
frecuencias de 5 y 15 hercios (Hz) fueron filtrados y suavizados a 1 Hz (1 dato por
segundo) con el objetivo de facilitar la sincronización con los datos obtenidos a partir de la
observación.

Los datos registrados permitieron identificar la secuencia de configuraciones de

movimiento definidas por la combinación de categorías activas e inactivas (grados de
libertad). De esta forma por cada segundo de juego se fue asignando un 1 a las categorías
que se activaban y un 0 para las que no se daban, obteniendo así una matriz binaria,
temporal y multivariada, de n*m dimensiones, donde n corresponde al número de
categorías y m a los segundos que duró cada una de las situaciones de juego (Tabla 1).

Tabla 1. Ejemplo de matriz binaria que recoge las configuraciones de movimiento en cada
segundo de la unidad observada.

Tiempo (s.)
1 2 3 … m
Categorías
1 1 1 0 … 1

2 0 0 1 … 0

3 0 1 0 … 0

… … … … … 0

n 0 1 0 1 1

Para ilustrarlo gráficamente, en la figura 2 se muestran diferentes categorías cuya

combinación definió las configuraciones de movimiento individuales y colectivas. Por
ejemplo, la configuración de movimiento del jugador número 7 quedaría definida por la
activación de las categorías que hacen referencia a la realización de una cobertura ofensiva
corriendo desde zona 9, partiendo de una distancia de más de 37,5 metros sobre la línea de
fondo, entre el compañero y oponente más alejado y avanzado, estando el más cercano
entre 5 y 6 metros. A nivel colectivo la configuración de movimiento se definiría porque el
equipo se expande sobre un radio de más de 16,5 metros con una profundidad y amplitud

24
media y grande respectivamente, posicionados sobre el carril central del sector medio-
ofensivo y desplazándose rápido hacia delante y basculando ligeramente hacia la izquierda;
activándose en este caso las categorías pertinentes. Esta categorización se determinó a
partir de variables discretas (ej. zonas del campo o acciones tácticas) o de la segmentación
de variables continuas (ej. distancia al oponente más cercano o la velocidad de dispersión y
contracción del equipo).

Figura 2. Ilustración de las categorías (valores de variables) cuya combinación definen las
configuraciones de movimiento individuales (a: acción táctica, b: contexto de interacción, c: zona,
d: velocidad, e: distancia al oponente más cercano, f: distancia a la línea de fondo) y colectivas (g:
índice de dispersión, h: velocidad de dispersión/contracción, i: profundidad, j: amplitud, k: sector, l:
carril, m: velocidad longitudinal centro del equipo, n: velocidad latitudinal del centro del equipo).

Por lo tanto, se puede diferenciar un primer nivel de componentes o grados de libertad

donde se definen las configuraciones de movimiento. Un segundo, donde se sintetizan
múltiples grados de libertad para capturar y describir la relación entre los componentes del
sistema utilizando variables coordinativas o colectivas, también denominadas parámetros
de orden. Y un tercero, donde los parámetros de control (ej. número de oponentes) cambian

25
la relación entre los componentes, es decir, los valores del parámetro de orden, modificando
así las propiedades de estabilidad del sistema (ver tabla 2).

2.3. Análisis de los datos

Como ya se ha explicado en el primer capítulo, para entender las dinámicas del

juego es necesario identificar aquellos parámetros de orden que permiten explicar el
comportamiento del sistema. Las superposiciones u overlaps en el marco de la mecánica
estadística de los sistemas desordenados se tratan como parámetros de orden, ya que
representan el grado en que las configuraciones de movimiento del sistema están
correlacionadas (Parisi, 1983). La correlación de Pearson o correlación de Fisher y el
coseno de similitud entre dos vectores (binarios) aleatorios son dos medidas de asociación
intrínsecamente relacionadas y que han sido identificadas como potenciales parámetros de
orden (Davenport & El-Sanhurry, 1991; Domany, Hed, Palassini y Young, 2001). Por lo
tanto la correlación existente entre las configuraciones de movimiento de jugadores y/o
equipos puede ser tratada como grado de orden o coordinación del sistema en cuestión. Por
ello el análisis jerárquico de componentes principales y el análisis de superposición
dinámica (dynamic overlap) qd(t) se utilizaron en los diferentes estudios de esta tesis.
Además, estos análisis se complementaron con medidas derivadas del cálculo de
probabilidades de las configuraciones de movimiento, así como el cálculo del tiempo
promedio en el que las categorías permanecieron activas. La utilización de todas estas
medidas permiten detectar la estructura y propiedades dinámicas del paisaje potencial de
comportamiento táctico de jugadores y equipos (Hristovski et al., 2013; Torrents et al.,
2015).

2.3.1. Análisis Jerárquico de Componentes Principales (hierarchical Principal

Component Analysis)

El análisis jerárquico de componentes principales ordinario o hierarchical principal

component analysis (hPCA) se utilizó con el objetivo de detectar la estructura compleja de
patrones de comportamiento de jugadores y/o equipos. Este análisis se basa en la reducción
de un número elevado de variables originales (configuraciones) a un conjunto de menor
dimensión derivado de la correlación entre los datos iniciales (para la idoneidad de utilizar

26
PCA en variables binarias ver Joliffe, 2002). Los resultados de los componentes principales
(CPs) extraídos se utilizaron para determinar las categorías más destacadas que definieron
cada CP (es decir, el patrón de comportamiento). Las puntuaciones de los CPs se estimaron
como: 𝐶 = 𝑅−1 𝑆 donde R es la matriz de correlación de las variables iniciales
(configuraciones) y S es la matriz de estructura (Fulgosi, 1988). Esta matriz de estructura
permite visualizar la dinámica de evolución de patrones de comportamiento a través de las
correlaciones entre las configuraciones ordenadas en el tiempo y los CPs extraídos. La
principal característica del hPCA es la ausencia del supuesto de ortogonalidad de los CPs.
Por ello se utilizó el método de rotación Oblim Directo con delta = 0 (Westerhuis, Kourti y
MacGregor, 1998) para considerar la posibilidad de una estructura de orden superior en los
datos debido a la correlación entre los componentes extraídos en cada orden. Por lo tanto,
los resultados del primer orden de CP extraídos son tratados como una nueva entrada para
una mayor reducción de la dimensionalidad, estableciendo un orden superior de CP. El
procedimiento continúa hasta el momento en que no se detectan correlaciones significativas
entre los CPs extraídos. El número de CP significativos se determinó por el criterio de
Kaiser-Gutmann, pero únicamente se seleccionaron aquellos que acumularon el 80% de la
varianza explicada (Fabrigar, Wegener, MacCallum y Strahan, 1999).

2.3.2. Análisis de superposición dinámica qd(t) (Dynamic overlap)

Esta medida se puede utilizar como parámetro de orden para determinar las
propiedades dinámicas de comportamiento exploratorio (Hristovski et al., 2013; Torrents et
al., 2015). Captura la similitud promedio de las configuraciones a distancias de tiempo cada
vez mayores entre sí. Por lo tanto es capaz de detectar el ritmo o velocidad, y la amplitud
del comportamiento exploratorio en diferentes escalas temporales. El promedio de
superposición dinámica <qd(t)> se calculó como el coseno de similitud promedio de las
superposiciones o solapamientos entre las configuraciones a diferentes lapsos de tiempo, y
se ajustó a la siguiente ecuación, derivada de sistemas con una estructura jerárquica
intrincada (Sibani & Dall, 2003):

< 𝑞𝑑 (𝑡) > = (1 − 𝑞𝑠𝑡𝑎𝑡 )𝑡 −𝛼 + 𝑞𝑠𝑡𝑎𝑡

27
Donde qstat es el valor asintótico (es decir, estacionario) de la superposición dinámica, t son
las distancias o lapsos de tiempo y α es el exponente dinámico. Mientras que qstat detecta la
amplitud exploratoria de los jugadores, el valor de α captura el ritmo, o velocidad, de
exploración. Si los jugadores realizaran el mismo comportamiento durante todo el tiempo
de observación, la media de la dinámica exploratoria sería constante e igual a uno (<qd(t)>
= 1) para el total de lapsos de tiempo. Sin embargo, si los jugadores desarrollaran
comportamientos completamente diferentes en cada unidad de tiempo durante toda la
práctica, la dinámica exploratoria tendería a cero (<qd(t)> = 0), es decir, todos los
comportamientos serían completamente aleatorios entre sí. El hecho de que la relajación de
la curva teórica no sea exponencial permite detectar la escala temporal en la que el
comportamiento exploratorio se satura, y, por lo tanto, diferenciar dos escalas temporales
distintas dentro de la dinámica de juego.

2.3.3. Análisis complementarios

La amplitud exploratoria se ha definido como la probabilidad de transición entre

estados sobre el total de estados posibles, es decir, a mayor variedad de comportamientos
sobre el total de comportamiento posibles por el jugador/equipo mayor será la amplitud
exploratoria (Hristovski et al., 2011). Formulado matemáticamente sería 𝑊𝑒 = 𝑛/𝑁, donde
n el total de configuraciones distintas y N el número total de configuraciones. Por lo tanto
la inversa (𝑊𝑐 = 1 − 𝑊𝑒) mide la fuerza de atracción de una configuración o estado. Este
valor da información sobre la flexibilidad del sistema en un contexto determinado, ya que
cuanto menor sea al valor de Wc menores serán las barreras que separan los estados de
configuración dentro del espacio de estados y por lo tanto más fácil será la transición entre
ellos.

Por otro lado, la incertidumbre de los comportamientos de los jugadores se calculó

utilizando la entropía de Shannon, S:

𝑆 = − ∑ 𝑝𝑖 ln 𝑝𝑖
𝑖=1

28
donde pi es la probabilidad de cada configuración calculada dividiendo la frecuencia de
cada configuración por la totalidad de las frecuencias (𝑝𝑖 = 𝑛𝑖 /𝑁). Estas probabilidades
nos han posibilitado representar la red de probabilidad de pases entre jugadores, así como
calcular el potencial (cuencas de atracción) del espacio de estados representado a partir de
la combinación de dos variables de comportamiento. Los valores del potencial se
calcularon a partir de la siguiente ecuación (Balescu, 1975):

𝑉𝑖 = 𝑄 ln(𝑝𝑖 ⁄𝑁)

donde Q es la varianza estandarizada del sistema asumiendo que el ruido del sistema es
constante (Q = 1), y N es el número total de configuraciones.

Finalmente, a través de las frecuencias, también se puede calcular el tiempo

promedio, en segundos, en el que las categorías permanecen activas. El tiempo de
persistencia (dwell time) se calculó con el objetivo de identificar las propiedades dinámicas
asociadas al tiempo en que el jugador o el equipo residen en un estado (comportamiento) en
concreto. Esta medida evalúa como ciertas variables y/o categorías permanecen más o
menos estables en un estado antes de cambiar a otro. Por lo tanto será útil para conocer la
velocidad de transición entre comportamientos determinados.

En la tabla 2 se detalla para cada artículo el contexto de juego, parámetro de control,

nivel de organización, parámetro de orden, análisis complementarios y conjunto de
variables que definen el comportamiento.

29
Tabla 2. Contextualización y resumen de los diferentes parámetros y análisis utilizados en los diferentes artículos.

Contexto Parámetro de Nivel de Parámetro de Análisis Relación de variables que definen las
Artículo
de juego control descripción orden complementarios configuraciones de movimiento

Índice de dispersión, velocidad de dispersión

/contracción, amplitud, profundidad, carril y
Set de Tiempo de
Partido Coordinación Componentes sector en el que está localizado el centro
I constreñimientos persistencia (Dwell
(FSG) interpersonal principales geométrico del equipo, velocidad de
del juego time)
movimiento del centro geométrico del
equipo

Fuerza de atracción
(Trapping strength)
Juego
campo Número de Coordinación Dynamic Entropía (Shannon Acción táctica, contexto de interacción, zona
II
reducido oponentes intrapersonal overlap qd(t) entropy) del campo, velocidad
(SSG)
Paisaje potencial de
acción
Componentes
Juego Acciones del jugador con balón,
Número de principales y
campo Coordinación características de la acción con balón,
III oponentes y -
reducido interpersonal acciones de los compañeros sin balón,
compañeros Dynamic
(SSG) acciones de los oponentes
overlap qd(t)
Análisis de redes
Juego
Limitación del Coordinación Acción táctica, distancia al oponente más
campo Dynamic Dwell time
IV espacio de intrapersonal e cercano, distancia a la zona de marca,
amplio overlap qd(t)
interacción interpersonal velocidad
(LSG) Paisaje potencial de
acción

30
 3. Estudio I - Soft-assembled Multilevel
Dynamics of Tactical Behaviors in
Soccer 2

2
Ric, A., Torrents, C., Gonçalves, B., Sampaio, J. y Hristovski, R. (2016). Soft-assembled

Multilevel Dynamics of Tactical Behaviors in Soccer. Frontiers in psychology, 7:1513. doi:

10.3389/fpsyg.2016.01513

31

Edited by:
Duarte Araújo,
University of Lisbon, Portugal
Reviewed by:
Raoul Huys,
Centre national de la recherche
scientifique, France
John Komar,
University of Rouen, France
*Correspondence:
Angel Ric
angelric10@outlook.com
Specialty section:
This article was submitted to
Movement Science and Sport
Psychology,
a section of the journal
Frontiers in Psychology
Received: 04 June 2016
Accepted: 20 September 2016
Published: 05 October 2016
Citation:
Ric A, Torrents C, Gonçalves B,
Sampaio J and Hristovski R (2016)
Soft-assembled Multilevel Dynamics
of Tactical Behaviors in Soccer.
Front. Psychol. 7:1513.
doi: 10.3389/fpsyg.2016.01513
Frontiers in Psychology | www.frontiersin.org
ORIGINAL RESEARCH
published: 05 October 2016
doi: 10.3389/fpsyg.2016.01513
Soft-assembled Multilevel Dynamics
of Tactical Behaviors in Soccer
Angel Ric 1*, Carlota Torrents 1 , Bruno Gonçalves 2 , Jaime Sampaio 2 and Robert Hristovski 3
1
Complex Systems in Sport Research Group, National Institute of Physical Education of Catalonia, University of Lleida,
Lleida, Spain, 2 CreativeLab Research Community, Research Centre in Sports Sciences, Health Sciences and Human
Development, CIDESD, Universidade de Trás-os-Montes e Alto Douro, Vila Real, Portugal, 3 Faculty of Physical Education,
Sport and Health, Saints Cyril and Methodius University, Skopje, Macedonia
This study aimed to identify the tactical patterns and the timescales of variables during
a soccer match, allowing understanding the multilevel organization of tactical behaviors,
and to determine the similarity of patterns performed by different groups of teammates
during the first and second halves. Positional data from 20 professional male soccer
players from the same team were collected using high frequency global positioning
systems (5 Hz). Twenty-nine categories of tactical behaviors were determined from eight
positioning-derived variables creating multivariate binary (Boolean) time-series matrices.
Hierarchical principal component analysis (PCA) was used to identify the multilevel
structure of tactical behaviors. The sequential reduction of each set level of principal
components revealed a sole principal component as the slowest collective variable,
forming the global basin of attraction of tactical patterns during each half of the match. In
addition, the mean dwell time of each positioning-derived variable helped to understand
the multilevel organization of collective tactical behavior during a soccer match. This
approach warrants further investigations to analyze the influence of task constraints
on the emergence of tactical behavior. Furthermore, PCA can help coaches to design
representative training tasks according to those tactical patterns captured during match
competitions and to compare them depending on situational variables.
Keywords: exploratory behavior, interpersonal coordination, complex systems, collective variables, principal
component analysis
INTRODUCTION
In team sports settings, the interaction between team players and environment gives rise to
interpersonal coordination movements that dynamically arise during the game. Interpersonal
interactions are thought to be non-linear and some studies have shown this explicitly (Schmidt
et al., 1990; Richardson et al., 2007). Under conditions of non-linear interpersonal interactions,
behavioral patterns are hypothesized to be spontaneously organized at the macroscopic level
as a result of functional grouping of components which are temporarily assembled, through a
self-organization process (Gréhaigne et al., 1997; McGarry et al., 2002; Araujo et al., 2006). Many
investigations have studied the emergence of coordination patterns on individual (Travassos et al.,
2012), dyadic (Sampaio and Maçãs, 2012) and collective levels of analysis (Silva et al., 2014).
Notwithstanding, there is little literature that attempts to identify the theoretically existing relation
among these levels of game constraints (Bourbousson et al., 2014).
1 October 2016 | Volume 7 | Article 1513
Ric et al.
During soccer matches, there is no particular and fixed
predetermined movement or social coordination to be executed
for the achievement of a performance goal (score a goal) or
sub-goal (avoid conceding one). Hence, there is a lack of
convergence of the exploratory behavior toward a pre-defined
intra- or inter-personal configuration. This situation allows very
subtle and consequently unpredictable interactions between the
environmental information flow and the performer’s organismic
constraints to decide which particular affordance, from the set
of affordances, will be realized at each moment. Gibson (1979)
originally used this term referring to opportunities for action. The
interrelatedness of affordances (Bruineberg and Rietveld, 2014)
allows us to understand the emergence of behavior at different
levels of game constraints. Affordances offered by situational
variables, such as quality or expertise of the opponent, can
constrain ball possession strategies (Lago, 2009). In the same
way, having ball possession far from or close to the opponent’s
goal typically constrains the emergence of specific tactical task
solutions, for example, deep mobility of players while keeping
the ball far from the opponent’s goal, or, choosing a pressing
strategy when the ball is lost in the opponent’s field. All these
interconnected situations are characterized by their evolution
on different timescales (Mendes et al., 2007; Ric et al., 2016).
This study of Ric et al. (2016) showed a clear separation of
dynamical timescales in small-sided games. Slower evolution
was characteristic of ball possesion or dispossesion and quicker
evolution was characteristic for specific actions belonging to the
former states. This state of affairs clearly reflects the model-
independent definition of dynamic property of metastability1 as a
1 Over time several models were proposed that can account for metastability:
Stochastic hopping within the enduring potential landscape (a simple bistable
potential scenario Eyring, 1935; Freidlin and Wentzell, 2012) or a much more
involved quenched disorder scenario (Cugliandolo and Kurchan, 1993), hopping
due to the changing stability of the fixed point attractors (minima become maxima
and vice versa, i.e., an annealed disorder scenario Badii and Politi, 1999), transient
dynamics in a homoclinic channel (Kelso, 1995), and transient dynamics in a stable
heteroclinic channel (Rabinovich et al., 2008) as well as the self-organized criticality
scenario (Bak et al., 1987; Kello et al., 2008).
All these models claim descriptions of metastability. However, they differ.
Quenched and annealed disorder models differ in the timescales ascribed to the
interactions (constraints) among the dynamic degrees of freedom of the system.
Both systems are multistable in the sense of having a vast number of well-defined
fixed point attractors i.e., minima. While in the quenched systems interactions are
assumed to be to a good approximation constant with respect to the dynamical
degrees of freedom, in annealed disorder systems they change on the same
timescale. Hence the difference in the behavior of these systems: enduring vs.
changing potential landscapes. These two models are stochastic in a sense that the
interactions between the dynamic degrees of freedom are random. Real complex
systems, such as teams of players, may even be a combination of both these polar
types, showing both types: constant and quickly changing interactions. This would
produce landscapes with some enduring and some changing basins of attraction.
Contrary to these models, homoclinic and heteroclinic channel metastability are
purely dynamic and no stochastic influence is needed although, as was the case
in the heteroclinic channel scenario, noise, i.e., stochasticity, can be present. The
self-organized criticality type of metastability arises as a domino effect, upsetting
minimally stable states.
On the other hand the data collected on the observational timescale enables
the calculation of the probabilities of configurations. This further enables the
presentation of the configurational probabilities as coming from the well-known
statistical mechanics relation (via the maximum entropy method) between
configuration probabilities and minima depths. Of course, such a depiction of
probabilities, taken on the observational timescale, makes whether this quicker
Frontiers in Psychology | www.frontiersin.org
Soft-Assembled Multilevel Dynamics in Soccer
property related to the existence of multiple separated timescales.
At quick timescales, the system appears to be in equilibrium,
i.e., in a temporarily stable state, but explores a limited part
of its available state space. At longer timescales, however, it
undergoes transitions between such metastable states (Bovier
and Den Hollander, 2016). These previous studies allow us to
hypothesize that the characteristic timescales of team behaviors
could define the interpersonal coordination at different levels of
system organization and help to understand the nested structure
of the dynamics of tactical behaviors.
The dynamical systems approach has established the bases for
understanding the emergence of interpersonal coordination in
team sports requiring the identification of relevant coordination
variables [described as collective variables or modes and
their amplitudes, i.e., order parameters (Haken, 2006)]. These
collective variables describe the spatiotemporal pattern of
interpersonal coordination and the changes of coordination
(Riley et al., 2011) that occur in response to game constraints.
In a team sport setting, the relative phase between two oscillating
components has been frequently used as a collective variable to
capture the macroscopic order or coordination of the dyadic
system. Bourbousson et al. (2010), through the relative phase
between the oscillations of mean player dispersion around the
team geometrical center (i.e., stretch index) of two confronted
basketball teams, found that team players tended to expand and
contract together when both teams were moving from basket to
basket showing clear inter-team coordination patterns. Travassos
et al. (2011) showed that interpersonal coordination between
defenders was stronger than between attackers demonstrating
that attacking players show greater variability in the movement
patterns performed during futsal matches. Gonçalves et al. (2014)
applied relative phase analysis between the mean position of
specific group players (i.e., team geometrical center) over time
during a soccer match. They found that stronger coordination
movements appeared between the consecutive line forces,
that is, defender-midfielders and midfielders-attackers. Other
techniques, commonly used to capture coordination patterns,
like cross-correlations and vector coding, have recently been
studied (Moura et al., 2016).
Whereas these studies focused on the coordination between
two oscillating variables, principal component analysis (PCA) has
dynamics comes from the enduring or changing landscape or any other form of
quick dynamics irrelevant. The enduring landscape scenario, in this case, becomes
a useful tool to represent the dynamics of the configurations (see Hristovski et al.,
2011; Ric et al., 2016). Of course, alternative models may exist as is the case in
many areas of science. Here we find the scenario of hopping dynamics on an
enduring landscape as being the most parsimonious at present, although a mixed
quenched-annealed disordered scenario would be rather more realistic.
Bovier and Den Hollander (2016) define metastability in a model-independent
way as a property related to the existence of multiple separated timescales. At
short time-scales, the system appears to be in equilibrium, but in fact, explores
only a limited part of its available state space. At longer timescales, it undergoes
transitions between such metastable states. This definition satisfies the models
with enduring and changing potential landscapes as well as the homoclinic and
heteroclinic channel models of metastability. However, it does not satisfy the
metastability of the self-organized criticality type in which timescales are not
separated and independent. This example shows that the definition of metastability
is somewhat flexible. Nevertheless, in all of these models there exist at least two
timescales (for convenience called quick and long-term).
2 October 2016 | Volume 7 | Article 1513
Ric et al.
been successfully proposed as a multivariate statistical method
for analyzing high dimensional (i.e., >2) movement coordination
patterns (Daffertshofer et al., 2004; Forner-Cordero et al., 2005).
In essence, PCA reduces the dimensionality of large data sets,
obtaining a smaller number of subjacent components that explain
most of the variance and summarizes the information of the
original variables. PCA has been used in the past three decades
as an efficient method of defining the essential, i.e., collective,
variables in a range of complex dynamical processes such
as protein folding (Matsunaga et al., 2007; Hayward and De
Groot, 2008; Maisuradze et al., 2009) and brain dynamics (Jirsa
et al., 2002). In sport science, PCA has allowed identifying the
performance of an interpersonal precision task (Ramenzoni et al.,
2012) to determine the impact of different training programs
on cardio-respiratory coordination (Balagué et al., 2016), and to
identify different dancing coordination patterns depending on
concrete task constraints (Bronner and Shippen, 2015; Torrents
et al., 2015). In soccer, PCA has been applied to capture player
positional patterns and their variability (Barros et al., 2006).
Moura et al. (2015) used this method by applying it to each
mean player position in the field in order to obtain the collective
organization during the matches of subsequent rounds in the
European Championship and the positional variability of all
players. Despite these investigations, the use of PCA to detect the
pattern-forming dynamics on a collective level over the course of
a competitive match and the relation between the adjacent levels
of game constraints remains unexplored. In this sense, several
authors have emphasized the need to analyze the timescales
of tactical behavior that unequivocally define the multi-level
game dynamics because nested levels in dynamical systems
are inevitably connected with characteristic timescales of their
evolution (Haken, 2006). Therefore, the aim of this study was to
identify the tactical patterns and the timescales of positioning-
derived variables that define the patterns during a soccer match,
allowing understanding the multilevel organization of tactical
behaviors.
MATERIALS AND METHODS
Participants and Procedure
Twenty male professional soccer players from the same team (age
= 22.8 ± 4.4 years; professional playing experience = 6.9 ± 4.1
years) participated in an official pre-season match across the
two halves against another team. Goalkeepers participated in
the match but were excluded from the analysis. Due to the
characteristics of pre-season matches, the ten starter outfield
soccer players participated in the first half. Eight of them
were substituted at halftime and the other two in the middle
of the second half in order to ensure a more constant team
formation and avoid the effects of cumulative fatigue (Mohr
et al., 2005). The match was played on a natural turf pitch
(100 × 68 m) following the official soccer rules. The players
analyzed belonged to the visiting team. The final score of the
match was 1-0. All players provided written informed consent
to participate in the experiment. The local institutional Research
Ethics Committee approved the study, which also conformed to
the recommendations of the Declaration of Helsinki.
Frontiers in Psychology | www.frontiersin.org
Soft-Assembled Multilevel Dynamics in Soccer
Data Collection and Preparation
Positional data from the outfield players were collected using
5 Hz GPS devices (SPI Pro, GPSports, Canberra, Australia).
Each one was placed on the upper back of the players. Latitude
and longitude coordinates were exported from the units and
computed using dedicated routines in Matlab R2014b software
(MathWorks, Inc., Massachusetts, USA; for complete guidelines,
see Folgado et al., 2014).
Configurations of tactical behaviors expressed in team-related
positioning-derived variables were computed to determine the
structural and dynamic characteristics of the team. Eight
collective measures were processed from the outfield players:
stretch index, team length, team width, longitudinal position of
team geometrical center (x axis) and lateral position of team
geometrical center (y axis) (Duarte et al., 2013). Whereas team
geometrical center (also named centroid) and stretch index have
been previously defined, team length, and width were calculated
as the difference between the maximum and minimum positions
of players in the field’s longitudinal and latitudinal dimensions,
respectively, in each time unit. Also included was the speed of
displacement (meters per second) at which team geometrical
center was moving (differentiated lateral and longitudinal axes)
and the speed of contraction and expansion in order to capture
how the team was behaving (Bourbousson et al., 2010). This last
variable was calculated by differencing the each data point of the
stretch index with the previous one.
Data were down sampled into 1 Hz in order to define the team
configuration at each second. A two-step cluster analysis was
performed to determine automatically the boundary values of
each positioning-derived variable. To determine the zone around
which the geometrical center was located, the field was divided
into four sectors and three corridors (Costa et al., 2011; Sarmento
et al., 2013). Finally, 29 categories were determined to create
multivariate binary (Boolean) time series matrices (Table 1). A
value of 1 was ascribed to the active categories and value of 0
to the inactive ones, representing the full configuration (tactical
pattern) during the same time interval of 1 s. Each 1-s window
was defined as a 29-component binary vector (column). A total
of 2700 configurations (45 min × 60 s) for each half were finally
obtained.
Data Analysis
Principal Component Analysis
Hierarchical PCA (hPCA) was performed to define the collective
(state) variables of the coordinated behaviors of teams on
separated levels of organization. The higher levels were
hypothesized to be dominantly defined by categories that possess
larger dwell times. The initial data were multivariate binary
matrices: 29 categories × 2700 time-ordered configurations (for
the suitability of using principal components analysis with binary
variables see Joliffe, 2002). The initial system of PCs was rotated
under the Direct Oblimin method with δ = 0 (Westerhuis
et al., 1998) to consider the possibility of a higher-order
structure in data because of the correlation between the extracted
components in each order. The number of significant principal
components (PCs) was determined by the Kaiser-Gutmann
criterion but only those that cumulatively accounted for ≥80%
3 October 2016 | Volume 7 | Article 1513
Ric et al. Soft-Assembled Multilevel Dynamics in Soccer

TABLE 1 | Categories of each collective positioning-derived variable with solutions. Principal component scores (see Joliffe, 2002) were
the boundary values, identifying a total of 29 categories. used to determine the most salient categories which defined each
1–2: Stretch index <16.75 m Small SIS
principal component (i.e., tactical pattern). Principal component
>16.75 m Large SIL
scores were estimated as: C = R−1 S, where R is the correlation
matrix of time-ordered game configurations and S is the PC
3–6: Speed of spread < −0.6 m/s Quick contraction QC structure matrix, i.e., the correlations between 2700 time-ordered
−0.6 to 0 m/s Slow contraction SC game configurations and PCs (Fulgosi, 1988).
0–0.6 m/s Slow expansion SE The structure matrix was used to visualize the dynamics of
>0.6 m/s Quick expansion QE team configurations in the space spanned by the extracted PCs.
Finally, to compare the structure of first-level PCs between both
7–10: Length <27.72 m Small LS halves, Tucker’s congruence coefficient was used to determine
27.72–35.4 m Medium LM the degree of similarity between principal components (Lorenzo-
35.4–42.71 m Large LL Seva and ten Berge, 2006).
>42.71 m Very large LVL
Analysis of Timescales of Positioning-Derived
11–14: Width <34.49 m Small WS
Variables
34.49–43.68 m Medium WM
The aim of the analysis was to identify the dynamic properties of
43.68–53.59 m Large WL the game assessed by the associated dwell (waiting or residence)
>53.59 m Very large WVL times of positioning-derived variables. Dwell times assess how
long a certain variable remains in a well-defined state before
15–18: Sector <25 m Ultra-defensive UDS
leaving it and switching to another. Hence, they are useful in
25–50 m Mid-defensive MDS
this respect since their averages show the speed of evolution of
50–75 m Mid-offensive MOS
the variable in question. The shorter the average dwell time the
>75 m Ultra-offensive UDS
quicker the evolution (changing the states) and vice versa. The
19–21: Corridor <21.33 m Right RC pooled averages of the active categories (i.e., with 1 ascribed) were
21.33–42.66 Central CC calculated in order to find out the average time, in seconds, that
>42.66 Left LC
the team was dwelling on each positioning-derived variable.
In addition, the video-recorded match was analyzed to
22–25: Longitudinal speed < −1.06 m/s Quick drop back QDB calculate the timescale (i.e., average dwell time) on which ball
of team center possession switched from one team to another. The beginning
−1.06 to 0 m/s Slow drop back SDB of ball possession started when: the goalkeeper took the ball in
0–1.06 m/s Slow forward move SFM his hands, the second touch of the player winning back the ball,
>1.06 m/s Quick forward move QFM or the first touch of any teammate after a pass, deflection, or
clearance by the teammate who first touched the ball (Castellano,
26–29: Lateral speed of < −0.6 m/s Quick to the right QR
2008). It is important to note that when play was stopped due to
team center
an interruption (e.g. corner kick, fouls, goals, etc.) ball possession
−0.6 to 0 m/s Slow to the right SR
was assigned to the team responsible for restarting the game.
0–0.6 m/s Slow to the left SL
Due to non-Gaussian distributions of the dwell times, the
>0.6 m/s Quick to the left QL
non-parametric Kruskal-Wallis test was performed in order to
compare dwell times of variables to identify possible slow-
and fast-evolving processes. It was conducted to compare the
of the explained variance were further selected (Fabrigar et al., timescales of all positioning-derived variables between both
1999). halves. The partial eta square (pη2 ) value is reported as a
The hPCA differs from the ordinary PCA in that it is measure of effect size and is interpreted according to the
free from the assumption of orthogonality of the principal following criteria: significant but weak (ES ≤ 0.04), moderate
components. In other words, it is based on more general (0.04 < ES ≤ 0.36) and strong (ES > 0.36) (Tabachnick and
assumptions of non-orthogonality of correlation/covariance Fidell, 2007). Comparisons between halves were assessed via
matrix eigenvectors (PCs) and treats orthogonality as a special standardized mean differences, computed with pooled variance
case. Just as the ordinary PCA is based on reducing the original and respective 90% confidence intervals. Dwell time data were
high dimensional data to a lower dimensional space based on log-transformed to reduce bias arising from a non-uniformity
correlations/covariance of observed data, the hPCA treats the error. Uncertainty in the differences was expressed as 90%
correlations of the extracted first order PCs as a new input to of confidence limits (CL) and as probabilities that the true
further reduce the dimensionality to fewer and higher-order effect was substantially greater or smaller than the smaller
PCs. The procedure continues until the moment when no practical difference at the threshold of 25% (declared possible).
further meaningful correlations/covariance between the PCs is These probabilities were used to make a qualitative probabilistic
detected. If the first order PCs share little meaningful variance, mechanistic inference about the true effect. The scale was as
the results of hPCA are comparable to the orthogonal PCA follows: 25−75%, possible; 75−95%, likely; 95−99%, very likely;

Frontiers in Psychology | www.frontiersin.org 4 October 2016 | Volume 7 | Article 1513

Ric et al.
>99%, most likely. A difference was assessed as being unclear
if the CI overlapped both substantially positive and negative
thresholds by ≥5%. The Cohen d effect size with 90% CL was
calculated using pooled standard deviation for comparisons and
the magnitude ranges for mean differences were: 0–0.2 trivial; >
0.2–0.6 small; > 0.6–1.2 moderate; > 1.2–2 large; > 2 very large
(Hopkins et al., 2009).
RESULTS
The Primary Level of Tactical Patterns
PCA initially revealed 22 principal components. Twelve of
these represented 80.8 and 80.91% of the total variance for
the first and second half, respectively. The high positive and
negative component scores on these twelve principal components
represented those tactical variables (categories) that occurred
and decayed jointly (Figure 1). The salient structure of PCs was
defined by those categories with a high absolute (positive or
negative) component score. Scores close to 0 indicate that the
corresponding category does not contribute or hardly contributes
to the PC. High positive scores (>1) informs that those categories
are simultaneously active while high negative scores (< − 1)
report that the corresponding categories are active together but
are active when the positive ones are inactive and inversely.
The structures of PCs showed significant congruence between
PC1 of the first half and PC3 of the second (rc = 0.70). They can
be defined as defensive patterns because of the medium width and
FIGURE 1 | Illustration of ten outfield players (numbered gray circles) and the eight positioning-derived variables: (a) stretch index, (b) speed of
dispersion, (c) length, (d) width, (e) sectors, (f) corridors, (g) longitudinal speed of team center, and (h) latitudinal speed of team center.
Frontiers in Psychology | www.frontiersin.org
Soft-Assembled Multilevel Dynamics in Soccer
small stretch index which slowly shrank, with the team centroid
located in middle offensive sector and central corridor while they
were dropping back (see Figure 2). The percentage of the total
variance that explained this pattern in the first part was twice that
of the second. These patterns could describe the main positioning
structure when they were defending in the first half. On the
other hand, the most frequent tactical pattern in the second
half (PC1) had a significant degree of similarity (rc = 0.78)
with PC9 of the first half. They are defined as defensive patterns
and were characterized by a small stretch index, related to the
medium length and small width, with the team centroid located
in the middle defensive sector and right corridor. The players
were slowly reducing their effective playing space and dropping
back in PC1 but this was not clearly defined in PC9. Defending
patterns were the most stable patterns in both halves, but,
whereas in the first half the team was located in the opponent’s
field for defending, in the second they were placed mostly in
their own field. The congruence coefficient between PC2 in both
halves showed a significant similarity between them (rc = 0.73).
These offensive patterns were defined by a large stretch index,
with the team keeping the distances between the players and their
geometrical center mostly stable. The team was moving forward
and quickly to the left, with the team centroid located in central
offensive sector and central corridor. The third PC of the first half
was defined by a small but slowly increasing dispersion of players,
with a long length and medium width. Besides, the location of
the team center was around the ultra-offensive sector and it was
5 October 2016 | Volume 7 | Article 1513
Ric et al. Soft-Assembled Multilevel Dynamics in Soccer

FIGURE 2 | Larger values than 1 (blue dashed line) and/or lower than −1 (red dashed line) refer to the tactical behaviors that defined the primary-level
of principal components.

Frontiers in Psychology | www.frontiersin.org 6 October 2016 | Volume 7 | Article 1513

Ric et al.
slowly moving forward to the right showing the dominance of
the game in the opponent’s field. Its highest congruence value
(rc = 0.61) is related with PC8 of the second half, explaining
two similar offensive patterns for each half. In addition, PCs also
showed significant congruence coefficients: PC12 of the first half
with PC11 of the second (rc = 0.70) PC11 with PC10 (rc = 0.76)
and PC10 with PC5 (rc = 0.79). The rest of combinations did not
reach coefficients of congruence higher than 0.70, however their
structure can be consulted in Figure 1.
Metastable Dynamics of Tactical Patterns
Figure 3 represents the loadings (correlations) of the time
configuration vector on/with the PCs. Note that the configuration
vector dwells for some time projecting dominantly on a few PCs
and then, on a longer timescale, quickly transits to form another
temporarily stable set of dominant projections. These sets of
dominant projections of configurations on the PCs represent the
attractors of the team dynamics. The dominance of projections
refers to the dominant PC content of the configurations, while the
dwell time of the dominant projection to its stability or attraction
strength (the longer the dwell time the greater the stability or
attraction strength and vice versa).
As can be seen in Figure 3, during the first 5 min the system
was projecting mostly on PC1, PC5, and PC9 defined by a small
stretch index that slowly contracted and team positioning was
slowly dropping back. Although the locations of the team center
were in different sectors, all these patterns were defensive. Then,
the switches between these PCs at the beginning of the match
were brought about by the small differences in some of the
categories that defined each one of these PCs. It is interesting
to note that the dominant projection of the game configuration
when the goal was scored (second 1148) corresponds to PC10.
The small dispersion of players was not altered much more,
but the team center quickly moved back to the right side from
the central corridor of the middle defensive sector. That pattern
clearly defines the defense of the opponent’s counterattack.
After the goal, the game configuration transited to a dominant
projection on PC11 defined by the large dispersion, slowly
growing up from the middle of the field quickly moving to
the opponent’s midfield. Finally, the team was mostly stable
performing PC2, previously defined, for 7 min before transiting
to another less stable tactical pattern.
Multilevel Organization of Tactical Patterns
First-level PCs correlations were then subjected to a further
higher-order analysis revealing second-level PCs. Four PCs in
the first half of the match and five in the second half were
identified. Furthermore, the second-level correlated structures
systematically produced a sole PC on the third-level for the first
half (see black bars in Figure 4), whereas two PCs were extracted
from the second half. Therefore, by a further iteration of the
procedure a fourth-level PC was extracted only in the second
half (see gray bars in Figure 4). The lower-order PCs are more
sensitive to detailed changes in impinging game constraints,
while the highest-level PC (third-level PC in the first half and
fourth-level PC in the second) captures the most robust and
stable structure of associations within the data, defining the most
Frontiers in Psychology | www.frontiersin.org
Soft-Assembled Multilevel Dynamics in Soccer
persistent patterns over time. The congruence coefficients of
these highest-level PCs reached a value of 0.69.
The Kruskal-Wallis test showed a significant effect between
the timescales of all positional variables including ball possession:
[H (8, N = 3156) = 845.8; X 2(8) = 513.51; P < 0.0001; pη2 =
0.314] for the first half and [H (8, N = 3257) = 911.66; X 2(8) =
527.63; P < 0.0001; pη2 = 0.324] for the second. Table 2 shows
means and standard deviations of timescales for each variable
differentiating both halves. The probability of finding true
differences between halves was possible for timescales of stretch
index and ball possession, with a positive and negative effect,
respectively. An unclear tendency was reported for the dwell
time of the centroids allocated in the corridors. Finally, trivial-to-
unlikely differences between the first and second half were found
in the timescales for the remaining collective positional variables.
Multiple comparisons between the means of all eight variables
(see Table 3) allowed identifying three groups of variables that
possessed different characteristic timescales. Those variables that
evolved over the timescales of tens of seconds, i.e., stretch index,
sectors, and corridors, showed statistical differences with all other
variables, whereas no significant differences were detected when
compared between them (stretch index vs. sectors p = 1.00;
stretch index vs. corridors p = 1.00; sectors vs. corridors p = 1.00
for the first half and p = 1.00; p = 1.00; p = 1.00 for the second).
Similarly, significant differences were absent in comparisons of
those variables that quickly evolved, i.e., over the timescale of a
few seconds. For the first half: speed dispersion vs. longitudinal
speed p = 0.05, speed dispersion vs. lateral speed p = 0.272
and longitudinal vs. lateral speed p = 1.00; and for the second
half: speed dispersion vs. longitudinal speed p = 0.11; speed
dispersion vs. lateral speed p = 1.00 and longitudinal vs. lateral
speed 8 p = 1.00. The significant differences of length and width
variables with the rest but not between them (for the first half
p = 0.272 and for the second half p = 0.375), lead us to think
that another intermediate timescale, which evolved over several
seconds, could exist.
DISCUSSION
The current study explores the soft-assembly of tactical patterns
and the timescales of positioning-derived variables that define
them during a soccer match, allowing understanding the
multilevel organization of tactical behaviors as defined by the
timescales of evolution of collective patterns. For this goal,
a hPCA and the dwell time measure were used to identify,
respectively, nested correlated movement configurations as well
as the characteristic timescales of their change (Hristovski
et al., 2013). The main results afford that most stable/persistent
movement patterns were well-related to defensive behaviors. In
fact, team contraction, expressed through lower stretch index
values and the drop-back of the team, are suggested as being key
indicators of team behavior in the defensive phase (Bourbousson
et al., 2010; Frencken et al., 2011). These results do not indicate
that the team performed more defensive patterns, rather the
observed patterns are likely to be most stable (Travassos et al.,
2011; Ric et al., 2016). The degree of similarity between the PC2
7 October 2016 | Volume 7 | Article 1513
Ric et al.
FIGURE 3 | Correlations (loadings) of time-ordered game configurations with the first-level PCs.
extracted for both halves elucidates that the offensive playing style
of the team is clearly defined. In this sense, it can be said that
independently of the players, the playing style proposed by the
coach strongly constrained the emergence of concrete tactical
behaviors. These patterns, which explain the larger percentage
of variance, describe collective behavior when the team or the
opponent perform long ball possessions, by stabilizing their
behavior and during which small reconfigurations take place.
Notwithstanding, some PCs also exhibit some patterns that
corresponded to game transitions (losing or winning back the
ball; Sarmento et al., 2013). For example, PC12 in the first half and
PC5 in the second defined counterattacks by the team analyzed.
The small stretch index was quickly expanded and the team was
moving rapidly forward starting from the mid-defensive sector.
However, it is possible to detect some tactical patterns defined by
a pressing strategy to win back the ball by using player distances
from their team geometrical center and its depth in respect of
goal (Frias and Duarte, 2014). In this hypothetical case, the team
geometrical center would move forward while team dispersion
would reduce in size. This structure of PCs could be related with
PC7 and PC8 in the first half and with PC6 and PC9 in the second.
Previous findings revealed that depending on the situational
context of the match the teams switch between different
functional states (Frencken et al., 2012). The results have
shown those of specific tactical patterns brought about by
the influence of key events, like goal-scoring (Lago-Peñas and
Dellal, 2010). Figure 2 allows identifying the types of patterns
that appeared at different game moments. In this sense, the
goal was preceded by a counterattack by the opposing team.
Consequently, coaches should consider training specifically those
patterns which might lead to conceding/scoring a goal. Even so,
such defensive patterns are obviously preceded by an attacking
phase, then ball recovery strategies and the context in which this
Frontiers in Psychology | www.frontiersin.org
Soft-Assembled Multilevel Dynamics in Soccer
would happen must be considered in the training task designed
(Barreira et al., 2014).
In this study the hypothesized soft-assembled multilevel
dynamics (Hristovski, 2012; Hristovski et al., 2013) in a soccer
match have been corroborated. The salient correlated patterns
(PCs) in the first level of analysis revealed tactical behaviors
on higher levels. This finding helps to understand the nested
organization of tactical behavior. At the highest level, a sole
collective variable (PC) for each half exhibits the essence of
tactics. The levels evolve on different timescales and only the
tactical variables that evolved over longer timescales significantly
contributed to the structure of highest-order PCs. That is,
the speed of team geometrical center evolves on a scale of
seconds, the stretch index (player dispersion) remains below
16.75 m during more than 20 s on average (timescales of tens of
seconds), before switching expansion, and vice versa. Therefore,
the hierarchy is a consequence of the correlated lower-order
PCs. The most time-persistent categories (those with long dwell
times) create correlations between the lower-order PCs. This was
revealed in the highest order PCs where two out of three most
time-persistent categories were those that had by far the highest
scores. In summary, whereas the highest level that captures the
essentials of team tactical behavior corresponds to slower changes
(a few tens of seconds), the lowest order of PCs quickly evolve,
being more susceptible to the sensitive changes of constraint-
induced actions. It is important to note that this emergence
of tactical behaviors in soccer as in other team sports results
from reciprocal influences (bottom-up, top-down) of performer-
environment interactions on different timescales (Hristovski
et al., 2011).
Although previous studies have demonstrated that the average
values of these positioning-derived variables were significantly
different between both halves and also dependent on ball
8 October 2016 | Volume 7 | Article 1513
Ric et al. Soft-Assembled Multilevel Dynamics in Soccer

FIGURE 4 | Component scores of the highest level PC for the first and second half.

TABLE 2 | Mean values ± SD for different collective positional variables and ball possession and corresponding difference comparisons between first
and second half.

Mean ± SD (n) Differences observed for the first half compared with second

First half Second half Difference in Chances* Qualitative Effect size

means (%) assessment

Stretch index 23.94 ± 22.70 (111) 26.44 ± 25.96 (102) 15.36; ± 25.97 34/65/1 Possibly ↑ 0.14; ± 0.23
Speed of dispersion 3.37 ± 3.32 (800) 3.24 ± 2.93 (833) −2.48; ± 6.08 0/100/0 Most likely trivial −0.03; ± 0.08
Length 7.27 ± 7.56 (370) 7.55 ± 8.46 (357) 2.12; ± 11.21 1/99/0 Very likely trivial 0.02; ± 0.12
Width 9.20 ± 9.10 (289) 9.27 ± 9.52 (291) 0.18; ± 12.57 1/98/1 Very likely trivial 0.00; ± 0.14
Sectors 22.05 ± 18.21 (119) 24.37 ± 25.31 (110) −3.27; ± 22.53 4/86/10 Likely trivial −0.03; ± 0.22
Corridors 37.10 ± 47.07 (73) 29.96 ± 32.69 (90) −0.68; ± 33.40 10/79/11 Unclear −0.01; ± 0.26
Longitudinal speed of team center 4.50 ± 5.03 (598) 3.98 ± 3.96 (678) −5.34; ± 7.71 0/99/1 Very likely trivial −0.06; ± 0.09
Lateral speed of team center 3.82 ± 3.84 (707) 3.89 ± 3.96 (693) 1.38; ± 7.67 0/100/0 Most likely trivial 0.02; ± 0.09
Ball possession 30.30 ± 32.30 (89) 26.19 ± 25.91 (103) −11.02; ± 22.82 2/72/27 Possibly ↓ −0.11; ± 0.24

*Percentage chance of having substantially positive/trivial/substantially negative effect.

↑, Increase; ↓, decrease.

Frontiers in Psychology | www.frontiersin.org 9 October 2016 | Volume 7 | Article 1513

Ric et al. Soft-Assembled Multilevel Dynamics in Soccer

TABLE 3 | Multiple Comparisons p-values (2-tailed) and Cohen’s d results for (a) the first half and (b) the second half.

Positioning-derived variables (a) First half (b) Second half

1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8

Stretch index 1 — 1 —
Speed of dispersion 2 *vl — 2 *vl —
Length 3 *l *m — 3 *l *m —
Width 4 *l *m — 4 *l *m —
Sectors 5 *vl *l *m — 5 *vl *l *m —
Corridors 6 *vl *l *l — 6 *vl *l *l —
Longitudinal speed of team center 7 *l *s *s *m *vl *vl — 7 *vl *s *m *vl *vl —
Lateral speed of team center 8 *vl *m *m *vl *vl — 8 *vl *m *m *vl *vl —

–, Diagonal cell. *p < 0.05. Letters denote the magnitude: s, small; m, moderate; l, large; vl, very large.

possession (Clemente et al., 2013), the dwell times of most of
these variables were not significantly different comparing the first
and second halves. This invariance in the timescales has given rise
to the hierarchical structure of tactical patterns in both halves
being mostly the same. However, the additional level identified
in the second half suggests that tactical patterns carried out by
the second group of teammates performed more varied behaviors
because of the less correlated behaviors in the second-level of
principal components. In spite of this, the principal component
extracted in the highest-level of both halves reached a high value
of similarity. This congruence in score distribution suggests that
independently of the group of teammates, the essence of team
tactics was similar during the match.
These findings open the way to analyzing how teams behave
during a competition and studying the influence of different
constraints (e.g., score, substitutions, instructional constraints,
opponent level, etc.) over the course of the match. Furthermore,
due to the limitation of the unknown opponent and ball position,
more research is needed to ascertain the soft-assemble action
landscape of different teams and to compare the collective
tactical patterns of two confronted teams. The positioning of
the opponent would allow analyzing the coupling between the
teams’ behavior, and to detect if some of them lead pattern-
forming. The results should be balanced considering that only
one match was used to capture the tactical behaviors. Despite this
limitation, this study leaves an open path to exploring match-to-
match invariant behaviors. In addition, this approach warrants
further investigations on using ecological task constraints
during training to develop the potential landscape of tactical
patterns allowing the spontaneous emergence of novel modes
of coordination and/or specific tactical performance solutions.
This study also provides the possibility to determine relevant
timescales for scoring goals and shooting at goal using a large
goal-scoring sample.
CONCLUSION
This approach can help to identify tactical patterns during
different matches, comparing the degree of similarity between
them. Further analysis would allow determining the influence
on team tactical behaviors of different situational variables, e.g.,
rank of opponent, score of the previous match, playing home or
away, which remain invariant during a longer timescale, i.e., days.
Moreover, this analysis can help coaches to verify if the essence
of tactics or playing style performed during training sessions
is definitively developed during competition and if it remains
invariant from match to match. Moreover, it would allow coaches
to optimize training drills developing their style of play and/or
identify the tactical behaviors performed during the match to
design a representative training task. Finally, the characteristic
timescales of collective behaviors allow understanding the
formation of the hierarchically nested structure of tactical
patterns in an ecological context and presents a rationale to define
soft-assemble multilevel dynamics in soccer matches.
AUTHOR CONTRIBUTIONS
AR worked on the design of the study, collection, analysis
and interpretation of data, and drafting the manuscript. CT
participated in the conceptualization and design of the study,
data collection and reviewed the content of the manuscript. BG
worked on the design of the study, data collection, statistical
analysis, and reviewed the content of the manuscript. JS
participated in the conceptualization and design of the study
and reviewed the content of the manuscript. RH conceived the
approach to data analysis, data interpretation, and drafting of the
manuscript. The authors approved the final version and agree to
be accountable for all aspects of the work.
FUNDING
This study is supported by the National Institute of Physical
Education of Catalonia, Generalitat de Catalunya.
ACKNOWLEDGMENTS
We would like to thank the players who volunteered to
participate in this study. We gratefully acknowledge the support
of the Generalitat de Catalunya government project Grup de
recerca en Sistemes Complexos i Esport (2014 SGR 975). We
would also thank the reviewers for their thorough review that
helped us a lot to improve the present manuscript.

Frontiers in Psychology | www.frontiersin.org 10 October 2016 | Volume 7 | Article 1513

Ric et al.
REFERENCES
Araujo, D., Davids, K., and Hristovski, R. (2006). The ecological dynamics
of decision making in sport. Psychol. Sport Exerc. 7, 653–676. doi:
10.1016/j.psychsport.2006.07.002
Badii, R., and Politi, A. (1999). Complexity: Hierarchical Structures and Scaling in
Physics. Cambridge: Cambridge University Press.
Bak, P., Tang, C., and Wiesenfeld, K. (1987). Self-organized criticality:
an explanation of the 1/f noise. Phys. Rev. Lett. 59, 381–384. doi:
10.1103/PhysRevLett.59.381
Balagué, N., González, J., Javierre, C., Hristovski, R., Aragonés, D., Álamo, J.,
et al. (2016). Cardiorespiratory coordination after training and detraining.
A principal component analysis approach. Front. Physiol. 7:35. doi:
10.3389/fphys.2016.00035
Barreira, D., Garganta, J., Guimarães, P., Machado, J. C., and Anguera, M. T.
(2014). Ball recovery patterns as a performance indicator in elite soccer.
Proc. Inst. Mech. Eng. Part P J. Sports Eng. Technol. 228, 61–72. doi:
10.1177/1754337113493083
Barros, R. M. L., Cunha, S. A., Magalhaes, W. J., and Guimarães, M. F.
(2006). Representation and analysis of soccer players’ actions using principal
components. J. Hum. Mov. Stud. 51, 103–116. Available online at: http://hdl.
handle.net/11449/31748
Bourbousson, J., Deschamps, T., and Travassos, B. (2014). From players to teams:
towards a multi-level approach of game constraints in team sports. Int. J. Sports
Sci. Coach. 9, 1393–1406. doi: 10.1260/1747-9541.9.6.1393
Bourbousson, J., Seve, C., and McGarry, T. (2010). Space–time coordination
dynamics in basketball: part 2. The interaction between the two teams. J. Sports
Sci. 28, 349–358. doi: 10.1080/02640410903503640
Bovier, A., and Den Hollander, F. (2016). Metastability: A Potential-Theoretic
Approach. Berlin: Springer International Publishing.
Bronner, S., and Shippen, J. (2015). Biomechanical metrics of aesthetic perception
in dance. Exp. Brain Res. 233, 3565–3581. doi: 10.1007/s00221-015-4424-4
Bruineberg, J., and Rietveld, E. (2014). Self-organization, free energy minimization,
and optimal grip on a field of affordances. Front. Hum. Neurosci. 8:599. doi:
10.3389/fnhum.2014.00599
Castellano, J. (2008). Análisis de las posesiones de balón en fútbol : frecuencia,
duración y transición. Mot. Eur. J. Hum. Movemen. 21, 179–196.
Clemente, F. M., Couceiro, M. S., Martins, F. M. L., Mendes, R., and Figueiredo,
A. J. (2013). Measuring collective behaviour in football teams: inspecting the
impact of each half of the match on ball possession. Int. J. Perform. Anal. Sport
13, 678–689.
Costa, I. T., Garganta, J., Greco, P. J., Mesquita, I., and Maia, J. (2011). System
of tactical assessment in Soccer (FUT-SAT): development and preliminary
validation. System 7, 69–83.
Cugliandolo, L. F., and Kurchan, J. (1993). Analytical solution of the off-
equilibrium dynamics of a long-range spin-glass model. Phys. Rev. Lett. 71,
173–176. doi: 10.1103/PhysRevLett.71.173
Daffertshofer, A., Lamoth, C. J. C., Meijer, O. G., and Beek, P. J. (2004). PCA in
studying coordination and variability: a tutorial. Clin. Biomech. 19, 415–428.
doi: 10.1016/j.clinbiomech.2004.01.005
Duarte, R., Araújo, D., Folgado, H., Esteves, P., Marques, P., and Davids, K. (2013).
Capturing complex, non-linear team behaviours during competitive football
performance. J. Syst. Sci. Complex. 26, 62–72. doi: 10.1007/s11424-013-2290-3
Eyring, H. (1935). The activated complex in chemical reactions. J. Chem. Phys. 3,
107–115. doi: 10.1063/1.1749604
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., and Strahan, E. J. (1999).
Evaluating the use of exploratory factor analysis in psychological research.
Psychol. Methods 4, 272–299. doi: 10.1037/1082-989X.4.3.272
Folgado, H., Duarte, R., Fernandes, O., and Sampaio, J. (2014). Competing with
lower level opponents decreases intra-team movement synchronization and
time-motion demands during pre-season soccer matches. PLoS ONE 9:e97145.
doi: 10.1371/journal.pone.0097145
Forner-Cordero, A., Levin, O., Li, Y., and Swinnen, S. P. (2005). Principal
component analysis of complex multijoint coordinative movements. Biol.
Cybern. 93, 63–78. doi: 10.1007/s00422-005-0582-y
Freidlin, M. I., and Wentzell, A. D. (2012). Random Perturbations of Dynamical
Systems, 3rd Edn. Berlin; Heidelberg: Springer.
Frontiers in Psychology | www.frontiersin.org
Soft-Assembled Multilevel Dynamics in Soccer
Frencken, W., Lemmink, K., Delleman, N., and Visscher, C. (2011). Oscillations of
centroid position and surface area of soccer teams in small-sided games. Eur. J.
Sport Sci. 11, 215–223. doi: 10.1080/17461391.2010.499967
Frencken, W., Poel, H., De Visscher, C., and Lemmink, K. (2012). Variability of
inter-team distances associated with match events in elite-standard soccer. J.
Sports Sci. 30, 1207–1213. doi: 10.1080/02640414.2012.703783
Frias, T., and Duarte, R. (2014). Man-to-man or zone defense? Measuring team
dispersion behaviors in small-sided soccer games. Trends Sport Sci. 3, 135–144.
Fulgosi, A. (1988). Faktorska Analiza. Zagreb: Školska Knjiga.
Gibson, J. J. (1979). The Ecological Approach to Visual Perception. Boston, MA:
Houghton, Mifflin and Company.
Gonçalves, B. V., Figueira, B. E., Maçãs, V., and Sampaio, J. (2014). Effect
of player position on movement behaviour, physical and physiological
performances during an 11-a-side football game. J. Sports Sci. 32, 191–199. doi:
10.1080/02640414.2013.816761
Gréhaigne, J. F., Bouthier, D., and David, B. (1997). Dynamic-system analysis of
opponent relationships in collective actions in soccer. J. Sports Sci. 15, 137–149.
doi: 10.1080/026404197367416
Haken, H. (2006). Information and Self-Organization: A Macroscopic Approach to
Complex Systems, 3rd Edn. Heidelberg: Springer Verlag.
Hayward, S., and De Groot, B. L. (2008). Normal modes and essential dynamics.
Methods Mol. Biol. 443, 89–106. doi: 10.1007/978-1-59745-177-2-5
Hopkins, W. G., Marshall, S. W., Batterham, A. M., and Hanin, J. (2009).
Progressive statistics for studies in sports medicine and exercise science. Med.
Sci. Sports Exerc. 41, 3–12. doi: 10.1249/MSS.0b013e31818cb278
Hristovski, R. (2012). Sport performance as a domain of creative problem solving
for self-organizing performer-environment systems. Open Sports Sci. J. 5,
26–35. doi: 10.2174/1875399X01205010026
Hristovski, R., Davids, K., Araújo, D., and Passos, P. (2011). Constraints-induced
emergence of functional novelty in complex neurobiological systems: a basis for
creativity in sport. Nonlinear Dynamics. Psychol. Life Sci. 15, 175–206.
Hristovski, R., Davids, K., Araújo, D., Passos, P., Torrents, C., Aceski, A., et al.
(2013). “Creativity in sport and dance: ecological dynamics on a hierarchically
soft-assembled perception-action landscape,” in Complex Systems in Sport, eds
K. Davids, R. Hristovski, D. Araújo, N. Balagué, C. Button, and P. Passos
(London: Routledge), 261–274.
Jirsa, V. K., Jantzen, K. J., Fuchs, A., and Kelso, J. A. S. (2002). Spatiotemporal
forward solution of the EEG and MEG using network modeling. IEEE Trans.
Med. Imaging. 21, 493–504. doi: 10.1109/TMI.2002.1009385
Joliffe, I. T. (2002). Principal Component Analysis, 2nd edn. New York, NY:
Springer.
Kello, C. T., Anderson, G. G., Holden, J. G., and Van Orden, G. C. (2008). The
pervasiveness of 1/f scaling in speech reflects the metastable basis of cognition.
Cogn. Sci. 32, 1217–1231. doi: 10.1080/03640210801944898
Kelso, J. A. S. (1995). Dynamic patterns: The Self-Organization of Brain and
Behavior. Complex adaptive systems series. Cambridge, MA: MIT Press,
Bradford Book.
Lago, C. (2009). The influence of match location, quality of opposition, and match
status on possession strategies in professional association football. J. Sports Sci.
27, 1463–1469. doi: 10.1080/02640410903131681
Lago-Peñas, C., and Dellal, A. (2010). Ball possession strategies in elite soccer
according to the evolution of the match-score: the influence of situational
variables. J. Hum. Kinet. 25, 93–100. doi: 10.2478/v10078-010-0036-z
Lorenzo-Seva, U., and ten Berge, J. M. F. (2006). Tucker’s congruence coefficient
as a meaningful index of factor similarity. Methodology 2, 57–64. doi:
10.1027/1614-2241.2.2.57
Maisuradze, G. G., Liwo, A., and Scheraga, H. A. (2009). Principal component
analysis for protein folding dynamics. J. Mol. Biol. 385, 312–329. doi:
10.1016/j.jmb.2008.10.018
Matsunaga, Y., Li, C.-B., and Komatsuzaki, T. (2007). Anomalous diffusion in
folding dynamics of minimalist protein landscape. Phys. Rev. Lett. 99:238103.
doi: 10.1103/PhysRevLett.99.238103
McGarry, T., Anderson, D. I., Wallace, S. A., Hughes, M. D., and Franks, I. M.
(2002). Sport competition as a dynamical self-organizing system. J. Sports Sci.
20, 771–781. doi: 10.1080/026404102320675620
Mendes, R. S., Malacarne, L. C., and Anteneodo, C. (2007). Statistics of football
dynamics. Eur. Phys. J. B 57, 357–363. doi: 10.1140/epjb/e2007-00177-4
11 October 2016 | Volume 7 | Article 1513
Ric et al.
Mohr, M., Krustrup, P., and Bangsbo, J. (2005). Fatigue in soccer: a brief review. J.
Sports Sci. 23, 593–599. doi: 10.1080/02640410400021286
Moura, F. A., Santana, J. E., Vieira, N. A., Santiago, P. R. P., and Cunha, S.
A. (2015). Analysis of soccer players’ positional variability during the 2012
UEFA European Championship: a case study. J. Hum. Kinet. 47, 225–236. doi:
10.1515/hukin-2015-0078
Moura, F. A., van Emmerik, R. E. A., Santana, J. E., Martins, L. E. B., de Barros, R.
M. L., and Cunha, S. A. (2016). Coordination analysis of players’ distribution in
football using cross-correlation and vector coding techniques. J. Sports Sci. doi:
10.1080/02640414.2016.1173222. [Epub ahead of print].
Rabinovich, M. I., Huerta, R., Varona, P., and Afraimovich, V. S. (2008). Transient
cognitive dynamics, metastability, and decision making. PLoS Comput. Biol.
4:e1000072. doi: 10.1371/journal.pcbi.1000072
Ramenzoni, V. C., Riley, M. A., Shockley, K., and Baker, A. A. (2012). Interpersonal
and intrapersonal coordinative modes for joint and single task performance.
Hum. Mov. Sci. 31, 1253–1267. doi: 10.1016/j.humov.2011.12.004
Ric, A., Hristovski, R., Gonçalves, B., Torres, L., Sampaio, J., and Torrents, C.
(2016). Timescales for exploratory tactical behaviour in football small-sided
games. J. Sports Sci. 34, 1723–1730. doi: 10.1080/02640414.2015.1136068
Richardson, M. J., Marsh, K. L., Isenhower, R. W., Goodman, J. R. L., and
Schmidt, R. C. (2007). Rocking together: dynamics of intentional and
unintentional interpersonal coordination. Hum. Mov. Sci. 26, 867–891. doi:
10.1016/j.humov.2007.07.002
Riley, M. A., Richardson, M. J., Shockley, K., and Ramenzoni, V. C. (2011).
Interpersonal synergies. Front. Psychol. 2:38. doi: 10.3389/fpsyg.2011.00038
Sampaio, J. E., and Maçãs, V. M. (2012). Measuring tactical behaviour in football.
Int. J. Sports Med. 33, 395–401. doi: 10.1055/s-0031-1301320
Sarmento, H., Barbosa, A., Anguera, M. T., Campaniço, J., and Leitão, J. (2013).
“Regular patterns of play in the counterattack of the FC Barcelona and
Manchester United football teams,” in Performance Analysis of Sport IX, eds
D. Peters and P. O’Donoghue (London: Routledge), 59–106.
Schmidt, R. C., Carello, C., and Turvey, M. T. (1990). Phase transitions and
critical fluctuations in the visual coordination of rhythmic movements between
Frontiers in Psychology | www.frontiersin.org
Soft-Assembled Multilevel Dynamics in Soccer
people. J. Exp. Psychol. Hum. Percept. Perform. 16, 227–247. doi: 10.1037/0096-
1523.16.2.227
Silva, P., Travassos, B., Vilar, L., Aguiar, P., Davids, K., Araújo, D., et al.
(2014). Numerical relations and skill level constrain co-adaptive behaviors of
agents in sports teams. PLoS ONE 9:e107112. doi: 10.1371/journal.pone.01
07112
Tabachnick, B. G., and Fidell, L. S. (2007). Multivariate analysis of variance and
covariance. Using Multivar. Stat. 3, 402–407.
Torrents, C., Ric, A., and Hristovski, R. (2015). Creativity and emergence of specific
dance movements using instructional constraints. Psychol. Aesthetics Creat.
Arts 9, 65–74. doi: 10.1037/a0038706
Travassos, B., Araújo, D., Davids, K., Vilar, L., Esteves, P., and Vanda, C.
(2012). Informational constraints shape emergent functional behaviours during
performance of interceptive actions in team sports. Psychol. Sport Exerc. 13,
216–223. doi: 10.1016/j.psychsport.2011.11.009
Travassos, B., Araújo, D., Vilar, L., and McGarry, T. (2011). Interpersonal
coordination and ball dynamics in futsal (indoor football). Hum. Mov. Sci. 30,
1245–1259. doi: 10.1016/j.humov.2011.04.003
Westerhuis, J. A., Kourti, T., and MacGregor, J. F. (1998). Analysis of
multiblock and hierarchical PCA and PLS models. J. Chemom. 12,
301–321.
Conflict of Interest Statement: The authors declare that the research was
conducted in the absence of any commercial or financial relationships that could
be construed as a potential conflict of interest.
Copyright © 2016 Ric, Torrents, Gonçalves, Sampaio and Hristovski. This is an
open-access article distributed under the terms of the Creative Commons Attribution
License (CC BY). The use, distribution or reproduction in other forums is permitted,
provided the original author(s) or licensor are credited and that the original
publication in this journal is cited, in accordance with accepted academic practice.
No use, distribution or reproduction is permitted which does not comply with these
terms.
12 October 2016 | Volume 7 | Article 1513
 4. Estudio II - Timescales for exploratory
tactical behaviour in football small-sided
games 3

3
Ric, A., Hristovski, R., Gonçalves, B., Torres, L., Sampaio, J. y Torrents, C. (2016).

Timescales for exploratory tactical behaviour in football small-sided games. Journal of

sports sciences, 34(18), 1723-1730. doi: 10.1080/02640414.2015.1136068

44
 5. Estudio III - Emergence of Exploratory,
Technical and Tactical Behavior in
Small-Sided Soccer Games when
Manipulating the Number of Teammates
and Opponents 4

4
Torrents, C., Ric, A., Hristovski, R., Torres-Ronda, L., Vicente, E. y Sampaio, J. (2016).

Emergence of Exploratory, Technical and Tactical Behavior in Small-Sided Soccer Games

when Manipulating the Number of Teammates and Opponents. PloS ONE, 11(12),

e0168866. doi: 10.1371/journal.pone.0168866

53

a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Torrents C, Ric A, Hristovski R, Torres-
Ronda L, Vicente E, Sampaio J (2016) Emergence
of Exploratory, Technical and Tactical Behavior in
Small-Sided Soccer Games when Manipulating the
Number of Teammates and Opponents. PLoS ONE
11(12): e0168866. doi:10.1371/journal.
pone.0168866
Editor: Sergio Pellis, University of Lethbridge,
CANADA
Received: June 8, 2016
Accepted: December 7, 2016
Published: December 22, 2016
Copyright: © 2016 Torrents et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All data files are
available from the Open Science Framework
database osf.io/8weh9.
Funding: This study was supported by the National
Institute of Physical Education of Catalonia
(INEFC), Generalitat de Catalunya.
Competing Interests: The authors have declared
that no competing interests exist.
PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016
RESEARCH ARTICLE
Emergence of Exploratory, Technical and
Tactical Behavior in Small-Sided Soccer
Games when Manipulating the Number of
Teammates and Opponents
Carlota Torrents1*, Angel Ric1, Robert Hristovski2, Lorena Torres-Ronda3, Emili Vicente1,
Jaime Sampaio4,5
1 National Institute of Physical Education of Catalonia (INEFC), University of Lleida, Lleida, Spain, 2 Sts.
Cyril and Methodius University, Skopje, Macedonia, 3 Department of Health and Kinesiology, Exercise and
Sport Nutrition Lab, Texas A&M University, College Station, Texas, United States of America, 4 CreativeLab,
Research Center in Sports Sciences, Health Sciences and Human Development (CIDESD), Vila Real,
Portugal, 5 Sport Sciences Department, Universidade de Trás-Os-Montes e Alto Douro, Vila Real, Portugal
* ctorrents@inefc.es
Abstract
The effects that different constraints have on the exploratory behavior, measured by the
variety and quantity of different responses within a game situation, is of the utmost impor-
tance for successful performance in team sports. The aim of this study was to determine
how the number of teammates and opponents affects the exploratory behavior of both pro-
fessional and amateur players in small-sided soccer games. Twenty-two professional (age
25.6 ± 4.9 years) and 22 amateur (age 23.1 ± 0.7 years) male soccer players played three
small-sided game formats (4 vs. 3, 4 vs. 5, and 4 vs. 7). These trials were video-recorded
and a systematic observation instrument was used to notate the actions, which were subse-
quently analyzed by means of a principal component analysis and the dynamic overlap
order parameter (measure to identify the rate and breadth of exploratory behavior on differ-
ent time scales). Results revealed that a higher the number of opponents required for more
frequent ball controls. Moreover, with a higher number of teammates, there were more
defensive actions focused on protecting the goal, with more players balancing. In relation to
attack, an increase in the number of opponents produced a decrease in passing, driving and
controlling actions, while an increase in the number of teammates led to more time being
spent in attacking situations. A numerical advantage led to less exploratory behavior, an
effect that was especially clear when playing within a team of seven players against four
opponents. All teams showed strong effects of the number of teammates on the exploratory
behavior when comparing 5 vs 7 or 3 vs 7 teammates. These results seem to be indepen-
dent of the players’ level.
1 / 15
 Exploratory Behavior in Soccer

Introduction
Tactical creativity refers to varying, atypical and flexible decisions and actions, and it plays a
key role in team ball sports due to the complex and chaotic determinants of performance [1,2].
Coaches are especially interested in developing a more flexible game to increase unpredictable
behavior through emergent properties [3] and gain advantage over the opponents [1]. Tactical
creativity can be described by the classical characteristics of creativity defined by Guilford [4]:
the unusualness of tactical actions defines originality; the variety of tactical actions defines flex-
ibility; and the number of tactical actions generated for certain situations defines fluency [5].
Therefore, it is important to design practice tasks that can foster players’ creative behavior and,
consequently, improve performance outcomes. Despite this, few studies have addressed the
topic of creativity in sport performance, and there remains a need to establish solid modeling
approaches for studying and explaining creative behavior [6,7].
The most widely used definitions of creativity refer to obtaining a useful and original prod-
uct or behavior [8–13], however, several authors with closer ties to education have particularly
focused the concept on individual and exploratory behavior. This conceptualization of creativ-
ity classifies as creative any exploratory behavior without regard to whether it leads to a novel
final outcome or not [14]. Exploratory behavior may be defined as a subsequent realization of
a large number of movement configurations under specific constraints of each performer [15].
Therefore, exploratory activity is strongly dependent on the set of constraints imposed on the
system. This observation implies that during training programs no explicitly prescriptive list of
possible actions needs to be provided beforehand to the performer so that s/he might realize
them [6]. This approach is relevant to team sports, which are based on improvisation and on
the interaction between the performers and the environment. The exploratory behavior that
emerges in team sports will depend on the affordances of the individuals involved, and on the
influence of the environment. Affordances have been defined as opportunities [16], invitations
[17] or solicitations [18] for action, but always as a scaled relation between certain properties
of the environment and a performer’s effectiveness, i.e. abilities [19,20]. It is the affordances
that channel the exploratory activity of the performer. The interrelatedness of affordances [18]
allows us to understand the emergence of behavior at different levels of game constraints.
Therefore, it may be necessary to provide task constraints allowing players to adapt constantly
their behavior to the unpredictable opponents’ behavior. Practice tasks should seek to promote
varied and flexible behavior so that players learn to be more adaptive, and the inclusion of
exploratory tasks can help them to be more creative.
A coordination dynamics approach, based on complex systems theories and ecological psy-
chology, offers a useful theoretical explanation of behavior based on performer-environment
relationships [21,22]. Under this approach, individuals and teams can be modeled as complex
social and nondeterministic systems [23]. From this perspective, functional patterns of coordi-
nated behavior emerge from ongoing interactions between system components through a pro-
cess of self-organization between players, the environment, and specific task constraints [24].
Hristovski, Davids, and Araújo [25] showed how the manipulation of task constraints
might enable the emergence of innovative and functional tactical behaviors. In boxing, it was
observed that changing the distance of the participants to a punching bag led to spontaneous
and novel behaviors, in this case, specific striking techniques that had not been practiced or
taught before. It was also shown how the spontaneous and innovative behaviors emerge only
for some specific task constraints. Extremes of very high and very low risk-of-being-hit con-
straints were suppressing the innovative behavior and only medium cost constraints enabled
emergence of novel actions. Torrents, Ric, and Hristovski [26] showed how the manipulation
of task constraints in dancers had a significant effect on the type of configurations performed,

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 2 / 15

 Exploratory Behavior in Soccer

as well as on their explorative behavior (i.e., the degree of exploration of dancers’ movement/
postures). Unusual rules resulted in a greater exploration of movement possibilities. Moreover,
results showed that novel dancers explored a significantly smaller region of the potential land-
scape than experts did, suggesting that this measure may help measuring the dancers’ perfor-
mances [27]. These observations of the emergence of human innovative behaviors are
important from the perspective of Boden’s [28] definition of transformational creativity, since
they explain how manipulating constraints, enables a transformation within a neurobiological
action system without reference to abstract rule-governed behaviors. Traditionally, coaches
design partial and separate practice exercises for technical, physiological, and tactical aspects,
and in some cases there is little correspondence with actual game situations [29]. However,
coaches need to develop practice scenarios that are representative of the actual competition in
order to achieve a better coupling between perception and action during the games [30].
In soccer, a common way to modify task constraints is implementing small-sided games, a
training task widely described in the literature [31,32]. These training scenarios exploit inher-
ent self-organizing tendencies and can simultaneously develop technical skills (related with the
movement of players’ body with the ball) and tactical behaviors (spatial-temporal movements
related with the shared task-goals in offense or defense) in competitive performance environ-
ments. The game is unpredictable and helps players maintain higher levels of motivation in
comparison with non-specific training situations. During small-sided games without external
instructions, the information flows, allowing continuous spontaneous interactions between
teammates and opponents [23]. The most common responses analyzed when playing small-
sided games have been those related to physiological or physical nature [31,33–35], in addition
to technical [36–38] and tactical aspects related to players’ positioning [36, 39]. In particular,
considering the number of players involved, Jones and Drust [40], and Katis and Kellis [41]
reported that the number of technical actions increased with a decrease in the number of play-
ers. However, Abrantes, Nunes, Maçãs, Leite, and Sampaio [42] observed no differences in
player effectiveness in relation to the density of players. Similarly, Duarte, Batalha, Folgado,
and Sampaio [43] founded that, in semi-professional futsal players, the lower the number of
players during the small-sided games, the more frequent the technical actions (e.g., successive
contacts with the ball, the number of dribbles, and the number of tackles). Casamichana and
Castellano [44] studied the frequency of appearance of technical actions in soccer and found
they became more common as the density of players, i.e. individual playing area was reduced.
Specifically, there were differences between small (32 x 23 m) and medium-sized (50 x 35 m)
pitches for the variables control and shoot, clearance, and putting the ball in play. Between
small and large (62 x 44 m) pitches there were also differences for interception, control and
dribble, clearance, and putting the ball in play. Owen et al. [45] found differences in the techni-
cal-related performance with the ball between small- (4 vs. 4), medium- (5 vs. 5 to 8 vs. 8), and
large-sided games (i.e., 9 vs. 9 to 11 vs. 11), where passing, receiving, dribbling, and shots, were
more frequent as the number of players decreased. Nevertheless, very little research has studied
technical responses in games involving unequal numbers of players, even though this kind of
numerical (dis-) advantage is common, and likely to be associated with more game perturba-
tions and, in consequence, more scoring opportunities [32]. Silva et al. [39] evaluated the effect
of different numerical relations (5vs5, 5vs4 and 5vs3) on positioning considering inter-individ-
ual, intra- and inter-team coordination. They found that these asymmetries constrained play-
ers’ individual dominant regions, modified the teams’ dispersion (decreased in inferiority), the
teams’ relative positioning on field (approached the opponent’s goal in superiority) and the
space between line-forces on wings and sectors.
Although small-sided games are well described in the literature, research related to their
effect on the creative behavior of the players is very scarce. In general, very limited research

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 3 / 15

 Exploratory Behavior in Soccer

has been conducted so far focused in the development of training methods that may foster cre-
ative behavior in sport. Some of the proposals to date include the use of diversification, deliber-
ate play, deliberate practice, or deliberate coaching (see [1] for a review). However, measuring
creativity in sport is challenging, and previous research has shown how difficult it can be to
reach an agreement about what constitutes an original response [46]. Due to the subjective
nature of this aspect of tactical creativity, the present study takes into account criteria related
to the variety and quantity of responses (i.e., flexibility and fluency) [4], but does not consider
the rarity or originality of responses. For this reason, we will address exploratory behavior
instead of tactical creativity to describe the players’ creative behavior. Moreover, we propose
unusual training situations with the aim of stimulating greater exploration of the tactical possi-
bilities, considering the suggestion by Stokes [47,48] where in order to produce a creative solu-
tion, constraints not only may preclude existing responses, but also promote novel, and
frequently, opposites. We hypothesize that playing in difficult scenarios such as playing in
numerical inferiority will facilitate the emergence of more varied behavior. We further hypoth-
esize that the modification of the number of opponents and teammates will produce the use of
different number and type of technical and tactical actions. Therefore, the aim of this study
was to examine how the constraints such as number of opponents and teammates affect the
technical, tactical, and exploratory behavior in small-sided games, in both professional and
amateur players.

Materials and Methods

Participants
Twenty-two professional (PRO) males from a single soccer team (age: 25.6 ± 4.9 years; height:
180.5 ± 4.3 cm; weight: 74.7 ± 4.8 kg), and twenty-two amateur (AMA) male players enrolled
in a sports sciences degree (age: 23.1 ± 0.7 years; height: 179.6 ± 6.1 cm; weight: 72.3 ± 5.9 kg)
were recruited for this study. Goalkeepers (GK) took part in the games, but were excluded
from the analysis. All players were informed about the research procedures, requirements,
benefits, and risks, and their written consent was obtained before the study began. The investi-
gation was approved by the local institutional Research Ethics Committee and it conformed to
the recommendations of the Declaration of Helsinki.

Procedure
The study was conducted over two separated sessions, one for PRO and one for AMA, both
following the same procedures. The players were distributed into two teams (a, b) with a fixed
number of players (4), and two teams (c, d) with a variable number of players (3, 5, or 7). The
head coach of the professional team and the teacher of the students was the same person and
used subjective evaluation to distribute the players in the different teams. Their physical, tech-
nical, and tactical performances were the criteria used to constitute balanced teams [31]. Both
sessions involved three different small-sided games: GK + 4 vs. 7 + GK; GK + 4 vs. 5 + GK;
and GK+ 4 vs. 3 + GK, lasting 3 min each, with 4 min of passive recovery (Fig 1). These three
small-sided games formats were repeated twice, yielding a total of six situations for each team.
All these games were played with a numerical mismatch between the teams involved, and in a
randomized order. Each team always competed against the same team (i.e., AMAa vs. AMAc;
AMAb vs. AMAd; PROa vs. PROc; PROb vs. PROd). Neither professional nor amateur players
were familiar with the 4 vs. 7 situations. All the small-sided games were played on a 40 × 30 m
artificial turf pitch, and in accordance with the official soccer rules. To encourage high work-
rate maintenance, coaches or other players were allowed to give verbal encouragement to all
players, but not specific feedback related with players’ performance or their technical/tactical

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 4 / 15

 Exploratory Behavior in Soccer

Fig 1. Procedure followed in the data collection.

doi:10.1371/journal.pone.0168866.g001

behavior. In addition, several balls were placed around the pitch perimeter with the aim of
increasing the effective playing time. In order to avoid the effect of the scoring, the scoreboard
turned to 0 when any team got two goals. Each session ended with a 10 min cool-down, con-
sisted of static stretching exercises.
All small-sided games were video-recorded using three cameras, and analyzed by two
independent experienced soccer coaches (Graduated in Sports Sciences, certified with UEFA
PRO license and with experience as coaches). For all small-sided games, observers took into
account the actions and characteristics of the actions performed by both teams. These actions
or characteristics were defined on a coarse-grained scale of 51 categories (see Table 1) adapt-
ing the proposals of Owen, Twist and Ford [49] and Costa, Garganta, Greco, Mesquita and
Maia [50]. These actions were evaluated each second, following the suggestions of Mendes
and Malacarne [51]. The two repetitions of the same small-sided games were sequenced and
summed, resulting in three 6 min situations for each AMA team and three 6 min situations
for each PRO team. The time during which the players analyzed were not playing (i.e., the
ball left the pitch or the goalkeeper was controlling the ball) was removed, and all trials were
equalized, resulting in an overall analysis of 204 seconds for each type of small-sided games.
The interval and ratio scales, i.e. positional data, are maximally objective and due to this
property they intrinsically have larger potential for generalization. However, space-time data
are not able to provide direct information on which tasks individual players were exploring
and in what context, within the game. Because we were interested in the technical-tactical
and exploratory behavior in the discrete task-space [52] the nominal scale was the scale of
choice. Each 1 s window was defined as a 51-component binary vector representing the full
action configuration during the same time interval. A value of 1 was ascribed to the active
category and a value of 0 to the inactive one. Both observers analyzed one game (6 minutes)
of each team to calculate reliability. The inter-observer agreement yielded a minimum Kappa
coefficient of 0.82, providing a satisfactory guarantee of data reliability. One of the observers
analyzed the other trials of all teams.

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 5 / 15

 Exploratory Behavior in Soccer

Table 1. Observation instrument defined on a coarse-grained scale and based on the technical/tacti-
cal actions of players (Adapted from Owen et al. [45] and Costa et al. [53])*.
Role Technical/tactical Definition
action
Attacker with Run to the ball Player who runs towards the ball with the intention of reaching it
ball Wait Player remains still until the ball is close to where he is, at a point
where another action can be performed
Control Player receives an intended pass or makes contact in attempting
to maintain possession of the ball
Pass Player in possession sends the ball to a teammate
Shoot Player in possession intentionally sends the ball towards the goal
in an attempt to score
Protect Putting one’s body between the ball and one or more opponents
with the aim of keeping possession
Drive Movement of ball carrier towards the goal or changing direction in
order to play in other areas of the pitch line
Feint Gesture or movement to deceive the opponent
Dribble Player in possession, with ball at feet, runs with ball, beats or
attempts to beat an opponent
Intercept Player makes contact with or stops the ball, enabling him to regain
possession and preventing an opponent’s pass from reaching its
intended destination
Deflect Unintentionally changing the trajectory of the ball after it was
kicked by the opponent
Clear Intentionally moving the ball away from a zone or situation close to
one’s own goal
Anticipate Action intending to dispossess an opponent who is in possession
of the ball.
Attackers Wait Remaining in a fixed position or walking, waiting the action of the
without ball ball carrier, teammates and opponent
Support Player’ movements towards the ball carrier offering a passing
option aimed on keeping the ball possession
Unmark Movement of players between the last defender and towards the
goal line amplifying the effective playing space and offering a long
pass option
Defenders Press Actions to regain the ball or attempt to make the opponent lose the
ball
Delay Actions to slow down the opponent’s attempt to move forward with
the ball
Dissuade Take up a position with respect to an opponent without the ball to
make it difficult for him to receive or to prevent him from receiving
the ball
Balance Positioning of off-ball defenders in response to movements of
attackers in an attempt to achieve numerical stability or superiority
with respect to the opposition
Withdraw Move back to a position between the line of the ball and one’s own
goal

*The number of players who performed each action (1, 2, 3 or more) was defined each second.

doi:10.1371/journal.pone.0168866.t001

Data analysis
In order to analyze the influence of the task constraints on the use of the different type and
number of actions, a hierarchical principal components analysis [54] under Direct Oblimin
rotation, with Delta = 0, aiming data reduction was performed (for the suitability of using
principal components analysis with binary variables, see [55]). The number of significant first

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 6 / 15

 Exploratory Behavior in Soccer

level principal components was determined by identifying those that accounted for 80% of
the explained variance. The component correlation matrix of the first-order principal compo-
nents was then subjected to a further higher-order analysis. The component score matrix was
used in order to detect the most salient collective and individual action categories that emerged
during play.
To identify a team’s collective exploratory behavior during games, the average dynamic
overlap, <qd(t)>, was calculated as an average cosine auto-similarity of the overlaps between
configurations with increasing time lag (for more details, see Hristovski et al. [7]). This mea-
sure captures the average similarity of configurations or game patterns at ever increasing time
distances (i.e., time lags) from each other. Hence, it is capable of detecting the rate and breadth
of exploratory behavior over different time scales. For example, if a team uses different types of
attacking actions (i.e. running to the ball, then control it and then pass it, but in the following
attacks the players use other combinations of actions) and defending actions (sometimes press,
sometimes delay and others balance, for instance) will have a larger exploratory breadth than a
team of players who perform the same actions over time. Had the players remained in the
same position or had they repeated the same action throughout the observation time, then the
average dynamic overlap would be a constant equal to 1 (i.e., <qstat> = 1) for all time lags. On
the other hand, if, during the game, the players explored distant configurations of all the possi-
ble combinations of actions, then the average overlap (<qstat)>) would be zero. This will be
detected in a short time scale of some seconds, but also for larger time scales of minutes.
The average dynamic overlap was fitted by the following equation, which is derived for sys-
tems with an intricate hierarchical structure [56]:
< qd ðtÞ >¼ ð1 qstat Þt a þ qstat ð1Þ

where qstat is the asymptotic (i.e., stationary) value of the dynamic overlap, t is the time lag,
and α is the dynamic exponent. qstat detects the long-term exploratory breadth of the team,
and α the rate of exploration. The fit between the theoretical prediction and the experimental
data was assessed by the quasi-Newton method of least squares estimation, with the goodness
of fit being based on the percentage of explained empirical variance.
In the current study, the sample average <qstat> values under different task constraints
using a one-way ANOVA were compared across the different scenarios. A statistical signifi-
cance level of 95% was selected and effect sizes (Cohen’s d) were computed to identify the mag-
nitude of the effects.

Results
Principal component analysis
Under the Kaiser-Guttman criterion the hierarchical principal components analysis initially
revealed between 9 and 14 primary principal components (PCs) in each file. The second-level
PC structure was formed by salient correlated clusters of first-order PCs. A significant dimen-
sional reduction was obtained, resulting in between three and five PCs to analyze. The second-
ary-level correlated structures systematically produced a sole third-level PCs under the
majority of the trials, although a fourth level was present in some cases. By way of an example,
the results from this highest level PC for AMAb and PROa (fixed teams) are presented in Fig
2. The highest level PC of the remaining teams was very similar.
These results showed that the essential technical/tactical actions of soccer were the most fre-
quent in all situations and in both AMA and PRO players. When attacking, the most common
actions used by the player in possession of the ball were controlling, passing, and driving the
ball, whereas waiting or supporting the player with the ball were the actions most used by

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 7 / 15

 Exploratory Behavior in Soccer

Fig 2. Highest level PCs for AMAb and PROa (fixed teams).
doi:10.1371/journal.pone.0168866.g002

attacking players without the ball. On the other side, when defending, results showed less vari-
ability in the different tactical actions used, being the most common combination: one player
pressing, one delaying, one dissuading, and one balancing and/or one withdrawing. Results
for the variable team were very similar, with the exception that the number of players without
the ball, who performed more technical/tactical actions as the number of players in the team
increased.
Due to these similar results for all situations in the highest level PC, which seems to capture
the essential characteristics of soccer, it was important to analyze the first set of primary PCs of
each trial. For the fixed team, an increase in the number of opponents led to an increase in the
time spent defending, this being the case for both PRO and AMA players. A further finding for
both levels when defending was a decrease in the number of players dissuading and an increase
in the number of players balancing. In 4 vs. 7 games, the most frequent scenario was all the
players balancing. In relation to attack, an increase in the number of opponents produced a
decrease in passing and driving actions among PRO players, whereas AMA players showed a
decrease in passing and controlling actions. For both levels there were fewer players waiting
during the game in inferiority. For the variable team, an increase in the number of teammates
led to more time being spent in attacking situations. No clear differences appeared with respect
to the player with the ball, but an increase in the number of teammates did produce an increase
in the number of players waiting and supporting. When defending, there was a clear increase
in the number of players involved in balancing. In all the situations, the most frequent combi-
nation was one player pressing, one delaying, one dissuading, and/or one or more balancing.

Exploratory dynamics
The goodness of fit between the theoretical curve and the experimental data (n = 102) for the
exploratory dynamics (see Eq 1) reached values of R2 = .91 ± 0.04. This result confirms that Eq
(1) is able to capture the essential hierarchical dynamics during small-sided games. Results
from the exploratory dynamics analysis are summarized in Table 2 and Fig 3, where it is pre-
sented the averages of the stationary overlap order parameter, <qstat>, for each trial, compar-
ing the values of each team in all conditions. Results show different values of exploration but
with clear differences between the different small-sided games. For the fixed teams, AMAa
showed a small effect of the number of opponents, just comparing 5 OPP vs 7 OPP. AMAb

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 8 / 15

 Exploratory Behavior in Soccer

Table 2. Mean values (±SD) for the stationary part of the overlap order parameter (<qstat)>) obtained for the three conditions in all teams.
Fixed teams 3 OPP 5 OPP 7 OPP p (Cohen’s d) Variable teams 3 TM 5 TM 7 TM p (Cohen’s d)
AMAa 0.19±0.04 0.19±0.03 0.19±0.02 a) - AMAc 0.2±0.04 0.19±0.02 0.28±0.02 d) p < .01 (0.21)
b) p < .05 (0.36) e) p < .01 (4.27)
c) - f) p < .01 (2.65)
AMAb 0.22±0.03 0.17±0.03 0.16±0.03 a) p < .01 (1.73) AMAd 0.18±0.03 0.19±0.04 0.27±0.03 d) -
b) - e) p < .01 (2.00)
c) p < .01 (2.01) f) p < .01 (2.74)
PROa 0.20±0.04 0.17±0.02 0.17±0.03 a) p < .01 (0.87) PROc 0.2±0.05 0.2±0.03 0.29±0.03 d) -
b) - e) p < .01 (3.27)
c) p < .01 (0.76) f) p < .01 (2.18)
PROb 0.20±0.03 0.19±0.03 0.18±0.03 a) p < .05 (0.43) PROd 0.15±0.04 0.19±0.04 0.28±0.03 d) p < .01 (1.01)
b) - e) p < .01 (2.8)
c) p < .05 (0.66) f) p < .01 (4.02)

Abbreviations: PRO = professional players; AMA = amateurs players; OPP = opponents; TM = teammates. Significant differences are identified as: a) 3
OPP vs. 5 OPP; b) 5 OPP vs. 7 OPP; c) 3 OPP vs. 7 OPP; d) 3 TM vs. 5 TM; e) 5 TM vs. 7 TM; f) 3 TM vs. 7 TM.

doi:10.1371/journal.pone.0168866.t002

Fig 3. Averages of the stationary overlap order parameter, <qstat>, for each trial.
doi:10.1371/journal.pone.0168866.g003

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 9 / 15

 Exploratory Behavior in Soccer

showed strong effects of the number of opponents when comparing 3 vs 5 OPP and 3OPP vs 7
OPP. PROa and PROb showed moderate effects when comparing 3 OPP vs 5 OPP and 3 OPP
vs 5 OPP. In the case of the variable teams, playing with seven teammates clearly produced a
lower exploratory breadth compared with the other conditions. All teams showed strong
effects of the number of teammates when comparing 5 TM vs 7 TM and 3 TM vs 7 TM.

Discussion
This study explored the use of constraints as a way of enhancing players’ exploratory behavior
in soccer, and also analyzed the emergence of individual and collective action configurations
when varying the number of opponents or teammates in the context of numerically unbal-
anced teams. The analysis was only focused on observable motor behavior, therefore, disre-
garding important behavioral factors such as participants’ emotions, thoughts, the influence of
the partners’ behavior, or the subjective meaning of their behavior. In addition, the effective-
ness of play was not analyzed, mainly due to the controversies found in using any objective
measure of overall performance.
The results showed a topological and temporal structure of relatedness between the config-
urations emerging when playing soccer [7]. In general, certain technical/tactical actions
appeared with a high frequency, while others were statistically rare and short-lived reconfigu-
rations that constitute fluctuations following dynamical systems principles [21]. Highly fre-
quent combination of actions loaded predominantly on the first PC, capturing most of the
variance. The variation in the number of opponents and teammates had a significant effect on
the type of action configurations performed by the players in lower order PCs, although this
effect was not appreciated in the highest level PC, which captures the essential characteristics
of the game but not the peculiarities of each situation. This result gives a clear picture of the
emergent hierarchical organization of soccer, in which the general task constraints of the game
give rise to a stable and general structure that is captured by the high similarity of the highest
order PCs [57]. However, more immediate and peculiar task constraints, specific to the relative
number of players on the pitch and, perhaps, to personal constraints of the players involved in
the game, induce a more differentiated structure of decisions-actions at the individual and col-
lective level, and this is captured in lower order PCs.
An increase in the number of opponents produced an increase of the frequency of defensive
patterns and, especially, an increase in the number of players controlling and a decrease in the
number waiting. An increase in the number of teammates produced an increase in the fre-
quency of attacking patterns. By contrast, an increase in the number of opponents produced a
decrease in the use of basic actions such as passing or driving by players in possession of the
ball, while an increase in the number of teammates produced more defensive actions focused
on protecting the goal, with more players balancing, although it did not produce an increase in
the number of players pressing or dissuading. This finding should be taken into account by
coaches when designing their training sessions, since it suggests that training in situations of
numerical inferiority could improve a team’s defensive skills, not only because of the increased
time spent in that situation but also because the players become more active when defending
in an attempt to prevent shooting opportunities by the opponents. In light of these results, it is
suggested that coaches alternate situations of numerical inferiority and superiority in training
sessions by using attacking jokers who play for the team in possession of the ball [58]. This
strategy could help coaches to take advantage of the observed effects of inferiority in defense
and superiority in attack. In general, compared with full-sided games, small-sided games offer
more opportunities for players to practice attacking actions such as passing, shooting, drib-
bling, heading, or supporting teammates with the ball, as well as other defense-related skills

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 10 / 15

 Exploratory Behavior in Soccer

such as intercepting the ball [44]. Moreover, small-sided games provide situations closer to
instability [23,41,45], since in full-sided team games the global system is harder to destabilize
because some subgroups may be inactive [23].
In terms of the emergence of flexible and fluent behavior, players seem to show more
exploratory behavior when playing with a numerical disadvantage. This situation seems to
force players to vary their game more, whereas a numerical advantage appears to produce
game that is less exploratory and less varied. This was especially evident when playing with
seven teammates, since the exploratory breadth was clearly lower in all cases. This result is in
agreement with the proposal of Stokes [48,59], who considers constraints not only as barriers
or structures but also as a possibility to promote a search in different parts of the problem
space. This seems to be a general principle, since it has also been found in individual sports
training sessions. For example, in order to enhance exploratory behavior one can induce con-
straints that suppress more habitual actions lying within the comfort zone [26]. The newly
induced constraints effectively release the level of constraints that play a crucial role in generat-
ing the habitual behavior. As a consequence, new functional behaviors may emerge [6]. Never-
theless, extreme difficult scenarios probably will produce the contrary effect, and players
would not be able to cope with the situation. These results may also be related to that obtained
when studying the positioning dynamics of small-sided games [60], as an increase in the num-
ber of teammates seemed to promote regularity in spatial organization in AMA teams,
although in PRO teams the variation in players’ irregularity ranged from trivial to small. PRO
players probably anticipate better the need to optimize collective decision-making. However,
by studying individual variables, it has been showed that playing in inferiority produces less
varied game patterns, especially in defense [57]. It may be the case that the smaller degree of
diversity at the individual tactical behavior level is compensated with an increased diversity at
the team level, and vice versa [61]. These possible synergic mutually compensating processes
may be a key to understanding the multi-level interdependencies in team sports. That is, each
individual performer may reciprocally influence and be influenced by team’s behavior in a
multilevel synergic relation [62].
In general, these results suggest that easier game situations can promote more regular and
less varied play, while more difficult scenarios force the players to explore the varieties of tacti-
cal/technical actions that they can perform. This conclusion is consistent with the results
obtained in other contexts, for example, the recent analysis of dance improvisation by Torrents
et al. [26]. In that study unusual and more difficult instructional constraints produced a more
varied dance and led to greater exploration of movement possibilities. Previous research on
tactical creativity has shown that it can be improved by practice, with deliberate play or delib-
erate practice [63], but to our knowledge, there is no literature related to the type of specific
task constraints that can enhance creative behavior. The present study suggests that discomfort
may be a good way of enhancing creativity, as has been suggested in other fields such as man-
agement, art or creative services [64], but more research is needed to verify that this theory is
applicable to sport specific settings. More research is also needed in order to evaluate if there is
a curvilinear effect of discomfort on the exploratory behavior in team sports as Hristovski,
Davids and Araújo [25] observed in boxing by modifying the risk-of-being-hit constraints.
The effect of the three small-sided games formats seems to be similar for both AMA and
PRO players. The effect of varying the number of opponents and teammates on the configura-
tions of actions or exploratory behavior seems to be independent of the players’ level. On the
contrary, Silva et al. [39] have found that skill level is determinant in the perception of different
possibilities for action when evaluating positioning of the team. This model could be useful for
studying the emergence of creative behavior in any sport activity, especially team sports or

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 11 / 15

 Exploratory Behavior in Soccer

improvisational situations, and also for exploring the influence of task constraints on the
teaching-learning process.

Acknowledgments
We would like to thank the players who volunteered to participate in this study. We gratefully
acknowledge the support of the Generalitat de Catalunya government project Grup de recerca
en Sistemes Complexos i Esport (2014 SGR 975) and Project Symbiotic technology for societal
efficiency gains: Deus ex Machina (DEM), NORTE-01-0145-FEDER-000026, co-financed by
Fundo Europeu de Desenvolvimento Regional (FEDER) by NORTE 2020.

Author Contributions
Conceptualization: CT AR RH EV LT JS.
Data curation: AR RH.
Formal analysis: AR RH.
Funding acquisition: CT JS.
Investigation: CT EV LT JS.
Methodology: CT AR RH EV LT JS.
Project administration: CT.
Resources: CT EV LT JS.
Software: RH.
Supervision: CT RH LT JS.
Validation: CT AR RH EV LT JS.
Visualization: CT AR.
Writing – original draft: CT AR RH EV LT JS.
Writing – review & editing: CT RH LT JS.

References
1. Memmert D. Teaching tactical creativity in sport: Research and practice. Teaching Tactical Creativity in
Sport: Research and Practice. Taylor and Francis Inc.; 2015. http://www.scopus.com/inward/record.
url?eid=2-s2.0-84942093386&partnerID=tZOtx3y1
2. Memmert D, Roth K. The effects of non-specific and specific concepts on tactical creativity in team ball
sports. J Sports Sci. 2007; 25: 1423–1432. doi: 10.1080/02640410601129755 PMID: 17786695
3. Yamamoto Y, Yokoyama K. Common and Unique Network Dynamics in Football Games. PLoS One.
2011; 6: e29638. doi: 10.1371/journal.pone.0029638 PMID: 22216336
4. Guilford JP. The Nature of Human Intelligence. McGraw-Hill, Inc.; 1967.
5. Memmert D. Tactical creativity in team sports. Res Phys Educ Sport Heal. 2014; 3: 13–18.
6. Hristovski R, Davids K, Araújo D, Passos P. Constraints-induced emergence of functional novelty in
complex neurobiological systems: a basis for creativity in sport. Nonlinear Dynamics Psychol Life Sci.
2011; 15: 175–206. PMID: 21382260
7. Hristovski R, Davids K, Araújo D, Passos P, Torrents C, Aceski A, et al. Creativity in sport and dance:
Ecological dynamics on a hierarchically soft-assembled perception-action landscape. In: Davids K,
Hristovski R, Araújo D, Balagué N, Button C, Passos P, editors. Complex systems in sport. London:
Routledge; 2013. pp. 261–274.
8. Amabile T. Growing up creative. Nurturing a lifetime of creativity. New York: CEF Press; 1989.

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 12 / 15

 Exploratory Behavior in Soccer

9. Kasof J. Explaining Creativity: The Attributional Perspective. Creat Res J. 1995; 8: 397–403.
10. Mumford MD, Simonton K. Creativity in the Workplace: People, Problems, and Structures. J Creat
Behav. 1997; 31: 1–6.
11. Runco MA. Creativity. Annu Rev Psychol. 2004; 55: 657–687. doi: 10.1146/annurev.psych.55.090902.
141502 PMID: 14744230
12. Runco MA, Charles RE. Judgments of originality and appropriateness as predictors of creativity. Pers
Individ Dif. 1993; 15: 537–546.
13. Sternberg RJ, Lubart TI. Investing in Creativity. Psychol Inq. 1993; 4: 229–232.
14. Drazin R, Glynn MA, Kazanjian RK. Multilevel theorizing about creativity in organizations: A sensemak-
ing perspective. Acad Manag Rev. 1999; 24: 286–307.
15. Hristovski R, Davids K, Passos P, Araújo D. Sport Performance as a Domain of Creative Problem Solv-
ing for Self-Organizing Performer-Environment Systems. Open Sports Sci J. 2012; 5: 26–35.
16. Gibson JJ. The ecological approach to visual perception. Boston: Houghton Mifflin; 1979.
17. Withagen R, de Poel H, Araújo D, Pepping GJ. Affordances can invite behavior: Reconsidering the rela-
tionship between affordances and agency. New Ideas Psychol. 2012; 30: 250–258.
18. Bruineberg J, Rietveld E. Self-organization, free energy minimization, and optimal grip on a field of affor-
dances. Front Hum Neurosci. 2014; 8: 1–14.
19. Turvey M. Affordances and prospective control: An outline of the ontology. Ecol Psychol. 1992; 4: 173–
187.
20. Chemero A. An outline of a theory of affordances. Ecol Psychol. 2003; 15: 181–195.
21. Kelso JAS. Dynamic patterns: the self-organization of brain and behavior. Complex Adaptive Systems
series. MIT Press, Bradford Book, Cambridge, MA; 1995.
22. Kelso JAS. Coordination Dynamics. In: Meyers RA, editor. Encyclopedia of complexity and system sci-
ence. Springer; 2009. pp. 1537–1564.
23. Davids K, Araújo D, Correia V, Vilar L. How small-sided and conditioned games enhance acquisition of
movement and decision-making skills. Exerc Sport Sci Rev. LWW; 2013; 41: 154–161.
24. Newell KM. Constraints on the development of coordination. In: Wade MG, Whiting TA, editors. Motor
development in children Aspects of coordination and control. Dordrecht: Martinus Nijhoff; 1986. pp.
341–360.
25. Hristovski R, Davids K, Araújo D. Information for Regulating Action in Sport: Metastability and Emer-
gence of Tactical Solutions. In: Araujo D, Ripoll H, Raab M, editors. Perspectives on Cognition and
Action in Sport. New York; 2009. pp. 43–57.
26. Torrents C, Ric A, Hristovski R. Creativity and emergence of specific dance movements using instruc-
tional constraints. Psychol Aesthetics, Creat Arts. 2015; 9: 65–74.
27. Torrents C, Hristovski R, Coterón J, Ric A. Interpersonal coordination in dance: Inspiration or constraint.
In: Passos P, Chow JY, Davids K, editors. Interpersonal Coordination and Performance in Social Sys-
tems. Oxon: Routledge; 2016. pp. 85–93.
28. Boden MA. What is Creativity. In: Boden MA, editor. Dimensions of Creativity. MIT Press; 1996. Pp.
75–118.
29. Mitchell S, Oslin J. Ecological task analysis in games teaching: tatical games model. In: Davis WE,
Broadhead GD, editors. Ecological task analysis and movement. Human Kinetics; 2007. pp. 161–178.
30. Araujo D, Davids K, Passos P. Ecological validity, representative design, and correspondence between
experimental task constraints and behavioral setting: comment on Rogers, Kadar, and Costall (2005).
Ecol Psychol. 2007; 19: 69–78. Available: http://www.tandf.co.uk/journals/titles/10407413.asp
31. Aguiar MVD, Botelho GMA, Gonçalves BS V, Sampaio JE. Physiological responses and activity profiles
of football small-sided games. J Strength Cond Res. LWW; 2013; 27: 1287–1294.
32. Hill-Haas S V., Dawson B, Impellizzeri FM, Coutts AJ. Physiology of small-sided games training in foot-
ball: A systematic review. Sport Med. 2011; 41: 199–220.
33. Los Arcos A, Vázquez JS, Martı́n J, Lerga J, Sánchez F, Villagra F, et al. Effects of small-sided games
vs. interval training in aerobic fitness and physical enjoyment in young elite soccer players. PLoS One.
Public Library of Science; 2015; 10.
34. Castellano J, Puente A, Echeazarra I, Usabiaga O, Casamichana D. Number of Players and Relative
Pitch Area per Player: Comparing Their Influence on Heart Rate and Physical Demands in Under-12
and Under-13 Football Players. PLoS One. 2016; 11.
35. Halouani J, Chtourou H, Dellal A, Chaouachi A, Chamari K. Physiological responses according to rules
changes during 3 vs. 3 small-sided games in youth soccer players: stop-ball vs. small-goals rules. J

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 13 / 15

 Exploratory Behavior in Soccer

Sports Sci. Routledge; 2014; 32: 1–6. Available: http://www.tandfonline.com/doi/abs/10.1080/

02640414.2014.899707
36. Sampaio JE, Lago C, Gonçalves B, Maçãs VM, Leite N. Effects of pacing, status and unbalance in time
motion variables, heart rate and tactical behaviour when playing 5-a-side football small-sided games. J
Sci Med Sport. 2014; 17: 229–233. doi: 10.1016/j.jsams.2013.04.005 PMID: 23683687
37. Dellal A, Hill-Haas S, Lago-Penas C, Chamari K. Small-Sided Games in Soccer: Amateur Vs. Profes-
sional Players’ Physiological Responses, Physical, and Technical Activities. J Strength Cond Res.
2011; 25: 2371–2381. Available: <Go to ISI>://WOS:000294280200004 doi: 10.1519/JSC.
0b013e3181fb4296 PMID: 21869625
38. Owen AL, Wong DP, McKenna M, Dellal A. Heart Rate Responses and Technical Comparison Between
Small- vs. Large-Sided Games in Elite Professional Soccer. J Strength Cond Res. 2011; 25: 2104–
2110. Available: <Go to ISI>://WOS:000293063500007 doi: 10.1519/JSC.0b013e3181f0a8a3 PMID:
21642858
39. Silva P, Travassos B, Vilar L, Aguiar P, Davids K, Araújo D, et al. Numerical relations and skill level con-
strain co-adaptive behaviors of agents in sports teams. PLoS One. 2014; 9: e107112. doi: 10.1371/
journal.pone.0107112 PMID: 25191870
40. Jones S, Drust B. Physiological and technical demands of 4 v 4 and 8 v 8 games in elite youth soccer
players. Kinesiology. 2007; 39: 150–156.
41. Katis A, Kellis E. Effects of small-sided games on physical conditioning and performance in young soc-
cer players. J Sport Sci Med. 2009; 8: 374–380.
42. Abrantes CI, Nunes MI, Maçãs VM, Leite NM, Sampaio JE. Effects of the number of players and game
type constraints on heart rate, rating of perceived exertion, and technical actions of small-sided soccer
games. J strength Cond Res. 2012; 26: 976–81. Available: http://www.ncbi.nlm.nih.gov/pubmed/
22446670 doi: 10.1519/JSC.0b013e31822dd398 PMID: 22446670
43. Duarte R, Batalha N, Folgado H, Sampaio J. Effects of Exercise Duration and Number of Players in
Heart Rate Responses and Technical Skills During Futsal Small-sided Games. Open Sports Sci J.
2009; 2: 37–41.
44. Casamichana D, Castellano J. Time-motion, heart rate, perceptual and motor behaviour demands in
small-sides soccer games: Effects of pitch size. J Sport Sci. 2010; 28: 1615–1623. Available: <Go to
ISI>://WOS:000284891700013
45. Owen AL, Wong DP, Paul D, Dellal A. Physical and technical comparisons between various-sided
games within professional soccer. Int J Sports Med. Georg Thieme Verlag; 2014; 35: 286–292.
46. Castañer M, Torrents C, Anguera M., Dinusová M. Instrumentos de observación ad hoc para el análisis
de las acciones motrices en Danza Contemporánea, Expresión Corporal y Danza Contact-Improvisa-
tion. Apunt Educ Fis y Deport. 2009; 2009: 14–23.
47. Stokes PD. Variability, constraints, and creativity. Shedding light on Claude Monet. Am Psychol. 2001;
56: 355–359. PMID: 11330236
48. Stokes PD. Creativity from Constraints: What can we learn from Motherwell? from Modrian? from Klee?
J Creat Behav. 2008; 42: 223–236.
49. Owen A, Twist C, Ford P. Small-sided games: the physiological and technical effect of altering pitch
size and player numbers. Insight. 2004; 7: 50–53.
50. Costa IT, Garganta J, Greco PJ, Mesquita I, Maia J. System of tactical assessment in Soccer (FUT-
SAT): Development and preliminary validation. System. 2011; 7: 69–83.
51. Mendes RS, Malacarne LC, Anteneodo C. Statistics of football dynamics. Eur Phys J B. 2007; 57: 357–
363.
52. Keith F, Wieber P, Mansard N, Kheddar A. Analysis of the Discontinuities in Prioritized Tasks-Space
Control Under Discrete Task Scheduling Operations. IROS’11: International Conference on Intelligent
Robots and Systems. San Francisco: IEEE/RSJ; 2011. pp. 3887–3892.
53. Costa IT, Garganta J, Greco PJ, Mesquita I, Maia J. System of tactical assessment in Soccer (FUT-
SAT): Development and preliminary validation. System. 2011; 7: 69–83.
54. Westerhuis JA, Kourti T, MacGregor JF. Analysis of multiblock and hierarchical PCA and PLS models.
J Chemom. 1998; 12: 301–321.
55. Joliffe IT. Principal component analysis. 2nd edition. New York: Springer; 2002.
56. Sibani P, Dall J. Log-Poisson statistics and full aging in glassy systems. EPL Europhysics Lett. IOP
Publishing; 2003; 64: 8.
57. Ric A, Hristovski R, Gonçalves B, Torres L, Sampaio J, Torrents C. Timescales for exploratory tactical
behaviour in football small-sided games. J Sports Sci. 2016; 0414: 1–8.

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 14 / 15

 Exploratory Behavior in Soccer

58. Ric A, Hristovski R, Torrents C. Can joker players favor the exploratory behaviour in football small-sided
games? Res Phys Educ Sport Heal. 2015; 4: 35–39.
59. Stokes PD. Using Constraints to Create Novelty: A Case Study. Psychol Aesthetics, Creat Arts. 2009;
3: 174–180. Available: http://205.186.157.44/imats/Serra2009.pdf
60. Gonçalves B, Marcelino R, Torres-Ronda L, Torrents C, Sampaio J. Effects of emphasising opposition
and cooperation on collective movement behaviour during football small-sided games. J Sport Sci.
2016; 34: 1346–1354.
61. Riley MA, Richardson MJ, Shockley K, Ramenzoni VC. Interpersonal synergies. Front Psychol. 2011;
2: 1–7.
62. Schöner G. Recent Developments and Problems in Human Movement Science and Their Conceptual
Implications. Ecol Psychol. 1995; 7: 291–314.
63. Memmert D. Sports and Creativity. In: Runco E-CMA, Pritzker SR, editors. Encyclopedia of Creativity (
Second Edition). Academic Press; 2011. pp. 373–378. http://dx.doi.org/10.1016/B978-0-12-375038-9.
00207-7
64. Hearn G, Rodrigues JHP, Bridgstock R. Learning processes in Creative Service teams: towards a
dynamic systems theory. In: Hearn G, Bridgstock R, Goldsmith B, Rodgers J, editors. Creative Work
Beyond the Creative Industries: Innovation, Employment and Education. Edward Elgar Publishing
Company; 2014. pp. 175–192.

PLOS ONE | DOI:10.1371/journal.pone.0168866 December 22, 2016 15 / 15

 6. Estudio IV - Dynamics of tactical
behavior in association football when
manipulating players' space of
interaction5

5
Ric, A., Torrents, C., Gonçalves, B., Torres-Ronda, L., Sampaio, J. y Hristovski, R.

(2017). Dynamics of tactical behavior in football when manipulating players' space of

interaction. PLoS ONE 12(7): e0180773. doi:10.1371/journal.pone.0180773

69

a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ric A, Torrents C, Gonçalves B, Torres-
Ronda L, Sampaio J, Hristovski R (2017)
Dynamics of tactical behaviour in association
football when manipulating players’ space of
interaction. PLoS ONE 12(7): e0180773. https://
doi.org/10.1371/journal.pone.0180773
Editor: Tiago M. Barbosa, Nanyang Technological
University, SINGAPORE
Received: January 18, 2017
Accepted: June 21, 2017
Published: July 14, 2017
Copyright: © 2017 Ric et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All data files are
available from the Open Science Framework
Database https://osf.io/gu5kp/?view_only=
e935e0a8a6dc4119b7e67ee7bfe18554.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017
RESEARCH ARTICLE
Dynamics of tactical behaviour in association
football when manipulating players’ space of
interaction
Angel Ric1*, Carlota Torrents1, Bruno Gonçalves2, Lorena Torres-Ronda3,
Jaime Sampaio2, Robert Hristovski4
1 Complex Systems in Sport Research Group, National Institute of Physical Education of Catalonia (INEFC),
University of Lleida, Lleida, Spain, 2 Research Center in Sports Sciences, Health Sciences and Human
Development (CIDESD), CreativeLab Research Community, Universidade de Trás-os-Montes e Alto Douro,
Vila Real, Portugal, 3 Department of Health and Kinesiology Exercise and Sport Nutrition Lab, Texas A&M
University, College Station, Texas, United States of America, 4 Faculty of Physical Education, Sport and
Health, Saint Cyril and Methodious University, Skopje, Macedonia
* angelric10@outlook.com
Abstract
The analysis of positional data in association football allows the spatial distribution of players
during matches to be described in order to improve the understanding of tactical-related
constraints on the behavioural dynamics of players. The aim of this study was to identify
how players’ spatial restrictions affected the exploratory tactical behaviour and constrained
the perceptual-motor workspace of players in possession of the ball, as well as inter-player
passing interactions. Nineteen professional outfield male players were divided into two
teams of 10 and 9 players, respectively. The game was played under three spatial con-
straints: a) players were not allowed to move out of their allocated zones, except for the
player in possession of the ball; b) players were allowed to move to an adjacent zone, and;
c) non-specific spatial constraints. Positional data was captured using a 5 Hz interpolated
GPS tracking system and used to define the configuration states of players for each second
in time. The configuration state comprised 37 categories derived from tactical actions, dis-
tance from the nearest opponent, distance from the target and movement speed. Notational
analysis of players in possession of the ball allowed the mean time of ball possession and
the probabilities of passing the ball between players to be calculated. The results revealed
that the players’ long-term exploratory behaviour decreased and their short-term exploration
increased when restricting their space of interaction. Relaxing players’ positional constraints
seemed to increase the speed of ball flow dynamics. Allowing players to move to an adja-
cent sub-area increased the probabilities of interaction with the full-back during play build-
up. The instability of the coordinative state defined by being free from opponents when play-
ers had the ball possession was an invariant feature under all three task constraints. By
allowing players to move to adjacent sub-areas, the coordinative state became highly
unstable when the distance from the target decreased. Ball location relative to the scoring
zone and interpersonal distance constitute key environmental information that constrains
the players’ coordinative behaviour. Based on our results, dynamic overlap is presented
as a good option to capture tactical performance. Moreover, the selected collective (i.e.
1 / 16
 Dynamics of tactical behaviour

relational) variables would allow coaches to identify the effects of training drills on teams
and players’ behaviour. More research is needed considering these type variables to under-
stand how the manipulation of constraints induce a more stable or flexible dynamical struc-
ture of tactical behaviour.

Introduction
The analysis of performance in sport through the collection and subsequent processing of data
has been widely used to provide useful information for coaches [1]. This performance analysis
has sought to obtain indicators of execution, such as offensive technical actions or successful
defensive events, and has ranked them using statistical procedures to characterize football per-
formance [2–5]. Although these previous studies have led to advances in football performance,
the notation of discrete actions and/or events has not provided information about the certain
performance contexts [6]. For instance, ball location relative to the scoring targets constrains
the emergence of spatiotemporal coordinated team behaviours [7]. Various studies have
employed other measures such as players’ trajectories, interpersonal distances, relative angles
between players or velocities as state variables [8–11]. They have served to define the coordina-
tion system states (patterns) at different levels of analysis (i.e., player, team or game). Hence,
the challenge in performance analysis is to capture key contextual information that helps to
describe and model the varied game scenarios explored by the team and players during both
training and competition [12,13].
A combination of ball events and positional data is needed to understand the players’ and
team’s performance. Thus, several indicators such as player-player and player-ball dyadic
coordination, intra- and inter-team synchronization, pattern-forming dynamics, time
required to regain ball possession, ball possession percentage, number of passes and their
length have been used to characterize individual and collective performance [14–17]. The
notation of passes has also helped to quantify the main interactions between the player in pos-
session of the ball and her or his teammates. These data can be used to measure closeness and
betweenness centrality scores [18–20], flow motifs [21] or a combination of zone and player
passing measures [22] to identify team and player performance and styles of play. Accordingly,
Yamamoto and Yokoyama [23] found the existence of a power law structure in the network of
passing behaviour. This degree distribution suggests that self-organisation phenomena occurs
during ball flow football dynamics and characterize a scale-free network where few players acts
as a game hub. Due to the small number of nodes (11 players), Narizuka and colleagues [24]
included the player location when performing a pass. They found that the degree distributions
fitted with a truncated gamma distribution, showing that each player moves around his own
home position when interacting with the ball. Cotta and colleagues [25] also identified this
positional feature when characterizing the Spanish national team style of play, by analysing
passing network during the 2010 FIFA world cup. Such ball-passing dynamics not only influ-
ence the teams’ dynamics, but also the local players’ dynamics (e.g., movement reconfigura-
tions) [26,27]. That is, after a pass, the player in possession of the ball switches, change the
topology of the passing network (i.e. ball-passing probability distribution) and the perceptual-
motor workspace of the player in possession of the ball.
Football players’ perceptual-motor workspace is the dynamic interface between informa-
tion flows coming from perception, and kinetic flows from action [28,29]. The channelling of
players’ exploratory activity is provided by affordances, i.e., possibilities for action mainly
offered by the environment [30,31]. In fact, they map environmental information onto

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 2 / 16

 Dynamics of tactical behaviour

movement behaviour within the perceptual-motor workspace in accordance with the demands
of the task goal. Due to the multiple ways to score a point in the football game, there is no
predetermined interpersonal coordination to be executed in order to attain the goal. Hence,
there is a lack of convergence of the exploratory behaviour towards a pre-defined intra- or
interpersonal configuration. This situation allows very subtle and consequently unpredictable
interactions between the environmental information flow and the performer’s organismic
constraints, to decide which particular action, from the perceptual-motor workspace, will be
performed at each moment.
These informational variables constrain the system, pushing it towards different states of
organization or concrete coordination patterns [32,33]. The manipulation of constraints, espe-
cially those related to the tasks, will induce the stability of patterns and the functional explora-
tion through the transitions between organizational states inherent to sports teams [34]. The
adequate modification or variations of this type of constraints can produce functional changes
in behaviour in competitive and performance contexts. These qualitative changes in the coor-
dinative system states are characterized by the loss of stability from a previous state to the cur-
rent one. In this sense, behavioural dynamics correspond to the change of trajectories, i.e.
bifurcations, that occur in the performance state space, making reference to the whole set of
possible coordinative states of a system [35].
Key environmental information, and the manipulation of task constraints, will allow play-
ers to explore a variety of functional movements. That is to say, in order to foster the emer-
gence of functional tactical behaviours, coaches should manipulate training drills enhancing
the number of possibilities for action that can satisfy goals, or suppressing habitual behaviours
[36,37]. In individual sports, it has been experimentally demonstrated that constraining such
habitual behavioural modes can promote the emergence of novel behaviours. In fact, this
manipulation promotes an indirect release of constraints and increases the likelihood of spe-
cific emergent behaviour and/or enhances the potential for exploratory behaviour [36,38].
This exploration enhancing mechanism was called ‘the connected door effect’, and it enables
greater fluency and flexibility as well as innovative actions [39].
In association football, several studies have reported the effects of manipulating task con-
straints. For example, increasing the number of goals amplifies possibilities for scoring and,
consequently, a stabilization of protecting goal patterns will likely emerge [40]. The higher the
number of players in small-sided games requires a higher level of collective organization and
optimized space occupation [41]. The number of players involved in the match clearly affects
the stability of the team structure, constrains players to using a specific pitch location and
defines roles during the match. Curiously, when increasing the relative space per player by
reducing the number of players, the players’ positioning tends to be more irregular; however,
when the field dimension increases that tendency changes and players’ positioning becomes
more regular [42].
Furthermore, Vilar and colleagues [43] found that, for most of the match time, two team
members were located in center-back and middle sub-areas of the effective playing space while
one player was located in the back- and front-wing and center-front sub-areas. However, their
results showed that players’ numerical relations were differently distributed depending on the
sub-areas, postulating that local numerical dominance plays a key role in offensive and defen-
sive success. Several studies have explored the influence of numerical imbalances on perfor-
mance [44–46]. For example, Gonçalves and colleagues [47] analyzed the effects on tactical
behaviour when playing with different numbers of opponents or teammates and found higher
values of irregularity and variability in players’ positioning relative to their centroid (i.e., the
geometrical center of the team) when playing with more than one teammate or opponent.
Moreover, under numerical superiority, the players’ distance from their nearest opponent

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 3 / 16

 Dynamics of tactical behaviour

increased, expanding the space and time for interaction. Conversely, Ric and colleagues [48]
analyzed the similarity of individual tactical patterns and found that numerical superiority
allowed players to explore more varied configurations. However, it was reported that this
capacity to explore different patterns expired in time lags up to 20-to-30 seconds, recommend-
ing the use of wild-card players that enter and leave the game in that timescale. In basketball,
Bourbousson and colleagues [49] found that disruption in inter-team coordination fosters the
drive action, proposing a task including an extra player, who systematically plays in offense to
perturb the coordination modes of defenders. These situations require an increase in players’
breadth of attention to solve the numerical disadvantage, allowing the player in possession of
the ball and the supporting players to act within an advantageous context of play. These find-
ings may encourage coaches to use task constraints to preserve the players’ specific locations
and roles, as well as the local and temporary nature of numerical imbalances in order to respect
the representative nature of training drills.
According to all the presented literature, a task was designed where players were con-
strained to create local imbalances while preserving their specific playing location and roles.
The effects of these constraints on the exploratory dynamics and behavioural tendencies
remain unknown. Therefore, the aim of this study was to identify how players’ spatial restric-
tions affected their exploration of tactical behaviour and constrained the perceptual-motor
workspace of players in possession of the ball. We hypothesized that spatial restrictions would
increase the rate and breath of player’s exploratory behaviour. Moreover, coach staff expected
that these task constraints would allow players to possess the ball far from the opponent in
closer zones to the target, enhancing the passing flow.

Methods
Participants and procedures
Twenty-one male professional football players (age: 25.1±4.1 years; playing experience: 18.8
±5.3 years) participated in the study. All the players were informed about the procedures and
signed a consent form to participate voluntarily in the study. Players were divided into two
teams, taking into account their playing positions and their physical, technical and tactical lev-
els according to a coach’s subjective criteria. This was done to ensure that the performance
level of each team to be comparable [50]. The analyzed team comprised ten outfield players
plus a goalkeeper. The opposing team comprised nine outfield players and a goalkeeper. The
opponent goalkeeper was placed in the opponent target zone situated 37.5 m away from the
penalty area, supporting the offensive action of his teammates as an extra player. Based on the
expert knowledge of experienced football coaches and on earlier studies [43,51], the pitch was
divided into nine different sub-areas between the penalty area and the target zone, and the
players were placed according to their position roles (see Fig 1). The head coach then con-
strained the players’ movements between the sub-areas according to the following instructions:
a) players were not allowed to move out of the sub-areas during the game, except for the player
in possession of the ball; b) all players were allowed to move to any adjacent sub-area, and; c)
non-specific spatial constraints. Each of these situations was played for a 5-minute period with
a 3-minute period of passive rest. The three situations were repeated twice in a randomized
order. The sample size was determined by a power analysis, computed using G Power 3.1 [42]
for an effect size of d = 1, α < 0.05, power (1 – β) = 0.95. In similar studies of exploratory
dynamics [38], large effect sizes were observed (the smallest value was d = 0.84). The study was
approved by the Clinical Research Ethics Committee of the Catalan Institute of Health, Uni-
versity Hospital Arnau de Vilanova, Generalitat de Catalunya (Project-CEIC 1325), which fol-
lows the recommendations of the Declaration of Helsinki.

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 4 / 16

 Dynamics of tactical behaviour

Fig 1. Schematic representation of the experimental task.

https://doi.org/10.1371/journal.pone.0180773.g001

Data collection
All players wore, on their upper backs, a GPS device (SPI-Pro, GPSports, Canberra, ACT, Aus-
tralia) that captured the latitudinal and longitudinal coordinates with a sampling frequency of
5 Hz. The time series were exported and transformed into a Cartesian coordinate system using
dedicated routines in Matlab R2014b (MathWorks, Inc., Massachusetts, USA) (see Folgado
et al. [15] for complete guidelines). The players’ positioning data were resampled using two-
frames moving average filter to reduce tracking error noise. In addition, the player who
received the ball was notated, as was the time that elapsed between his reception of the ball and
that of another teammate before a pass. The dynamic players’ positional data was used to cal-
culate movement speed [15], distance from the nearest opponent [52], distance from the target
[53] and tactical actions [48,54]. Tactical actions were estimated from players’ trajectories rela-
tive to the position of the player in possession of the ball. The data collected for each player
yielded configuration states derived from the 37 variables pertaining to the four tactical mea-
sures named above (see Table 1). Players changed their states during the 5-minute game (i.e.,
300 seconds each game). Thus, every window of one second was defined as a 37-component
binary vector (column) representing the full configuration state, ascribing a value of 1 for
active categories and 0 for the inactive ones. This enabled the formation of a temporal 37 x 300
multivariate binary (Boolean) matrix.

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 5 / 16

 Dynamics of tactical behaviour

Table 1. Data collected to assess the tactical patterns of each player, formed by 37 categories. Each
data vector represented a player’s configuration in a 4D-category space.

VARIABLE (number of categories per Category CATEGORIES (37)

variable) number
Tactical actions (10) 1 Penetration
2 Offensive coverage
3 Depth mobility
4 Width and length
5 Offensive unity
6 Delay
7 Defensive coverage
8 Balance
9 Concentration
10 Defensive unity
Distance from the target (9) 11 >37.45 m.
12 32.1–37.45 m.
13 36.75–32.1 m.
14 21,4–26.75 m.
15 16.05–21.4 m.
16 10.7–16.05 m.
17 5.35–10.7 m.
18 0–5.35 m.
19 <0 m.
Distance from the nearest opponent (12) 20 <1 m.
21 1–2 m.
22 2–3 m.
23 3–4 m.
24 4–5 m.
25 5–6 m.
26 6–7 m.
27 7–8 m.
28 8–9 m.
29 9–10 m.
30 10–11 m.
31 >11 m.
Movement speed (6) 32 <0.7 km
h−1 (stand)
33 0.7–3.6 km
h−1 (walk)
34 3.6–7.2 km
h−1 (jog)
35 7.2–14.4 km
h−1 (medium-intensity
running)
36 14.4–19.8 km
h−1 (high-intensity
running)
37 >19.8 km
h−1 (sprint)

https://doi.org/10.1371/journal.pone.0180773.t001

Data analysis
Dynamic overlap qd(t) was used to determine the region of the performer-environment state
space explored by the players and the rate of exploration on different timescales [38,48,55].
The overlap was defined as a cosine similarity between two binary configuration vectors at
ever-increasing time distances (i.e., time lags), capturing the mean similarity of configuration

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 6 / 16

 Dynamics of tactical behaviour

states. The mean dynamic overlap was then fitted by the following equation, which is derived
for systems with an intricate hierarchical structure (see Sibani & Dall [56]):
< qd ðtÞ >¼ ð1 qstat Þt a þ qstat

From the known mean dynamic overlap (<qd(t)>) at every time lag (t), three different
parameters were calculated. qstat is the asymptotic value of the dynamic overlap (i.e. the hori-
zontal line towards which the curve, adjusted by the equation, tends to infinity). The α param-
eter is the slope of the curve. Finally, T is the time point where, for a fixed value of 0.05, the
asymptotic value intersects with the curve of the non-linear model (see upper panels of Fig 2
for a better interpretation).
In order to understand the ball dynamics, the pooled means of time that a player was in
possession of the ball before passing it to a teammate, losing it or scoring a goal were calcu-
lated. The total passes performed with successful reception by a teammate, as well as turnovers
(losing possession of the ball) and goals scored were also taken into account, obtaining the
relative frequencies of players’ passing interactions. Transition probabilities were calculated
dividing the number of each player’s passes to his teammates, turnovers and goals by the total
number of player interactions. Thus, the most probable sequences of passes between players
allowed the inter-player interaction tendencies and the characteristic passing channels to
reach the target/scoring zone to be identified.

Fig 2. The upper panels show an example of evolution in the mean dynamic overlap of the same player for three different task constraints of a
player: restricted (left), semi (center), free (right). The blue lines represent the adjusted curve to the non-linear function, the grey lines represent the
stationary <qstat> value when the curve tends to infinity, the red lines represent the time lag in which, for a fixed value of 0.05, the asymptotic value intersects
with the curve. The lower panels show the mean values for <qstat> (left), α exponent (center) and time lag (right). Differences in mean are expressed as
percentages (±90% CL). The asterisks indicate the likelihood for the magnitude of the true difference in means as follows: *possible; **likely; ***very likely;
****most likely. The letters denote the effect sizes: T = trivial; S = small; M = moderate; L = large.
https://doi.org/10.1371/journal.pone.0180773.g002

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 7 / 16

 Dynamics of tactical behaviour

In addition, distance from the nearest opponent and distance from the target were the state
variables, and the 2D space that they spanned was the configuration (state) space of the player
in possession of the ball. The configuration space contains all possible configurations under
the task constraints. The relative frequencies of the player in possession of the ball’s configura-
tion space were taken into account for each task constraint. The probabilities of each configu-
ration were calculated as limit (large N) relative frequencies for stationary distributions:
pi ¼ ni =N , where ni is the frequency of the configuration and N is the total frequencies. This
data allowed a potential landscape to be depicted, which represented the configuration space
of the player in possession of the ball for each task constraint. The potential values Vi were cal-
culated following [57] as: Vi ¼ Q lnðPi =N Þ; where Q = 1 is the standardized variance term; and
N is the total number of configurations in the space spanned by the two state variables.

Statistical analysis
Practical differences among task constraints were analysed using magnitude-based inferences
[58]. All data were first log-transformed to reduce bias arising from non-uniformity error. A
descriptive analysis was performed using mean and standard deviations for each variable (the
mean shown is the back-transformed mean of the log transformation). Uncertainty in the dif-
ferences was expressed as 90% of confidence limits (CL). Quantitative chances were assessed
qualitatively and reported using the following scale: 25−75%, possible; 75−95%, likely; 95
−99%, very likely; >99%, most likely. A difference was assessed as being unclear if the confi-
dence interval (CI) overlapped both substantially positive and negative thresholds. Cohen’s d
effect size at 90% CL was calculated using pooled standard deviation for comparisons, and the
magnitude ranges for mean differences were: 0–0.2 trivial; >0.2–0.6 small; >0.6–1.2 moderate;
>1.2–2 large; >2 very large [48]. In addition, the Chi-square test was used to compare the fre-
quencies of players’ interactions and frequencies of the configuration space between the three
task constraints.

Results
Dynamics of players’ exploratory behaviour
Dynamic overlap analysis allowed the slow dynamics on a long timescale (where players’
exploration became sufficiently saturated), and the quick dynamics on a shorter timescale
(related to the initial relaxation part of the overlap) to be determined (see upper panels of Fig 2
as an example of one player for each task constraint). The lower panel of Fig 2 shows the
descriptive and statistical analysis for all three parameters extracted from the non-linear
model. Long-term exploratory breadth most likely decreased (higher <qstat>) when playing
under restricted space compared to semi (large effect) and free space (large effect) conditions.
Also, a possible decrease was identified by comparing semi and free space scenarios (small
effect). When playing under restricted space conditions, players’ exploratory dynamics quickly
attained the stationary value (higher α) and likely decreased during the other two training situ-
ations. So, the larger the slope (α exponent), the quicker the players’ exploration. There were
unclear differences between semi and free space conditions. Finally, the time lag in which the
exploratory behaviour became saturated most likely (moderate/large effect) increased in the
tasks under semi and free space conditions, respectively, compared to the restricted condition.

Ball flow dynamics

Table 2 shows the results of ball flow dynamics. The mean time (in seconds) during which
players were in possession of the ball (Δt ball possession) possibly decreased under the free-

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 8 / 16

 Dynamics of tactical behaviour

Table 2. Descriptive analysis (mean±SD) of mean time of players’ ball possession, frequencies of passing interaction and frequencies of the
player in possession of the ball’s configurations. Difference in means, uncertainty in the true differences, based on probability chances, and Standardized
Cohen’s d differences among training game situations.

Restricted Semi Free Difference in means, Chances for smaller/ Uncertainty in the Standardized Cohen’s
%; ±90% CL similar/ greater value true differences d; ± 90% CL
a) -4.6 ±17.3 11/52/36 unclear -0.11±0.41
Δt ball 3.19±1.47 3.03 2.71 b) -7.1 ±21.0 12/41/47 unclear -0.17±0.51
possession ±1.08 ±0.73
n 20 20 20 c) -8.7 ±13.8 3/44/53 possibly # -0.21±0.35
a) -15.6 ±24.2 5/30/65 unclear -0.32±0.52
Passing 0.72±1.27 0.57 0.72 b) -1.3 ±20.4 17/61/22 unclear -0.02±0.38
interactions ±0.94 ±1.01
n 110 110 110 c) 25.1 ±26.4 83/17/1 likely " 0.42±0.39
a) -26.8 ±13.4 0/9/91 likely # -0.37±0.22
Configuration 3.22±6.18 2.63 2.40 b) -34.9 ±9.9 0/0/100 most likely # -0.51±0.18
space ±4.26 ±3.08
n 108 108 108 c) -4.6 ±15.9 2/86/12 likely trivial -0.06±0.20

Note: Δt = mean time; CL = confidence limits; " = increase; # = decrease. Comparisons between the three different training situations are identified as: a)
restricted vs semi, b) restricted vs free and c) semi vs free

https://doi.org/10.1371/journal.pone.0180773.t002

space condition (~-9%) compared to the semi space condition, while an unclear effect was
shown between the restricted and the other two task constraints. Fig 3 shows that the mean
time of players in ball possession was larger under restricted condition, specially central
defenders.
The Chi-square test did not show any significant differences (p<0.05) for each pair compar-
ison. The passing interactions likely increased under the free space condition compared to the
semi space condition (difference in means, %; ±90% CL: -25.1; ±26.4). Unclear differences
were shown when comparing the restricted space condition with the other two task constraints,
although a small effect (-0.32±0.52) was found between restricted and semi space conditions
(see Table 2). As shown in the network for the restricted space situation (see upper panels of
Fig 3), the most probable passing channel was found among players situated in the central cor-
ridor, connecting with the wings to score a goal or to pass to the forward for scoring. Not
allowing the players to move out of their zone decreased the possibilities of passing interaction
between full-backs and their teammates. This tendency changed when players were allowed to
move to the adjacent sub-area, as did the probability of interaction between midfielders and
forwards. In contrast, the probability of interaction with forwards decreased in the situation
where players’ movements were not restricted. In addition, the probability of a player passing
to the central midfielder increased.

Player in possession of the ball’s configuration space

The Chi-square test shows significant differences in the frequency values when comparing
restricted and semi space conditions [χ2 (107) = 369.96; P<0.001], restricted and free space con-
ditions [χ2 (107) = 348.05; P<0.001] and semi and free space conditions [χ2 (107) = 222.30;
P<0.001; η2 = 0.16]. Inferential magnitude-based analyses are presented in Table 2 (restricted
vs semi space conditions, restricted vs free space conditions and semi vs free space conditions,
respectively). The results showed a moderate negative effect when playing in the restricted
space scenario (likely 26.8±13.4% and most likely 34.9±9.9% decrease compared to semi and

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 9 / 16

 Dynamics of tactical behaviour

Fig 3. Upper panels: Network diagrams obtained from each task constraint. Size of nodes represents the mean
time of players in possession of the ball. The width represents the frequency number of passes. Probability of passing
interactions was depicted as the following soften scale: 0 –blue, 0.5 –yellow, 1 –red. Lower panels: Potential landscapes
formed by two state (coordinative) variables of the player in possession of the ball (distance from the target and nearest
opponent) under the three different task constraints. The 3D deeper wells correspond to 2D-projected more stable (i.e.,
more probable) red areas. The blue areas correspond to unstable coordinative states. Less stable coordinative states
are more likely to decay into more stable states.
https://doi.org/10.1371/journal.pone.0180773.g003

free space conditions). When comparing semi and free space conditions, there were likely triv-
ial (-4.6±15.9%) differences with a trivial effect (Table 2).
The high-value surfaces of the potential landscape define unstable coordinative states while
the minimum values define highly stable coordinative states (see lower panels of Fig 3). The
first thing to notice is that the changing task constraints brought about not only quantitative
changes but also qualitative changes, i.e. bifurcations, in the configuration space. The stability
profile and also the number of stable minima changed as a consequence of varying the task
constraints. However, under all three task constraints, there were also invariant features of the
potential landscape. On the one hand, a “free player” coordinative state was constantly unsta-
ble. This meant that the player in possession of the ball was five meters away from his nearest
opponent. Under the restricted space condition, deeper valleys were depicted, showing that the
player in possession of the ball worked within highly stable contexts. Semi space conditions,
such as those where the ball was taken into distant zones, afforded less stable local states. How-
ever, when the ball was fewer than 30 meters from the scoring zone, the opponent delayed the
ball carrier’s action (i.e., distance from the nearest opponent less than five meters), depicting
other stable configurations for all three conditions. Moreover, all potential landscapes showed
how the pressure increased when the players took the ball closer to the target zone.

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 10 / 16

 Dynamics of tactical behaviour

Discussion
This study aimed to present the influence of spatial restrictions on players’ movements on the
exploratory dynamics of tactical behaviour, ball flow dynamics and performance contexts
when players were in possession of the ball. The main finding suggests that restricting players’
movements out of their home sub-area enhanced their exploration in a short timescale, but
not their long-term exploratory behaviour. Furthermore, results did not reveal that spatial
restrictions help players to possess the ball in advantageous conditions, i.e. far from the oppo-
nent in advanced zones. However, allowing players to move out of their home sub-area fos-
tered the ball flow dynamics and the inter-player passing relations.
In terms of exploratory dynamics, players produced a higher long-term exploratory breadth
(i.e., lower <qstat>) when allowed to move out of their specific locations. While the results
might seem obvious, previous studies have shown that increasing the space of interaction leads
to decreases in the variability of players’ spatial distribution [42,59]. These differences may be
due to the multivariate analysis used on here, which involved tactical actions, as well as to tra-
jectories, distances, and movement-speed state variables. The players’ exploration quickly
became saturated under restricted spatial condition. This means that players were constantly
exploring new task solutions but quickly expired the whole set of possibilities for action. On
the contrary, relaxing spatial restrictions increased long-term exploratory behaviour. Some
previous results showed that a high number of opponents generally impaired long-term tacti-
cal exploration [48]. However, these constraints increased the probability of exploring certain
specific coordinative states, e.g., depth mobility back to the defence.
The manipulation of task constraints changes the probabilities of emergent tactical be-
haviours and the exploration in different regions of the task solution space [55,60] Players’
behaviour can be defined by several variables that shows the relations between the system com-
ponents. In this sense, performance coordinative variables, such as distance to opponents and
the target can serve as parsimonious macroscopic descriptors of what happens at the micro-
scopic level [53,61].
The depiction of these state variables on a performance configuration space gives a clear
picture, thus allowing it to be determined whether or not the task constraints analysed foster
an advantageous performance game scenario. For example, the restricted space condition
allowed players to play the ball far from their nearest opponent when decreasing the distance
from the target zone. This task constraint forced players to keep a large team surface area,
thereby fostering penetration actions or passing the ball to a free teammate [7,49]. During the
semi condition, this coordinative state had a greater degree of instability. Conversely, allowing
players to move to the adjacent sub-area increased the stability of coordinative states, with the
player in possession of the ball delayed in zones far from the target. The depth of the potential
of the player in possession of the ball clearly showed how forbidding players to move out of
their home sub-area can limit the exploration of task solutions yet foster the stability of con-
crete task solution states [48,62]. Moreover, it was shown that the change of task constraints
changes not only the quantitative aspects of the game, e.g. modifying the stability degree of
extant stable coordinative states, but also causes a formation of new stable states. This means
that changes of task constraints may create qualitatively new contexts of exploration for players
that cannot be interchangeable with respect to expected effects.
Network analysis allowed quantifying players performance [20] and identifying different
player solicitations. In that sense, the right fullbacks were no able to succeed during their inter-
ventions with the ball in all three scenarios. However, playing under the semi space condition
reduced the barriers to interacting with full-backs compared to the other two conditions. Pre-
vious network analysis in football has shown preferred passing channels and how they changed

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 11 / 16

 Dynamics of tactical behaviour

depending on the opposing team [25,63]. Positional constrains might be considered to simu-
late opponents’ defensive strategies. Allowing players to move on the adjacent sub-areas, wing
opponents were able to press players located in the central corridor, leaving the fullbacks and
wing forward players free of a marker, fostering the association with them. However, that task
constraints lead to a lower centralized passing interaction patterns and higher intra-team well-
connected passing relations, these behaviours are related with a better team performance
[18,19]. Grund [64] found that high levels of interaction (i.e., passes per minute weighted by
possession) led to increased team performance (i.e., goals). The more the players were in pos-
session of the ball, the fewer passes per minute could be performed. According to this state-
ment, relaxing players’ positional constraints would enable an increase in team performance
regarding the quickness of ball flow dynamics and decreasing the players’ dwell time in posses-
sion of the ball. Larger passes samples would probably allow to identify the common features
and emergent properties of the dynamic network topology under the effects of different task
constraints [23,24].
In light of these results, when the coach used sub-spaces to constrain players’ behaviours,
the exploratory behaviours of players quickly attained the stationary value, showing larger val-
ues of overlapped behaviours on a shorter timescale. Nevertheless, these positional restrictions
helped to find advantageous scenarios for the player in possession of the ball. It has been dem-
onstrated that stability in the degree of free movement of players can be constrained by chang-
ing the relative space per player by increasing the number of players [52]. Five-a-side formats
increase the percentage of time spent from four to six meters away from the nearest opponent
but decrease the probability of taking action from six meters free of opponents, as more advan-
tageous contexts of play. Interestingly, Silva and colleagues [59] established that, with a larger
(152 m2) relative space per player, the percentage of time spent in numerical advantage very
likely increased by changing the field dimension instead of the number of players. Under
imbalanced small-sided game conditions, the radius of free movement most likely increased
for those players who competed in numerical superiority [47]. From these and previous results,
a clear role of the constraints with respect to coaches’ goals arises. If the goal of the coach is to
generally increase the individual player’s exploratory behaviour, he/she would have to allow
players to move to other sub-areas of the pitch for the training task. If the goal is to elicit and/
or stabilize specific coordinative states, then restricting positions to a concrete zone may be an
optimal training context. Therefore, coaches should carefully design their training tasks. For
this purpose, reading later studies that evidence the tactical effects of modifying tasks con-
straints is advisable. In addition, more research is needed in order to better understand the
effects of task constraints involving large numbers of players, in which the spatiotemporal
dynamics of game constraints is emphasized.
Previous research using a football match highlighted the importance to create local imbal-
ance on specific zones of the area of play [43]. Further research should include other relational
variables, such as angles, in order to analyse the relation between at least three system compo-
nents (e.g. ball-player-target). Moreover, multivariate analysis (e.g. Principal Component
Analysis) would allow knowing the (co) relation between the already relational variables cap-
tured characterizing the dynamical structure of tactical behaviour under the influence of task
and environmental constraints. The dynamic nature of constraints and their evolution on dif-
ferent time scales should be also considered for investigation.

Conclusion
Performance analysis can be improved by providing information about the dynamics of foot-
ball games. The combination of ball events and positional data can help coaches to understand

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 12 / 16

 Dynamics of tactical behaviour

the effect of a task constraint on individual and collective behaviour. Moreover, the combina-
tion of relational variables can give a clear picture about the tactical performance of players
and its dynamics. Network analysis is widely used during football matches, but its use in train-
ing settings is still limited. Here, it has been reported new knowledge on how ball flow dynam-
ics could be constrained through pitch positioning restrictions. Spatial restrictions did not
stimulate the long-term players’ exploratory breadth, but increased the rate of exploration to
perform different tactical solutions on a shorter timescale. Dynamic overlap might be consid-
ered as a potential order parameter for performance analysis in sport. The depiction of rela-
tional variables also provide relevant information for tactical performance in a macroscopic
level. Overall, coaches can foster players’ exploration and/or to stabilize concrete coordinative
states by constraining the players’ space of interaction behaviour. In this sense, depending on
the training goals of the coaches i.e. enhancing the exploration of task solutions or focusing on
smaller set of task solutions, they can manipulate the constraints to attain each of them.

Author Contributions
Conceptualization: Angel Ric, Carlota Torrents, Bruno Gonçalves, Jaime Sampaio, Robert
Hristovski.
Data curation: Angel Ric, Bruno Gonçalves, Robert Hristovski.
Formal analysis: Angel Ric, Bruno Gonçalves, Jaime Sampaio, Robert Hristovski.
Funding acquisition: Carlota Torrents, Lorena Torres-Ronda.
Investigation: Angel Ric, Carlota Torrents, Bruno Gonçalves, Jaime Sampaio, Robert
Hristovski.
Methodology: Angel Ric, Carlota Torrents, Bruno Gonçalves, Robert Hristovski.
Project administration: Carlota Torrents.
Resources: Angel Ric, Carlota Torrents, Jaime Sampaio.
Software: Angel Ric, Bruno Gonçalves, Jaime Sampaio.
Supervision: Carlota Torrents, Bruno Gonçalves, Lorena Torres-Ronda, Jaime Sampaio, Rob-
ert Hristovski.
Validation: Carlota Torrents, Bruno Gonçalves, Jaime Sampaio, Robert Hristovski.
Visualization: Angel Ric, Bruno Gonçalves, Robert Hristovski.
Writing – original draft: Angel Ric, Carlota Torrents, Bruno Gonçalves, Lorena Torres-
Ronda, Jaime Sampaio, Robert Hristovski.
Writing – review & editing: Angel Ric, Carlota Torrents, Bruno Gonçalves, Lorena Torres-
Ronda, Jaime Sampaio, Robert Hristovski.

References
1. Hughes M, Franks IM. Notational analysis of sport: Systems for better coaching and performance in
sport. Psychology Press; 2004.
2. Castellano J, Casamichana D, Lago C. The use of match statistics that discriminate between successful
and unsuccessful soccer teams. J Hum Kinet. 2012; 31: 137–147.
3. Rampinini E, Impellizzeri FM, Castagna C, Coutts AJ, Wisløff U. Technical performance during soccer
matches of the Italian Serie A league: Effect of fatigue and competitive level. J Sci Med Sport. 2009; 12:
227–233. https://doi.org/10.1016/j.jsams.2007.10.002 PMID: 18083631

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 13 / 16

 Dynamics of tactical behaviour

4. Liu H, Gómez M-A, Gonçalves B, Sampaio J. Technical performance and match-to-match variation in
elite football teams. J Sports Sci. 2016; 34: 509–18. https://doi.org/10.1080/02640414.2015.1117121
PMID: 26613399
5. Sampaio J, McGarry T, Calleja-González J, Jiménez Sáiz S, Schelling I Del Alcázar X, Balciunas M.
Exploring game performance in the National Basketball Association using player tracking data. PLoS
One. 2015; 10. https://doi.org/10.1371/journal.pone.0132894 PMID: 26171606
6. James N. Notational analysis in soccer: past, present and future. Int J Perform Anal Sport. 2006; 6: 67–
81.
7. Travassos B, Araújo D, Duarte R, McGarry T. Spatiotemporal coordination behaviors in futsal (indoor
football) are guided by informational game constraints. Hum Mov Sci. 2012; 31: 932–945. https://doi.
org/10.1016/j.humov.2011.10.004 PMID: 22672740
8. Vilar L, Araújo D, Davids K, Travassos B, Duarte R, Parreira J. Interpersonal coordination tendencies
supporting the creation/prevention of goal scoring opportunities in futsal. Eur J Sport Sci. Taylor & Fran-
cis; 2014; 14: 28–35. https://doi.org/10.1080/17461391.2012.725103 PMID: 24533492
9. Travassos B, Araújo D, Davids K, Esteves PT, Fernandes O. Improving passing actions in team sports
by developing interpersonal interactions between players. Int J Sport Sci Coach. Multi Science Publish-
ing; 2012; 7: 677–688.
10. Corrêa UC, Pinho ST De, Letı́cia S, Clavijo AR, Souza TDO, Tani G, et al. Revealing the decision-mak-
ing of dribbling in the sport of futsal Revealing the decision-making of dribbling in the sport of futsal. J
Sports Sci. Routledge; 2016; 34: 2321–2328. https://doi.org/10.1080/02640414.2016.1232488 PMID:
27868480
11. Passos P, Araújo D, Davids K, Gouveia L, Milho J, Serpa S. Information-governing dynamics of
attacker-defender interactions in youth Rugby Union. J Sport Sci. 2008; 26: 1421–1429. https://doi.org/
10.1080/02640410802208986 PMID: 18923958
12. McGarry T. Applied and theoretical perspectives of performance analysis in sport: Scientific issues and
challenges. Int J Perform Anal Sport. 2009; 9: 128–140.
13. Balagué N, Torrents C, Pol R, Seirul-lo F. Entrenamiento integrado. Principios dinámicos y aplica-
ciones. Apunt Educ fı́sica y deportes 2014; 116:60–68. https://doi.org/10.5672/apunts.2014-0983.es.
(2014/2).116.06
14. Travassos B, Araújo D, Vilar L, McGarry T. Interpersonal coordination and ball dynamics in futsal
(indoor football). Hum Mov Sci. 2011; 30: 1245–1259. https://doi.org/10.1016/j.humov.2011.04.003
PMID: 21683464
15. Folgado H, Duarte R, Fernandes O, Sampaio J. Competing with lower level opponents decreases intra-
team movement synchronization and time-motion demands during pre-season soccer matches. PLoS
One. 2014; 9: e97145. https://doi.org/10.1371/journal.pone.0097145 PMID: 24817186
16. Tenga A, Larsen Ø. Testing the Validity of Match Analysis to describe Playing Styles in Football. Inter-
national Journal of Performance Analysis in Sport. 2003; 3:90–102.
17. Kempe M, Vogelbein M, Memmert D, Nopp S. Possession vs. Direct Play: Evaluating Tactical Behavior
in Elite Soccer. Int J Sport Sci. 2014; 4: 35–41. https://doi.org/10.5923/s.sports.201401.05
18. Gonçalves B, Coutinho D, Santos S, Lago-Penas C, Jiménez S, Sampaio J. Exploring Team Passing
Networks and Player Movement Dynamics in Youth Association Football. PLoS One. Public Library of
Science; 2017; 12: e0171156. https://doi.org/10.1371/journal.pone.0171156 PMID: 28141823
19. López Peña J, Touchette H. A network theory analysis of football strategies. In: (ed.) CC, editor. Sports
Physics: Proc 2012 Euromech Physics of Sports Conference; Palaiseau: Editions de l’Ecole Polytechni-
que; 2012. p. 517–28.
20. Duch J, Waitzman JS, Amaral LA. Quantifying the performance of individual players in a team activity.
PLoS One. 2010; 5(6):e10937. https://doi.org/10.1371/journal.pone.0010937 PMID: 20585387
21. Gyarmati L., Kwak H., Rodriguez P. Searching for a unique style in soccer, in Proceedings of the 2014
KDD Workshop on Large-Scale Sports Analytics. arXiv: 1409.0308.
22. Cintia P, Giannotti F, Pappalardo L, Pedreschi D, Malvaldi M. The harsh rule of the goals: data-driven
performance indicators for football teams. Data Science and Advanced Analytics (DSAA), 2015 36678
2015 IEEE International Conference on. IEEE; 2015. pp. 1–10.
23. Yamamoto Y, Yokoyama K. Common and Unique Network Dynamics in Football Games. PLoS One.
2011; 6: e29638. https://doi.org/10.1371/journal.pone.0029638 PMID: 22216336
24. Narizuka T, Yamamoto K, Yamazaki Y. Statistical properties of position-dependent ball-passing net-
works in football games. Phys A Stat Mech its Appl. 2014; 412: 157–168. https://doi.org/10.1016/j.
physa.2014.06.037
25. Cotta C, Mora AM, Merelo JJ, Merelo-Molina C. A network analysis of the 2010 FIFA world cup cham-
pion team play. J Syst Sci Complex. 2013; 26: 21–42. https://doi.org/10.1007/s11424-013-2291-2

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 14 / 16

 Dynamics of tactical behaviour

26. Ric A, Torrents C, Gonçalves B, Sampaio J, Hristovski R. Soft-Assembled Multilevel Dynamics of Tacti-
cal Behaviors in Soccer. Front Psychol. 2016; 7: 1513. https://doi.org/10.3389/fpsyg.2016.01513 PMID:
27761120
27. Wang Q, Zhu H, Hu W, Shen Z, Yao Y. Discerning Tactical Patterns for Professional Soccer Teams: An
Enhanced Topic Model with Applications. Proceedings of the 21th ACM SIGKDD International Confer-
ence on Knowledge Discovery and Data Mining. New York, NY, USA: ACM; 2015. pp. 2197–2206.
10.1145/2783258.2788577
28. Newell KM, McDonald P V, Kugler PN. The perceptual-motor workspace and the acquisition of skill. In:
Requin J, Stelmach GE, editors. Tutorials in motor neuroscience. Springer; 1991. pp. 95–108.
29. Kugler PN, Turvey MT. Resources for ecological psychology. Information, natural law, and the self-
assembly of rhythmic movement. Hillsdale, NJ, England: Lawrence Erlbaum Associates; 1987.
30. Gibson JJ. The ecological approach to visual perception. Boston, MA, US: Houghton, Mifflin and Com-
pany; 1979.
31. Bruineberg J, Rietveld E. Self-organization, free energy minimization, and optimal grip on a field of affor-
dances. Front Hum Neurosci. 2014; 8: 599. https://doi.org/10.3389/fnhum.2014.00599 PMID:
25161615
32. Haken H. An approach to self-organization. Self-organizing systems: The emergence of order. New
York; 1987. pp. 417–437.
33. Kelso JAS. Dynamic patterns: the self-organization of brain and behavior. Complex Adaptive Systems
series. MIT Press, Bradford Book, Cambridge, MA; 1995.
34. Hristovski R, Davids K, Araújo D, Passos P. Constraints-induced emergence of functional novelty in
complex neurobiological systems: a basis for creativity in sport. Nonlinear Dynamics Psychol Life Sci.
2011; 15: 175–206. PMID: 21382260
35. Araujo D, Davids K, Hristovski R. The ecological dynamics of decision making in sport. Psychol Sport
Exerc. Elsevier; 2006; 7: 653–676. https://doi.org/10.1016/j.psychsport.2006.07.002
36. Hristovski R, Davids K, Araújo D. Information for Regulating Action in Sport: Metastability and Emer-
gence of Tactical Solutions. In: Araujo D, Ripoll H, Raab M, editors. Perspectives on Cognition and
Action in Sport. New York; 2009. pp. 43–57.
37. Santos SDL, Memmert D, Sampaio J, Leite N. The Spawns of Creative Behavior in Team Sports: A Cre-
ativity Developmental Framework. Front Psychol. 2016; 7: 1282. https://doi.org/10.3389/fpsyg.2016.
01282 PMID: 27617000
38. Torrents C, Ric A, Hristovski R. Creativity and emergence of specific dance movements using instruc-
tional constraints. Psychol Aesthetics, Creat Arts. 2015; 9: 65–74. https://doi.org/10.1037/a0038706
39. Hristovski R, Davids K, Passos P, Araújo D. Sport Performance as a Domain of Creative Problem Solv-
ing for Self-Organizing Performer-Environment Systems. Open Sports Sci J. 2012; 5: 26–35. https://
doi.org/10.2174/1875399X01205010026
40. Travassos B, Gonçalves B, Marcelino R, Monteiro R, Sampaio J. How perceiving additional targets
modifies teams’ tactical behavior during football small-sided games. Hum Mov Sci. Elsevier B.V.; 2014;
38: 241–50. https://doi.org/10.1016/j.humov.2014.10.005 PMID: 25457422
41. Aguiar M, Gonçalves B, Botelho G, Lemmink K, Sampaio J. Footballers’ movement behaviour during 2-
, 3-, 4- and 5-a-side small-sided games. J Sports Sci. Routledge; 2015; 33: 1259–1266. https://doi.org/
10.1080/02640414.2015.1022571 PMID: 25782702
42. Silva P, Esteves P, Correia V, Davids K, Araújo D, Garganta J. Effects of manipulations of player num-
bers vs. field dimensions on inter-individual coordination during small-sided games in youth football.
International Journal of Performance Analysis in Sport. 2015. pp. 641–659.
43. Vilar L, Araújo D, Davids K, Bar-Yam Y. Science of winning soccer: Emergent pattern-forming dynamics
in association football. J Syst Sci Complex. 2013; 26: 73–84. https://doi.org/10.1007/s11424-013-2286-
z
44. Sampaio JE, Lago C, Gonçalves B, Maçãs VM, Leite N. Effects of pacing, status and unbalance in time
motion variables, heart rate and tactical behaviour when playing 5-a-side football small-sided games. J
Sci Med Sport. 2014; 17: 229–233. https://doi.org/10.1016/j.jsams.2013.04.005 PMID: 23683687
45. Silva P, Travassos B, Vilar L, Aguiar P, Davids K, Araújo D, et al. Numerical relations and skill level con-
strain co-adaptive behaviors of agents in sports teams. PLoS One. 2014; 9: e107112. https://doi.org/10.
1371/journal.pone.0107112 PMID: 25191870
46. Vilar L, Esteves P, Travassos B, Passos P, Lago-Peñas C, Davids K. Varying numbers of players in
small-sided soccer games modifies action opportunities during training. Int J Sport Sci Coach. Multi Sci-
ence Publishing; 2014; 9: 1007–1018.

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 15 / 16

 Dynamics of tactical behaviour

47. Gonçalves B, Marcelino R, Torres-Ronda L, Torrents C, Sampaio J. Effects of emphasising opposition

and cooperation on collective movement behaviour during football small-sided games. J Sport Sci.
2016; 34: 1346–1354. https://doi.org/10.1080/02640414.2016.1143111 PMID: 26928336
48. Ric A, Hristovski R, Gonçalves B, Torres L, Sampaio J, Torrents C. Timescales for exploratory tactical
behaviour in football small-sided games. J Sports Sci. Routledge; 2016; 34: 1723–1730. https://doi.org/
10.1080/02640414.2015.1136068 PMID: 26758958
49. Bourbousson J, Deschamps T, Travassos B. From Players to Teams: Towards a Multi-Level Approach
of Game Constraints in Team Sports. Int J Sport Sci Coach. 2014; 9(6): 1393–1406. https://doi.org/10.
1260/1747-9541.9.6.1393
50. Aguiar MVD, Botelho GMA, Gonçalves BS V, Sampaio JE. Physiological responses and activity profiles
of football small-sided games. J Strength Cond Res. LWW; 2013; 27: 1287–1294. https://doi.org/10.
1519/JSC.0b013e318267a35c PMID: 22820211
51. Di Salvo V, Baron R, Tschan H, Calderon Montero FJ, Bachl N, Pigozzi F. Performance characteristics
according to playing position in elite soccer. Int J Sports Med. 2007; 28: 222–227. https://doi.org/10.
1055/s-2006-924294 PMID: 17024626
52. Silva P, Vilar L, Davids K, Araújo D, Garganta J. Sports teams as complex adaptive systems: manipulat-
ing player numbers shapes behaviours during football small-sided games. Springerplus. Springer Inter-
national Publishing; 2016; 5: 191. https://doi.org/10.1186/s40064-016-1813-5 PMID: 27026887
53. Duarte R, Araújo D, Davids K, Travassos B, Gazimba V, Sampaio JE. Interpersonal coordination ten-
dencies shape 1-vs-1 sub-phase performance outcomes in youth soccer. J Sports Sci. 2012; 30: 871–
877. https://doi.org/10.1080/02640414.2012.675081 PMID: 22545768
54. Costa IT, Garganta J, Greco PJ, Mesquita I, Maia J. System of tactical assessment in Soccer (FUT-
SAT): Development and preliminary validation. System. 2011; 7: 69–83. https://doi.org/10.6063/
motricidade.7(1).121
55. Hristovski R, Davids K, Araújo D, Passos P, Torrents C, Aceski A, et al. Creativity in sport and dance:
Ecological dynamics on a hierarchically soft-assembled perception-action landscape. In: Davids K,
Hristovski R, Araújo D, Balagué N, Button C, Passos P, editors. Complex systems in sport. London:
Routledge; 2013. pp. 261–274.
56. Sibani P, Dall J. Log-Poisson statistics and full aging in glassy systems. EPL (Europhysics Lett. IOP
Publishing; 2003; 64: 8. https://doi.org/10.1209/epl/i2003-00109-0
57. Balescu R. Equilibrium and nonequilibrium statistical mechanics. NASA STI/Recon Tech Rep A. 1975;
76: 32809.
58. Hopkins WG, Marshall SW, Batterham AM, Hanin J. Progressive statistics for studies in sports medicine
and exercise science. Med Sci Sports Exerc. 2009; 41: 3–12. https://doi.org/10.1249/MSS.
0b013e31818cb278 PMID: 19092709
59. Silva P, Aguiar P, Duarte R, Davids K, Araújo D, Garganta J. Effects of Pitch Size and Skill Level on
Tactical Behaviours of Association Football Players During Small-Sided and Conditioned Games. Int J
Sports Sci Coach. 2014; 9: 993–1006. https://doi.org/10.1260/1747-9541.9.5.993
60. Pinder R a., Davids K, Renshaw I. Metastability and emergent performance of dynamic interceptive
actions. J Sci Med Sport. Sports Medicine Australia; 2012; 15: 437–443. https://doi.org/10.1016/j.
jsams.2012.01.002 PMID: 22326853
61. Passos P, Araújo D, Davids K, Gouveia L, Serpa S, Milho J. Interpersonal pattern dynamics and adap-
tative behavior in multiagent neurobiological systems: conceptual model and data. J Mot Behav. 2009;
41: 445–459. 10.3200/35-08-061 https://doi.org/10.3200/35-08-061 PMID: 19482724
62. Torrents C, Ric A, Hristovski R, Torres-Ronda L, Vicente E, Sampaio J. Emergence of Exploratory,
Technical and Tactical Behavior in Small-Sided Soccer Games when Manipulating the Number of
Teammates and Opponents. PLoS One. Public Library of Science; 2016; 11: e0168866. https://doi.org/
10.1371/journal.pone.0168866 PMID: 28005978
63. Gama J, Passos P, Davids K, Relvas H, Ribeiro J, Vaz V, et al. Network analysis and intra-team activity
in attacking phases of professional football. Int J Perform Anal Sport. 2014; 14: 692–708.
64. Grund TU. Network structure and team performance: The case of English Premier League soccer
teams. Soc Networks. 2012; 34: 682–690. https://doi.org/10.1016/j.socnet.2012.08.004

PLOS ONE | https://doi.org/10.1371/journal.pone.0180773 July 14, 2017 16 / 16

 7. Discusión
La conceptualización teórica para la comprensión del fútbol como un fenómeno
complejo, dinámico y no lineal ha sido ampliamente narrada en la literatura (Davids,
Araujo y Shuttleworth, 2005; Gréhaigne, Bouthier y David, 1997; McGarry, Anderson,
Wallace, Hughes y Franks, 2002). Recientemente, se han aportado evidencias empíricas
que han explicado la coordinación interpersonal a diferentes niveles de análisis en el fútbol
masculino (Duarte, 2012) y la emergencia de comportamiento táctico a partir de la
manipulación de constreñimientos (Silva, 2014), sin embargo consideramos imprescindible
profundizar en la compresión de la dinámica inherente a estos procesos. Para ello la
distribución de probabilidades y correlaciones de las configuraciones de acción ha
permitido capturar la dinámica del comportamiento táctico. Esta dinámica se caracteriza
por su metaestabilidad, es decir, por la estabilidad temporal de los diferentes
comportamientos del sistema en escalas temporales cortas, pero su inestabilidad a largo
plazo. Esto, además, ha permitido corroborar la estructura jerárquica de la variabilidad de
comportamiento táctico, determinada por su organización en diferentes niveles en función
de la coherencia (similitud) entre comportamientos y la evolución de éstos en escalas
temporales distintas. Por otro lado, la manipulación de constreñimientos se postula como
una estrategia, de validez empírica, para diseñar contextos de juego que estimulen
comportamientos más o menos variados y flexibles dependiendo de los objetivos del
entrenamiento. Concretamente, en esta tesis se han manipulado los desequilibrios
numéricos en juegos reducidos de fútbol masculino, a nivel individual y colectivo, en
futbolistas profesionales y amateurs; y la limitación del espacio de interacción durante
situaciones que involucran un número de jugadores y espacio de juego mayor.

7.1. Síntesis de los principales hallazgos

En el estudio I se ha investigado la dinámica de comportamiento táctico durante un

partido. Lo resultados confirmaron la hipótesis de que las escalas temporales en las que

86
evolucionan un conjunto de variables tácticas derivadas de los datos de posición de los
jugadores (Duarte, Araújo, Folgado, et al., 2012) definen diferentes niveles de organización
dentro de la dinámica de comportamiento táctico. Concretamente, el índice de dispersión
(Stretch index) y la localización del centro geométrico del equipo evolucionaron en una
escala temporal superior a veinte segundos. El centro geométrico del equipo permaneció
localizado en un sector y carril del campo durante un promedio de 22 y 37 segundos,
respectivamente, durante la primera parte, y de 24 y 30 durante la segunda. Precisamente la
estabilidad temporal de estas variables (prolongación de una categoría activa conforme
transcurre el tiempo) fueron las que arrojaron correlaciones significativas en el último
orden de componentes principales indicando las bases sobre las que se estructura el
comportamiento táctico del equipo. Curiosamente la alternancia en la posesión del balón
transcurrió en una escala temporal similar por lo que las dinámicas de cambio de estas
variables pueden estar asociadas a ella (Bourbousson, Seve, et al., 2010; Frencken et al.,
2011). En cambio, las variables que mostraban la velocidad de expansión y contracción y
de movimiento del centro geométrico del equipo (en metros por segundo) cambiaron en una
escala de pocos segundos y únicamente arrojaron valores significativos en los componentes
principales extraídos en el primer nivel de análisis. A pesar de que los once jugadores que
finalizaron la primera parte fueron distintos a los once que terminaron la segunda,
existieron tendencias colectivas similares. Por ejemplo, en ambas partes se pudo identificar
que el equipo se replegaba hacia un posicionamiento intermedio en torno al centro del
campo basculando ligeramente hacia la izquierda, sin embargo en la primera parte lo
hicieron en campo contrario desde el carril central, mientras que en la segunda en campo
propio desde el carril derecho. Esto, entre otros aspectos, puede indicar que el equipo
contrario atacaba preferentemente por su banda derecha. El hecho de que este patrón sea
defensivo no implica que el equipo estuviera la mayor parte del tiempo defendiendo, sino
que los comportamientos defensivos de los jugadores son más estables (mayor correlación)
que los ofensivos (Travassos, Araújo, Duarte y McGarry, 2012; Travassos et al., 2011). A
pesar de ello este tipo de análisis también permitió identificar patrones ofensivos e
inmediatos a la recuperación o pérdida del balón así como de transición entre ellos. Por otro
lado, la correlación entre estos patrones de primer orden captura la estructura jerárquica de
variabilidad de comportamiento que culmina con la obtención de una base global de

87
comportamiento táctico que atrae al sistema. Por lo tanto, la emergencia de patrones
tácticos se realiza de forma totalmente correlacionada, es decir que las bases de
comportamiento que atraen al sistema están integradas en otras, formando subatractores
dentro de atractores. En resumen, la correlación horizontal en la dinámica de
comportamiento táctico forma y estabiliza patrones de comportamiento en un orden
superior (vertical), que a su vez gobierna aquellos de orden inferior, los cuales residen en
escalas espacio-temporales más pequeñas. Por todo ello, en la adquisición y desarrollo de
comportamiento táctico no existe una ruta o dirección preferida (de arriba hacia abajo o de
abajo hacia arriba), sino que debe entenderse como una relación circular resultante de la
influencia recíproca entre los diferentes nivel o escalas espacio-temporales (Haken, 1983).
Estos resultados confirmarían pues una estructura jerárquica de la dinámica de
comportamiento en deportes colectivos, previamente ya identificada en deportes o prácticas
motrices de oposición (Hristovski et al., 2011) y de cooperación (Torrents et al., 2015).

Bajo la influencia de ciertos constreñimientos, en el caso anterior los derivados de la

competición (modelo de juego, instrucciones del entrenador, momento de la temporada,
affordances, etc.), el sistema jugador-equipo no puede explorar arbitrariamente la totalidad
de posibilidades de acción, sintiéndose presionado y provocando la aparición de réplicas de
comportamiento, es decir, comportamientos similares, por lo que ciertas configuraciones de
comportamiento, individual o colectivo, correlacionarán más entres sí o se darán con una
mayor probabilidad que otras. En el estudio II y III se manipuló el número de oponentes en
una situación de juego reducido con el objetivo de investigar cómo este constreñimiento de
la tarea puede moldear la estructura dinámica de comportamiento táctico de los jugadores,
así como determinar los efectos en la variedad y flexibilidad de comportamiento. En el
estudio II, donde se estudió únicamente a jugadores profesionales, la dinámica exploratoria
de comportamiento táctico tuvo una evolución similar en las tres situaciones estudiadas
(4x3, 4x5 y 4x7). Los jugadores exploraron configuraciones de movimiento cuya similitud
disminuía a medida que aumentaba el lapso de tiempo entre ellas, evolucionando en una
escala temporal de segundos. En una escala de unas pocas decenas de segundos, la
dinámica exploratoria se estabilizó, alcanzando valores de similitud estacionarios.
Finalmente, en escalas de tiempo aún mayores (minutos), el grado de exploración fue
estacionario, lo que significa que los comportamientos de los jugadores separados por

88
lapsos de tiempo de más de sesenta segundos presentaron altos valores de similitud. Por lo
tanto, la dinámica exploratoria muestra tendencias distintas dependiendo de la escala
temporal, lo que sugiere nuevamente una dinámica metaestable, o lo que es lo mismo,
temporalmente estable a corto plazo pero inestable a largo. El cálculo del tiempo promedio
(en segundos) que los jugadores tardaron en cambiar de acción táctica, contexto de
interacción, zona del campo o velocidad de movimiento (τ1 = 3,09 ± 3,01 seg.); y el
correspondiente al conjunto de acciones ofensivas o defensivas hasta que el equipo perdía o
recuperaba el balón (τ2 = 12,71 ± 9,32 seg.) corroboró la existencia de escalas temporales
significativamente separadas, definiendo inequívocamente dos niveles dentro de la
dinámica de comportamiento táctico. La representación de las probabilidades condicionales
de las configuraciones de comportamiento en un paisaje potencial permitió comparar la
dinámica metaestable de las tres situaciones en desequilibrio numérico. La rugosidad de los
diferentes paisajes mostró claramente como la situación en superioridad numérica provocó
una dinámica de comportamiento más flexible, mientras que las situaciones en inferioridad
numérica mostraron un paisaje mucho más abrupto. Esto supuso una dinámica de
comportamiento más rígida al jugar en inferioridad, donde los jugadores tendieron a
realizar principalmente coberturas ofensivas y balances defensivos en campo propio. Los
resultados del promedio de los valores estacionarios de la dinámica exploratoria, entropía y
fuerza de atrapamiento de las configuraciones (la probabilidad de ser capaz de repetir las
mismas configuraciones secuencialmente) también demostraron que a medida que
aumentaba el número de oponentes el comportamiento táctico individual fue menos
variado, irregular (impredecible) y flexible.

Curiosamente, en el estudio III, teniendo en cuentas las acciones técnico-tácticas de

todos los jugadores de un mismo equipo, se pudieron apreciar tendencias distintas a las
anteriores. A nivel colectivo, los equipos formados por cuatro jugadores profesionales
mostraron comportamientos significativamente más similares cuando jugaron en
superioridad numérica. O dicho de otra manera, a medida que aumentaba el número de
oponentes aumentaba la amplitud exploratoria, es decir, la variedad de comportamiento.
Estos resultados pueden parecer contradictorios con los del estudio anterior, sin embargo
Duarte y colaboradores (2013) encontraron valores más bajos de irregularidad (mayor
estabilidad) en la sincronización colectiva de los jugadores, en comparación con los

89
obtenidos en la sincronización de cada jugador con el equipo. Estos resultados sugieren que
la estabilidad del comportamiento a nivel colectivo puede compensar los valores superiores
de irregularidad (entropía) del movimiento a nivel individual, estableciéndose algún tipo de
compensación recíproca entre los comportamiento individuales y el colectivo (Riley et al.,
2011). Más adelante entraremos en detalle sobre este tipo de sinergia multinivel
considerándola de gran importancia, ya no solo a nivel teórico, sino también a nivel
práctico.

En este tercer estudio también se comparó la amplitud exploratoria en equipos

amateurs. Mientras en uno de los equipos la amplitud exploratoria fue significativamente
inferior en la situación en superioridad comparada con las otras dos en inferioridad, en el
otro, a pesar de seguir la misma tendencia que en los profesionales, únicamente se
encontraron diferencias entre las dos situaciones en inferioridad. Además, se comparó a los
equipos cuyo número de jugadores fue variando, los equipos de 3, 5 y 7 jugadores. La
tendencia en general fue que al aumentar el número de compañeros disminuyó la amplitud
exploratoria a nivel de respuestas técnico-tácticas, al igual que ocurrió en relación a la
irregularidad de distintas variables de posicionamiento de los jugadores (Aguiar et al.,
2015; Gonçalves et al., 2016). Sin embargo, comparando la relación numérica de +/- 1 (5 y
3 compañeros respectivamente) solo se apreció un efecto moderado positivo favorable a
uno de los equipos profesionales en superioridad numérica, lo que puede suponer que los
jugadores u oponentes puedan compensar la falta de un jugador (Silva, Travassos, et al.,
2014). Finalmente es importante destacar que cuando los equipos jugaron con tres
jugadores más que el oponente (7vs.4) disminuyó significativamente la amplitud
exploratoria (p<0,01; d=2,99±0,81), lo que tendrá fuertes connotaciones a la hora de
diseñar situaciones de juego reducido en las que se manipule la relación de desequilibrio
numérico entre los equipos, ya que contextos excesivamente favorables pueden disminuir la
cantidad, variedad y/o flexibilidad de soluciones o comportamientos. En contraposición con
la utilización de constreñimientos más restrictivos (gran inferioridad numérica), que
parecen estimular el comportamiento creativo del equipo (Rosso, 2014).

Finalmente, la aplicación del análisis jerárquico de componentes principales reveló

la estructura jerárquica de comportamiento colectivo. En este caso cada configuración

90
estuvo formada por la combinación de acciones desarrolladas por los jugadores en cada
segundo de juego. La correlación entre las configuraciones del equipo en un primer nivel de
análisis y la aplicación secuencial del mismo permitió capturar la esencia de
comportamiento colectivo constituyendo la base global de atracción en la división o
coordinación de tareas entre los jugadores (Duarte, Arajo, Correia y Davids, 2012; Eccles,
2010). Precisamente la heterogeneidad de las tareas fue una de las características esenciales
que mostró el comportamiento de los jugadores independientemente del aumento de
oponentes o compañeros. Sin embargo, lo que se pudo apreciar con el aumento del número
de oponentes es que el número de acciones que mostraron correlaciones significativas fue
superior, lo que supone un aumento de las posibilidades de combinar acciones por parte del
equipo. A nivel defensivo, disminuyeron las acciones de disuasión en pro de dar balance
defensivo. La gran superioridad numérica obligó a los jugadores a proteger el espacio
cercano a la portería disminuyendo la distancia al oponente más cercano seguramente
porque más jugadores trataron de buscar líneas de pase interiores, lo que generó amplitud
por fuera (Gonçalves et al., 2016). La alta densidad de oponentes mermó las posibilidades
de pase, probablemente por ello el recurso ofensivo del pase dejo de ser tan significativo
con el aumento del número de oponentes. Por el contrario, el incremento de compañeros
disminuyó el número de correlaciones significativas entre acciones, lo que muestra una
mayor rigidez en la base general de comportamiento colectivo, confirmando los resultados
obtenidos de amplitud exploratoria del equipo. Concretamente, el incremento del número
de compañeros hizo que el otro equipo se replegara, posiblemente por ello las acciones de
desmarque dejaron de aparecer en la estructura de comportamiento. Defensivamente, el
aumento de compañeros supuso que más jugadores dejaran de tener un par, es decir, que de
uno a tres jugadores centraron sus tareas defensivas en dar balance defensivo sosteniendo
las acciones individuales de los compañeros que buscaban la recuperación del balón.
Aunque con baja resolución, la base global de coordinación colectiva de acciones
individuales crea un núcleo de comportamiento colectivo sobre el que se vertebran
combinaciones de acciones más concretas. Éstas se desarrollan en una escala temporal más
corta, debido a los sutiles y rápidos cambios de constreñimientos físicos e informacionales
que guían las reconfiguraciones y adaptación del equipo.

91
 Los juegos reducidos en desequilibrio numérico son tareas habituales de
entrenamiento, sin embargo, se ha demostrado que, durante un partido, este tipo de
desequilibrios a los que se enfrentan los jugadores varía a lo largo del tiempo y es diferente
dependiendo de la posición que ocupe cada jugador (Vilar, Araújo, Davids y Bar-Yam,
2013). Por ello, en el estudio IV, manipulando la restricción del espacio en el que los
jugadores podían actuar, se pretendía provocar situaciones de desequilibrio numérico con el
objetivo de conocer si este constreñimiento provocaba contextos favorables para actuar con
el balón, y conocer sus efectos sobre la dinámica de pases y sobre la dinámica de
comportamiento exploratorio. Los paisajes potenciales de acción del jugador con balón,
definido por la distancia respecto al oponente más cercano y a la zona de marca, arrojó de
forma probable y muy probable diferencias significativas entre restringir el espacio de
interacción a los jugadores y hacerlo de forma parcial o no restringirlo. El hecho de
restringir el espacio dio como resultado un paisaje mucho más abrupto, lo que supone un
comportamiento más rígido que en las otras dos situaciones. A pesar de las diferencias
generales, la probabilidad de que los jugadores del equipo tuvieran el balón en zonas
avanzadas libres de marca (> 5 metros) fue prácticamente nula, por lo que se puede decir
que fue un estado inestable independientemente de los constreñimientos impuestos por el
entrenador. La limitación del espacio de interacción hizo que aumentara el tiempo
promedio en el que lo jugadores poseían el balón, aunque los resultados únicamente
mostraron posibles diferencias significativas entre la limitación parcial del espacio de
interacción y el juego libre. Las restricciones espaciales de los jugadores incrementaron la
relación espacio-temporal del atacante con el balón y el oponente más cercano aunque esto
únicamente se tradujo en una mayor distribución de probabilidades de pase entre
compañeros en la situación con limitación parcial. En cuanto a la dinámica exploratoria, la
restricción total del espacio de interacción disminuyó la amplitud de exploración de los
jugadores. Sin embargo, aumentó el ratio o rapidez de exploración en escalas temporales
más cortas, es decir, las reconfiguraciones de los jugadores fueron mayores en intervalos de
tiempo menores.

92
 7.2. Aportaciones teóricas y metodológicas

El fútbol es un sistema o fenómeno complejo porque implica a un conjunto de

componentes que interaccionan entre ellos (jugadores o jugadoras) como un todo, y con su
entorno (tarea/partido), y de la cual emergen propiedades diferentes a las de los
componentes que lo forman. Además es dinámico porque, como cualquier otro fenómeno
relacionado con el deporte, cambia conforme pasa el tiempo, aunque estos cambios puedan
darse en escalas temporales diferentes. A diferencia de los sistemas lineales, los no lineales
se caracterizan por la no proporcionalidad de respuestas ante un mismo conjunto de
constreñimientos o situación dada. Es decir, un pequeño cambio de constreñimientos puede
suponer un gran cambio en la variable que captura el comportamiento y viceversa. Los
sistemas no lineales, por lo tanto, posibilitan la aparición de múltiples soluciones o
alternativas de comportamiento.

La presión del conjunto de variables intrínsecas o extrínsecas, es decir

constreñimientos personales o del entorno, hará que el sistema converja hacia múltiples
estados de atracción metaestables, es decir, estados temporalmente estables, pero inestables
a largo plazo. La metaestabilidad es una propiedad fundamental para comprender las
dinámicas de comportamiento de sistemas complejos (Bovier, 2006).

En el fútbol masculino el parámetro de orden fase relativa (la diferencia en las fases
de oscilación de dos variables continuas) ha sido el más utilizado para capturar el
comportamiento coordinado o estados coordinativos que surgen de la interacción jugador-
jugador o equipo-equipo (Duarte, Araújo, Davids, et al., 2012; Folgado et al., 2014, 2015;
Gonçalves, Figueira, Maçãs y Sampaio, 2014; Silva, Chung, et al., 2016). A partir de
oscilaciones en el posicionamiento de dos jugadores o variables derivadas del equipo en su
conjunto, como el centro geométrico del equipo o el índice de dispersión, la fase relativa
describe los patrones espacio-temporales de la coordinación rítmica y los cambios en la
sincronización que se producen en respuesta a los constreñimientos del juego. Las
correlaciones cruzadas (cross-correlations) y vector coding (Moura et al., 2016), las
running correlations (Duarte, Araújo, Freire, et al., 2012) o el análisis cluster phase
(Duarte et al., 2013) también han permitido detectar la existencia de estos estados
metaestables siendo identificados como valores del parámetro de orden. Estos tipos de
93
análisis permiten detectar la simetría entre dos variables continuas. Sin embargo, la
heterogeneidad de las tareas ejecutadas por los jugadores (acciones técnico-tácticas) y su
desarrollo en diferentes situaciones, es decir, zona del campo, posicionamiento relativo
entre compañeros y oponentes o la posesión o no del balón, entre otras, son variables
discretas que, generalmente, no pueden ser analizadas como simetrías simples o espejo
(ɛ·ɜ), o reajustes en una variable continua. Por eso el parámetro de orden overlap q permite
medir la simetría o similitud (correlación) entre este tipo de variables discretas (Torrents et
al., 2015), así como entre variables continuas, e incluso la combinación de ambas. En este
sentido este parámetro de orden puede ser más general y versátil, y puede utilizarse en
casos multidimensionales, es decir, que involucren a más de dos variables, y que no sean
exclusivamente continuas. Además es muy importante destacar que captura la dinámica en
múltiples escalas temporales, ya que recoge los valores de similitud en lapsos de tiempo
incrementales (Hristovski et al., 2013).

Los diferentes análisis de datos utilizados en esta tesis abren un amplio abanico de
posibilidades ofreciendo nuevas herramientas y procedimientos estadísticos para estudiar el
comportamiento táctico. Por un lado, el análisis jerárquico de componentes principales
permite identificar patrones combinando múltiples variables que pueden definir el
comportamiento táctico colectivo o individual. En el estudio I se presenta una selección de
las variables más estudiadas en la literatura reciente, sin embargo, este tipo de análisis
permite la utilización de aquellas que puedan tener mayor interés para un investigador o
investigadora, o incluso aquellas que los miembros de un staff técnico puedan considerar
como variables de rendimiento. En el estudio III se pudo identificar la correlación existente
entre las acciones que realizan los diferentes jugadores de un equipo y las de los oponentes,
de manera que se puede obtener información relevante para verificar que los jugadores de
un equipo hacen una presión adecuada respetando los diferentes roles derivados de ella, o si
ofensivamente los compañeros sin balón ofrecen diferentes soluciones al portador. Aunque
los resultados no fueron publicados, este tipo de análisis también se pudo aplicar sobre las
matrices de datos utilizados en los estudios II y IV, de los que se podría haber identificado
la correlación entre las acciones de cada jugador con otras variables contextuales como la
zona del campo donde se realizó, la posición relativa respecto a compañeros y oponentes, la
distancia a la portería y al oponente más cercano, y la velocidad de movimiento al realizar

94
cada acción, permitiendo obtener una información más contextualizada de la acción
individual de cada jugador (McGarry, 2009). Sin embargo, a partir de las probabilidades
condicionales de las configuraciones de comportamiento derivadas de la combinación de
dos de las variables anteriores, se definió el paisaje potencial de comportamiento de cada
jugador. Estos paisajes permiten dar información sobre aquellos estados más estables o
comportamientos que atraen al sistema (jugador/equipo), es decir que poseen una alta
probabilidad de que se den, así como los más inestables. Además, la rugosidad de estos
paisajes permite identificar qué tareas o modificaciones de ellas promueven
comportamientos más variados y/o flexibles, así como cuales favorecen la estabilización de
ciertos comportamientos, en función de la localización y la profundidad de las bases de
atracción del comportamiento.

En esta misma línea, la dinámica exploratoria, como ya se ha explicado

anteriormente, puede mostrar tendencias distintas derivadas de los parámetros alfa y qstat
que permiten identificar el ratio o velocidad de exploración y la amplitud exploratoria del
jugador o equipo, respectivamente. Por un lado, el parámetro alfa que muestra la pendiente
de la curva da información sobre la flexibilidad del comportamiento, ya que, contra mayor
sea la pendiente más rápido explorará la amplitud total de posibilidades de acción definida
por el parámetro qstat. Es importante tener en cuenta que el valor de la pendiente α y el valor
estacionario qstat no están teóricamente conectados, por lo que la combinación de estos
permite definir diferentes perfiles de deportistas o clasificar diferentes tipos de ejercicios o
constreñimientos en función de la evolución del parámetro de orden (<qd(t)>).

7.3. Aplicaciones prácticas

En esta sección discutiremos algunas de las aplicaciones prácticas que se derivan de

la presente tesis, haciendo referencia al aprendizaje motor, la planificación y diseño del
entrenamiento, y el análisis del rendimiento.

Se puede decir que el fútbol, como otros deportes o prácticas motrices que se
desarrollan en entornos con gran incertidumbre y donde, por lo tanto, prima la adaptación,
debe ser pensado como un banco emergente de comportamiento donde deportista y/o
equipo continuamente exploran diferentes soluciones sobre las múltiples tareas o

95
situaciones impuestas por el comportamiento del equipo oponente. Tradicionalmente, los
programas de formación y entrenamiento se han basado en la prescripción de movimientos
predeterminados proporcionando una serie de técnicas o movimientos preestablecidos para
que el jugador o jugadora sea capaz de reproducirlos durante la competición. Sin embargo,
tal y como se ha mostrado, el comportamiento no se puede aislar de su contexto, es decir,
una misma acción será diferente en función de si se realiza en campo propio o rival, si un
jugador es el último o si está situado entre compañeros y oponentes, si el oponente más
cercano está a 2 o a 10 metros, o que a pesar de su correcta ejecución, ésta provoque un
contexto desfavorable para la consecución del objetivo a corto o largo plazo. En el fútbol,
por lo tanto, se pueden identificar una amplia variedad de comportamientos que vienen
definidos por su contexto. Dentro de esa variedad se pueden identificar configuraciones o
patrones concretos sin necesidad de ser impuestos por un agente externo (staff técnico o
docente). Así, mediante la manipulación de constreñimientos se pueden diseñar contextos
favorables que provoquen cambios significativos en la distribución de probabilidades de
comportamientos concretos modulando el paisaje potencial de acción de jugadores,
jugadoras o equipos. La modificación de uno o varios parámetros de control en la dirección
y con la intensidad adecuada puede reducir la probabilidad de que se den comportamientos
no deseados o incrementar la de los deseados para que sean más estables. Sin embargo, se
debe considerar que si desarrollamos comportamientos excesivamente estables aumentará
la rigidez del sistema perdiendo la capacidad de adaptación ante situaciones imprevistas.
Junto con la estabilidad otra característica fundamental para tener una buena capacidad de
adaptación es la flexibilidad. La estimulación de ambas durante el entrenamiento permitirá
la exploración de múltiples soluciones funcionales de mayor o menor estabilidad. El
sistema deportista-equipo deberá ser lo suficientemente flexible como para que, ante
comportamientos del oponente que disminuyan la funcionalidad de su comportamiento, sea
capaz de transitar hacia otro comportamiento o solución funcional.

Uno de los principales hallazgos de esta tesis es que en el fútbol masculino la

dinámica exploratoria de los jugadores y equipos se estabiliza en una escala temporal de
alrededor de 30-40 ±10 segundos, es decir, los comportamientos separados por 45 segundos
muestran el mismo grado de similitud que los desarrollados entre el primer y el último
segundo de juego. Ante estos resultados se decidió hacer una simulación para ver cómo

96
podría afectar la perturbación de la dinámica de juego a partir de manipulación del número
de jugadores en estas escalas temporales donde la exploración tiende a saturarse (Ric,
Hristovski y Torrents, 2015). Los resultados mostraron que provocar desequilibrios
numéricos de diferente magnitud en periodos de 15, 30, 45 y 60 segundos permite a los
jugadores ampliar su capacidad exploratoria, es decir, posibilita una mayor variedad de
comportamientos. Esto implica que si el objetivo del entrenador o entrenadora es
desarrollar más flexibilidad de comportamiento en sus equipos, la modificación constante
en breves periodos de tiempo de las condiciones iniciales de juego durante el desarrollo de
la tarea puede ser una estrategia interesante. Esta estrategia favorecerá la exploración de
soluciones en un entorno donde los desequilibrios cambian temporalmente, provocando
mayor o menor incertidumbre en función de la posición y variación del espacio efectivo de
juego (Vilar et al., 2013),.

Los resultados de los estudios II y III muestran como los mismos constreñimientos
pueden provocar diferentes tendencias a nivel individual y colectivo. Si aumentamos el
número de oponentes el comportamiento exploratorio de los jugadores a nivel individual
aumenta, sin embargo a nivel colectivo disminuye. Estos resultados pueden parecer
contradictorios pero muestran una compensación recíproca o sinergia multinivel. El
aumento de oponentes provoca configuraciones o patrones de comportamiento colectivos
mucho más estables, pero menos a nivel individual. El comportamiento jugando en
inferioridad numérica a nivel colectivo puede ser muy rígido, es decir, puede ser que haya
un jugador presionando, otro en cobertura, otro haciendo balance defensivo y un cuarto
replegando. Sin embargo, a nivel individual puede ser que cada jugador esté cambiando
constantemente de tarea, es decir, puede ser que el mismo jugador pase de estar haciendo
un balance defensivo a hacerle la cobertura a un compañero que ha pasado a presionar. Por
lo tanto, la superioridad del oponente puede implicar que el jugador a nivel individual deba
realizar múltiples tareas para favorecer la estabilidad a nivel colectivo. Estos hallazgos
hacen que se deba tener en cuenta el nivel del sistema (deportista o equipo) sobre el que se
quiere incidir, ya que la manipulación de la tarea puede provocar unos efectos a nivel
individual y no tenerlos o que sean contrarios a nivel colectivo, y viceversa.

97
 Tal y como se ha comentado anteriormente, la dinámica exploratoria se define por
el parámetro de orden <qd(t)>, y concretamente por los parámetros alfa y qstat, cuya
combinación podría determinar el perfil de comportamiento de los jugadores a diferentes
escalas temporales. A modo de ejemplo, en la figura 2 se diferencian cuatro combinaciones
posibles entre estos dos parámetros: a) valor qstat alto y valor α bajo (línea azul), lo que
significa que el jugador explora una región reducida dentro del abanico de posibilidades de
comportamientos a realizar y desarrolla configuraciones similares (acciones, zonas,
contexto, distancias, velocidades, etc.) de forma más prolongada en el tiempo; b) valor α y
qstat altos (línea verde), esto significa que el jugador cambia rápidamente de configuración
de movimiento, pero no explora un gran conjunto de configuraciones diferentes; c) valor α
y qstat bajos (línea amarilla), esto se asocia con un cambio lento de comportamientos, pero,
a largo plazo, una exploración de un amplio conjunto de configuraciones de movimiento; y
d) valor qstat bajo y valor α alto (línea roja), este perfil se asocia con cambios rápidos en las
configuraciones de movimiento y con la exploración de un amplio abanico de
comportamientos a largo plazo.

La identificación de este parámetro de orden de carácter comportamental y táctico

puede aplicarse para determinar el perfil de rendimiento de deportistas o equipos, para
monitorear la evolución del aprendizaje o control del entrenamiento, o para cuantificar los
efectos de los diferentes constreñimientos sobre el juego. Esto permitiría clasificar las
tareas del entrenamiento, algo que tradicionalmente ha venido determinado por parámetros
físicos y fisiológicos.

98
Figura 2. Cuatro perfiles de la superposición dinámica <qd(t)> para distancias de tiempo
incrementales definidos por los parámetros alfa y qstat. Su dinámica evoluciona en escalas
temporales distintas (de segundos a varios minutos) y no converge a 0 (comportamiento aleatorio)
durante el periodo de observación.

En la misma línea, el análisis jerárquico de componentes principales y su aplicación

desde momentos iniciales de la temporada permitirían identificar la evolución de modelo de
juego, verificando si las tareas de entrenamiento e instrucciones propuestas por el
entrenador o la entrenadora generan la estructura dinámica de comportamiento táctico
deseada. Además, la vertebración de patrones tácticos de cada partido permitiría identificar
si existen comportamientos comunes dependiendo de variables situacionales como pueden
ser el resultado final, jugar en casa o fuera, o jugar contra equipos situados en la parte alta o
baja de la clasificación. Del mismo modo se podría evaluar la respuesta táctica al
enfrentarse a equipos que desarrollen una presión avanzada o repliegue en campo propio,
defensivamente; o un juego de posición frente a otros equipos que propongan una estrategia
ofensiva basada en el juego directo o contraataque, por ejemplo. Disponiendo de datos de
posición de equipos rivales este tipo de análisis también permitiría identificar cuál es su
estructura de comportamiento táctico, para informar a los equipos sobre el rival y diseñar
las tareas de entrenamiento más adecuadas.

99
 7.4. Limitaciones y perspectivas de investigación futuras

A pesar de que este tipo de análisis podría tener transferencia inmediata a la realidad
de un club profesional, no resulta fácil obtener datos de posicionamiento para avanzar en el
análisis de rendimiento táctico. Actualmente, la última tecnología en sistemas de
seguimiento a través de video cámaras instaladas en la mayoría de estadios de los equipos
masculinos de las grandes ligas europeas permite conocer el posicionamiento de los
jugadores y del balón con una gran frecuencia de recogida, precisión y de forma
instantánea. Sin embargo, la tecnología GPS utilizada en esta tesis para recoger los datos de
posicionamiento impidió capturar el posicionamiento del balón durante las tareas
experimentales, y el posicionamiento del equipo rival durante el partido. El
posicionamiento del rival y del balón hubiera permitido comprender mejor la dinámica de
juego.

La localización del balón, así como el contexto que envuelve al que lo porta, tendrá
grandes implicaciones en el análisis del juego y obviamente será uno de los principales
condicionantes de la dinámica de comportamiento táctico. El hecho de tener que registrar
las acciones con balón y determinar el poseedor del balón de forma manual ralentizó la
preparación de los datos. Este hecho supone una limitación porque a pesar de que los datos
de posicionamiento de los jugadores se registraron con una frecuencia de 5 y 15 hercios
(Hz), tuvieron que ajustarse a 1 Hz para sincronizarlos con los datos obtenidos a través de
la observación sistemática y mantener unos valores de fiabilidad elevados. Si bien es cierto
que comprobamos que la similitud de configuraciones en una escala temporal inferior a 1
segundo mostró valores muy próximos al máximo (≈1), es decir, se mantuvo estacionaria
en una escala de décimas de segundo previa a la relajación de la curva (escala de
segundos). Esto implica que haberlo hecho con una frecuencia superior podría haber
alterado los resultados, ya que como se aprecia en la figura 3 la función de ajuste a la
distribución de los datos hubiera tenido que ser redefinida utilizado una función alternativa
(ej. stretched exponential function). Por ello, para futuros trabajos donde la frecuencia de
recogida sea mayor y pueda suponer una distribución diferente de los datos será necesario
utilizar funciones o modelos matemáticos distintos que se ajusten a dicha distribución.

100
Figura 3. Ejemplo de ajuste de la función de la curva teórica y datos experimentales.

Otro de los motivos por los que se utilizó una frecuencia de 1Hz es porque los
softwares estadísticos actuales no son capaces de procesar grandes matrices de datos. Por
ello, y para próximos trabajos que supongan el tratamiento de un mayor volumen de datos
será imprescindible desarrollar algoritmos que permitan el procesamiento de matrices
mayores y permitan disponer de los resultados con una mayor brevedad.

Del mismo modo, a pesar de que el análisis jerárquico de componentes principales

permite definir los patrones de comportamiento incluyendo más de dos variables, nuevas
técnicas de visualización podrían facilitar la interpretación y transmisión de los resultados.
En esta tesis se ha utilizado como técnica de representación el paisaje potencial, ya que
facilita la comprensión e interpretación de la dinámica metaestable de comportamiento de
deportistas, pero solo pueden introducirse dos variables.

A pesar de que este modelo conceptual ha permitido explicar y comprender la

estructura jerárquica y la dinámica intrínseca de fenómenos naturales complejos (Badii &
Politi, 1999), en el deporte únicamente encontramos algunos estudios recientes que lo han
utilizado (Hristovski et al., 2011; Torrents et al., 2015). Su aplicación en otros deportes
permitiría ahondar en la comprensión de la dinámica propia de cada deporte e identificar el
comportamiento emergente. La falta de estudios que profundicen en conocer los efectos de
la manipulación de las tareas hace que sea necesario continuar con esta línea de
investigación. Además, la manipulación dinámica o temporal de los constreñimientos

101
impuestos por el entrenador o la entrenadora todavía no ha estado sujeta a una investigación
sistemática, postulándose como una propuesta de estudio innovadora para el futuro.

Por otro lado, existe otro tipo de modelos matemáticos que podrían ser utilizados
para explorar los cambios discretos de patrones tácticos. Los modelos de cambio de
régimen (regime switching models), basándose en el enfoque de cadenas ocultas de Markov
(hidden Markov chain), permitirían estimar la probabilidad condicional de que se
produzcan cambios discretos entre diferentes regímenes (patrones) (Smith et al., 2017). Por
ejemplo, lateral y extremo intercambiando su posición en profundidad durante el juego, o la
alternancia en la presión de punta y media-punta en la salida de balón con defensas
centrales. Esto permitiría investigar la manera en que los comportamientos a nivel
individual responden a cambios cualitativos en un nivel superior del sistema. Es decir,
evaluar cómo los parámetros de control que evolucionan más lentamente (goles,
expulsiones, sustituciones, etc.) constriñen el comportamiento a nivel colectivo (valores de
parámetros de orden) que cambian más rápido, y cómo éstos constriñen las acciones
individuales de los jugadores o jugadoras que evolucionan más rápido, es decir, en una
escala temporal más corta.

El antes mencionado posicionamiento del balón podría ser considerado para estudiar
si el intercambio de pases entre deportistas sigue algún patrón en concreto. La anotación de
pases, aunque no durante el entrenamiento, ha permitido explorar las interacciones entre los
jugadores durante la competición mediante análisis de redes (Duch, Waitzman y Amaral,
2010; Gama et al., 2014; Gonçalves et al., 2017). Sin embargo, el conocimiento de la
longitud de esos pases permitiría explorar la posibilidad de que el desplazamiento del balón
pudiera seguir algún tipo de patrón que combine la alternancia de pases cortos con otros de
trayectorias más largas. Este tipo de patrones son parte de una estrategia general de
movimiento que adoptan muchas organismos o sistemas en una amplia variedad de
contextos, como por ejemplo, patrones identificados en moléculas de gas, patrones de caza
de tiburones o patrones de desplazamiento de personas en un parque de atracciones, y
reciben el nombre de vuelos de Levy. La distribución de distancias o longitud en los
desplazamientos (ej. pases) en forma de ley de potencia encierra en sí mismo una propiedad

102
fractal o de escala invariante, es decir, que estaríamos hablando de un patrón repetido a
diferentes escalas.

103
 8. Conclusiones
Las conclusiones generales de esta tesis son:

 Bajo la influencia de ciertos constreñimientos, la dinámica de comportamiento

forma una estructura jerárquica de patrones tácticos. Los patrones identificados en
cada nivel de jerarquía se caracterizan por aquellos valores del parámetro de orden
que emergen, persisten y se diluyen en diferentes escalas de tiempo. Los cambios de
comportamiento más rápidos, que evolucionan en escalas temporales cortas, se
situarán en un nivel más bajo de jerarquía; mientras que los comportamientos que
persistan en el tiempo, es decir que evolucionen en escalas temporales más largas,
definirán aquellos patrones en los niveles de jerarquía superiores.

 Para comprender la dinámica de comportamiento táctico en el fútbol será necesario

identificar la existencia de múltiples escalas temporales. Este hecho está relacionado
con la metaestabilidad, que se define por el desarrollo de un comportamiento estable
de forma temporal, en escalas cortas de tiempo, pero inestable a largo plazo. Es
decir, que en escalas temporales más largas, los/las deportistas o equipos
experimentan transiciones entre estados o comportamientos estables a corto plazo.

 La dinámica del comportamiento táctico se caracteriza por una sutil combinación

entre estabilidad (comportamientos persistentes) y flexibilidad (cambios de
comportamiento). El comportamiento individual o colectivo se basa en la continua
exploración de soluciones tácticas y respuestas motrices para adaptarse
funcionalmente a los constreñimientos del entorno que van apareciendo a diferentes
escalas temporales.

104
 La manipulación de la relación numérica en juegos reducidos tiene efectos
diferentes en la dinámica de comportamiento exploratorio del equipo e individual. A
nivel colectivo, los equipos que jugaron en superioridad mostraron una amplitud
exploratoria inferior, observándose mayores diferencias cuando aumentó el número
de compañeros. Sin embargo, a nivel individual, el aumento del número de
oponentes disminuyó la flexibilidad, variedad e incertidumbre de soluciones tácticas
individuales. En este sentido, se debe tener en cuenta que los mismos
constreñimientos pueden tener efectos diferentes sobre el comportamiento
exploratorio individual y colectivo. Esto también implica que existe algún tipo de
sinergia multinivel, donde el bajo grado de variedad táctica a nivel colectivo puede
ser compensado a nivel individual y viceversa.

 El análisis jerárquico de componentes principales permitió detectar tanto la esencia

de comportamiento técnico-táctico como los cambios en la estructura de
comportamiento colectivo provocados por el desequilibrio numérico. El aumento
del número de oponentes aumentó las posibilidades de combinar acciones,
sobretodo defensivamente. En cambio, el incremento de compañeros generó una
mayor rigidez en la base general de comportamiento colectivo. La imposición de
constreñimientos a priori limitantes puede promover la búsqueda de
comportamientos colectivos o combinaciones de tareas individuales nunca antes
explorados. Por lo tanto, situaciones de dificultad podrían favorecer la aparición de
soluciones novedosas y creativas.

 Análisis complementarios como análisis de redes, el cálculo del tiempo promedio en

la persistencia de comportamientos (dwell time), o el cálculo del potencial a partir
de las probabilidades condicionales de dos variables específicas han permitido
encontrar diferencias en la dinámica de comportamiento táctico manipulando la
restricción en el espacio de interacción. Limitar el movimiento de los jugadores a
zonas específicas aumenta el tiempo de retención del balón. Esto puede ser debido a
la dificultad de encontrar diversidad de líneas de pase o a la falta de flexibilidad del
comportamiento de los jugadores.

105
 Dependiendo de los objetivos del entrenador o de la entrenadora, las tareas de
entrenamiento mejorarán la exploración de soluciones tácticas o se centrarán en
estabilizar un conjunto más reducido de comportamientos. La manipulación
adecuada de los constreñimientos permite moldear una estructura y dinámica de
comportamiento táctico más o menos estable y/o flexible y más o menos regular y/o
variada. El análisis jerárquico de componentes principales y la superposición
dinámica o dynamic overlap <qd(t)> se postulan como parámetros de orden capaces
de caracterizar esta estructura dinámica.

106
 9. Referencias
Abrantes, C. I., Nunes, M. I., Maçãs, V. M., Leite, N. M. y Sampaio, J. E. (2012). Effects
of the number of players and game type constraints on heart rate, rating of perceived
exertion, and technical actions of small-sided soccer games. Journal of Strength and
Conditioning Research, 26(4), 976–981. doi: 10.1519/JSC.0b013e31822dd398

Aguiar, M., Gonçalves, B., Botelho, G., Lemmink, K. y Sampaio, J. (2015). Footballers’
movement behaviour during 2-, 3-, 4- and 5-a-side small-sided games. Journal of
Sports Sciences, 33(12), 1259–1266. doi:10.1080/02640414.2015.1022571

Almeida, C. H., Duarte, R., Volossovitch, A. y Ferreira, A. P. (2016). Scoring mode and
age-related effects on youth soccer teams’ defensive performance during small-sided
games. Journal of Sports Sciences, 34(14), 1355-1362.
doi:10.1080/02640414.2016.1150602

Araujo, D., Davids, K. y Hristovski, R. (2006). The ecological dynamics of decision

making in sport. Psychology of Sport and Exercise, 7(6), 653–676.

Araújo, D., Passos, P., Esteves, P., Duarte, R., Lopes, J., Hristovski, R. y Davids, K.
(2015). The micro-macro link is understanding sport tactical behaviours: Integrating
information and action at different levels of system analysis in sport. Movement &
Sport Sciences, 2015(89), 1–32. doi:10.1051/sm/2015028

Badii, R. y Politi, A. (1999). Complexity: Hierarchical structures and scaling in physics

(Vol. 6). Cambridge: Cambridge University Press.

Balescu, R. (1975). Equilibrium and nonequilibrium statistical mechanics. NASA STI/Recon

Technical Report A, 76, 32809.

Bourbousson, J., Deschamps, T. y Travassos, B. (2014). From players to teams: towards a

multi-level approach of game constraints in team sports. International Journal of
Sports Science & Coaching, 9(6), 1393–1406. doi:10.1260/1747-9541.9.6.1393

107
Bourbousson, J., Seve, C. y McGarry, T. (2010). Space–time coordination dynamics in
basketball: Part 2. The interaction between the two teams. Journal of Sports Sciences,
28(3), 349–358.

Bourbousson, J., Sève, C. y McGarry, T. (2010). Space–time coordination dynamics in

basketball: Part 1. Intra- and inter-couplings among player dyads. Journal of Sports
Sciences, 28(3), 339–347. doi:10.1080/02640410903503632

Bovier, A. (2006). Metastability: a potential theoretic approach. In Proceedings of the

International Congress of Mathematicians (pp. 499–518). Madrid.

Bruineberg, J. y Rietveld, E. (2014). Self-organization, free energy minimization, and

optimal grip on a field of affordances. Frontiers in Human Neuroscience, 8(August),
1–14. doi:10.3389/fnhum.2014.00599

Chemero, A. (2003). An Outline of a Theory of Affordances. Ecological Psychology, 15(2),

181–195. doi:10.1207/S15326969ECO1502_5

Davenport, E. C. y El-Sanhurry, N. A. (1991). Phi/Phimax: Review and Synthesis.

Educational and Psychological Measurement, 51(4), 821–828.
doi:10.1177/001316449105100403

Davids, K., Araujo, D. y Shuttleworth, R. (2005). Applications of Dynamical Systems

Theory to Football. En T. Reilly, J. Cabri y D. Araújo (Eds.), Science and Football V:
The Proceedings of the Fifth World Congress on Science and Football (pp. 537–550).
Routledge.

Davids, K., Button, C., Araújo, D., Renshaw, I. y Hristovski, R. (2006). Movement models
from sports provide representative task constraints for studying adaptive behaviour in
human movement studies. Adapt Behav, 14. doi:10.1177/105971230601400103

Domany, E., Hed, G., Palassini, M. y Young, A. P. (2001). State hierarchy induced by
correlated spin domains in short-range spin glasses. Phys. Rev. B, 64(22), 224406.
doi:10.1103/PhysRevB.64.224406

108
Duarte, R. (2012). Capturing Interpersonal coordination processes in association football:
from dyads to collectives. (Tesis doctoral, Universidade Técnica de Lisboa, Lisboa,
Portugal). Recuperado de: https://www.repository.utl.pt/handle/10400.5/4522

Duarte, R., Arajo, D., Correia, V. y Davids, K. (2012). Sports teams as superorganisms:
Implications of sociobiological models of behaviour for research and practice in team
sports performance analysis. Sports Medicine, 42(8), 633–642. doi:10.2165/11632450-
000000000-00000

Duarte, R., Araújo, D., Correia, V., Davids, K., Marques, P. y Richardson, M. J. (2013).
Competing together: Assessing the dynamics of team–team and player–team
synchrony in professional association football. Human Movement Science, 32(4), 555–
566. doi:10.1016/j.humov.2013.01.011

Duarte, R., Araújo, D., Davids, K., Travassos, B., Gazimba, V. y Sampaio, J. E. (2012).
Interpersonal coordination tendencies shape 1-vs-1 sub-phase performance outcomes
in youth soccer. Journal of Sports Sciences, 30(9), 871–877.
doi:10.1080/02640414.2012.675081

Duarte, R., Araújo, D., Folgado, H., Esteves, P., Marques, P. y Davids, K. (2012).
Capturing complex, non-linear team behaviours during competitive football
performance. Journal of Systems Science and Complexity, 26(1), 62–72.
doi:10.1007/s11424-013-2290-3

Duarte, R., Araújo, D., Freire, L., Folgado, H., Fernandes, O. y Davids, K. (2012). Intra-
and inter-group coordination patterns reveal collective behaviors of football players
near the scoring zone. Human Movement Science, 31(6), 1639–1651.
doi:10.1016/j.humov.2012.03.001

Duch, J., Waitzman, J. S. y Amaral, L. A. N. (2010). Quantifying the performance of

individual players in a team activity. PloS One, 5(6), e10937.

Eccles, D. (2010). The coordination of labour in sports teams. International Review of

Sport and Exercise Psychology, 3(2), 154–170. doi:10.1080/1750984X.2010.519400

109
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C. y Strahan, E. J. (1999). Evaluating the
use of exploratory factor analysis in psychological research. Psychological Methods,
4(3), 272–299. doi:10.1037/1082-989X.4.3.272

Fajen, B. R. y Warren, W. H. (2003). Behavioral dynamics of steering, obstacle avoidance,

and route selection. Journal of Experimental Psychology. Human Perception and
Performance, 29(2), 343–362. doi:10.1167/1.3.184

Folgado, H., Duarte, R., Fernandes, O. y Sampaio, J. (2014). Competing with lower level
opponents decreases intra-team movement synchronization and time-motion demands
during pre-season soccer matches. PloS One, 9(5), e97145.
doi:10.1371/journal.pone.0097145

Folgado, H., Duarte, R., Marques, P. y Sampaio, J. (2015). The effects of congested
fixtures period on tactical and physical performance in elite football. Journal of Sports
Sciences, 33(12), 1238–1247. doi:10.1080/02640414.2015.1022576

Frencken, W., Lemmink, K., Delleman, N. y Visscher, C. (2011). Oscillations of centroid

position and surface area of soccer teams in small-sided games. European Journal of
Sport Science, 11(4), 215–223. doi:10.1080/17461391.2010.499967

Fulgosi, A. (1988). Faktorska analiza. Zagreb: Školska knjiga. Retrieved from

https://books.google.mk/books?id=cmpfmgEACAAJ

Gabin, B., Camerino, O., Anguera, M. T. y Castañer, M. (2012). Lince: Multiplatform sport
analysis software. Procedia -Social and Behavioral Sciences, 46, 4692–4694.
doi:10.1016/j.sbspro.2012.06.320

Gama, J., Passos, P., Davids, K., Relvas, H., Ribeiro, J., Vaz, V. y Dias, G. (2014).
Network analysis and intra-team activity in attacking phases of professional football.
International Journal of Performance Analysis in Sport, 14(3), 692–708.

Gibson, J. J. (1979). The ecological approach to visual perception. Boston, MA, US:
Houghton, Mifflin and Company.

110
Gonçalves, B., Coutinho, D., Santos, S., Lago-Penas, C., Jiménez, S. y Sampaio, J. (2017).
Exploring Team Passing Networks and Player Movement Dynamics in Youth
Association Football. PloS One, 12(1), e0171156.

Gonçalves, B., Marcelino, R., Torres-Ronda, L., Torrents, C. y Sampaio, J. (2016). Effects
of emphasising opposition and cooperation on collective movement behaviour during
football small-sided games. Journal of Sports Science, 34(14), 1346–1354.
doi:10.1080/02640414.2016.1143111

Gonçalves, B. V., Figueira, B. E., Maçãs, V. y Sampaio, J. (2014). Effect of player position
on movement behaviour, physical and physiological performances during an 11-a-side
football game. Journal of Sports Sciences, 32(2), 191–199.
doi:10.1080/02640414.2013.816761

Gréhaigne, J. F., Bouthier, D. y David, B. (1997). Dynamic-system analysis of opponent

relationships in collective actions in soccer. Journal of Sports Sciences, 15(2), 137–
149. doi:10.1080/026404197367416

Haken, H. (1983). Synergetics, an Introduction: Nonequilibrium. Phase Transitions and

Self-Organization in Physics, Chemistry, and Biology.

Harbourne, R. T. y Stergiou, N. (2009). Movement Variability and the Use of Nonlinear

Tools: Principles to Guide Physical Therapist Practice. Physical Therapy, 89(3), 267–
282. doi:10.2522/ptj.20080130.

Hill-Haas, S. V., Dawson, B., Impellizzeri, F. M. y Coutts, A. J. (2011). Physiology of

small-sided games training in football: A systematic review. Sports Medicine.
doi:10.2165/11539740-000000000-00000

Hristovski, R., Davids, K., Araújo, D. y Passos, P. (2011). Constraints-induced emergence

of functional novelty in complex neurobiological systems: a basis for creativity in
sport. Nonlinear Dynamics, Psychology, and Life Sciences, 15(2), 175–206.

Hristovski, R., Davids, K., Araújo, D., Passos, P., Torrents, C., Aceski, A. y Tufekcievski,
A. (2013). Creativity in sport and dance: Ecological dynamics on a hierarchically soft-

111
 assembled perception-action landscape. En K. Davids, R. Hristovski, D. Araújo, N.
Balagué, C. Button y P. Passos (Eds.), Complex systems in sport (pp. 261–274).
London: Routledge.

Hristovski, R., Davids, K., Passos, P. y Araújo, D. (2012). Sport Performance as a Domain
of Creative Problem Solving for Self-Organizing Performer-Environment Systems.
The Open Sports Sciences Journal, 5(1), 26–35. doi:10.2174/1875399X01205010026

Joliffe, I. T. (2002). Principal component analysis (2nd editio). New York: Springer.

Kelly, D. M. y Drust, B. (2009). The effect of pitch dimensions on heart rate responses and
technical demands of small-sided soccer games in elite players. Journal of Science and
Medicine in Sport, 12(4), 475–479. doi:10.1016/j.jsams.2008.01.010

Kelso, J. A. S. (1984). Phase transitions and critical behaviour in human bimanual

coordination. American Journal of Physiology: Regulatory, Integrative and
Comparative Physiology, 246, R1000-1004.

Kelso, J. A. S. (1995). Dynamic patterns: the self-organization of brain and behavior.

Complex Adaptive Systems series. MIT Press, Bradford Book, Cambridge, MA.

Liu, H., Gómez, M.-A., Gonçalves, B. y Sampaio, J. (2016). Technical performance and
match-to-match variation in elite football teams. Journal of Sports Sciences, 34(6),
509–18. doi:10.1080/02640414.2015.1117121

McGarry, T. (2006). Identifying patterns in squash contests using dynamical analysis and
human perception. International Journal of Performance Analysis in Sport, 6(2), 134–
147.

McGarry, T. (2009). Applied and theoretical perspectives of performance analysis in sport:

Scientific issues and challenges. International Journal of Performance Analysis in
Sport, 9(1), 128–140.

McGarry, T., Anderson, D. I., Wallace, S. A., Hughes, M. D. y Franks, I. M. (2002). Sport
competition as a dynamical self-organizing system. Journal of Sports Sciences,

112
 20(10), 771–781. doi:10.1080/026404102320675620

Memmert, D., Lemmink, K. A. P. M. y Sampaio, J. (2017). Current Approaches to Tactical

Performance Analyses in Soccer Using Position Data. Sports Medicine, 47(1), 1–10.
doi:10.1007/s40279-016-0562-5

Moura, F. A., van Emmerik, R. E. A., Santana, J. E., Martins, L. E. B., Barros, R. M. L. de
y Cunha, S. A. (2016). Coordination analysis of players’ distribution in football using
cross-correlation and vector coding techniques. Journal of Sports Sciences, 34(24),
2224-2232. doi:10.1080/02640414.2016.1173222

Newell, K. M., McDonald, P. V y Kugler, P. N. (1991). The perceptual-motor workspace

and the acquisition of skill. In J. Requin & G. E. Stelmach (Eds.), Tutorials in motor
neuroscience (pp. 95–108). Springer.

Palut, Y. y Zanone, P.-G. (2005). A dynamical analysis of tennis: Concepts and data.
Journal of Sports Sciences, 23(10), 1021–1032. doi:10.1080/02640410400021682

Parisi, G. (1983). Order Parameter for Spin-Glasses. Phys. Rev. Lett., 50(24), 1946–1948.
doi:10.1103/PhysRevLett.50.1946

Passos, P., Araújo, D., Davids, K., Gouveia, L., Milho, J. y Serpa, S. (2008). Information-
governing dynamics of attacker-defender interactions in youth Rugby Union. Journal
of Sports Science, 26(13), 1421-1429. doi:10.1080/02640410802208986

Rampinini, E., Impellizzeri, F. M., Castagna, C., Coutts, A. J. y Wisløff, U. (2009).

Technical performance during soccer matches of the Italian Serie A league: Effect of
fatigue and competitive level. Journal of Science and Medicine in Sport, 12(1), 227–
233. doi:10.1016/j.jsams.2007.10.002

Ric, A., Hristovski, R. y Torrents, C. (2015). Can joker players favor the exploratoy
behavior in football small-sided games? Research in Physical Education , Sport and
Health, 4(2), 35–39.

Riley, M. A., Richardson, M. J., Shockley, K. y Ramenzoni, V. C. (2011). Interpersonal

113
 synergies. Frontiers in Psychology, 2, 1–7. doi:10.3389/fpsyg.2011.00038

Rosso, B. D. (2014). Creativity and Constraints: Exploring the Role of Constraints in the
Creative Processes of Research and Development Teams. Organization Studies, 35(4),
551–585. doi:10.1177/0170840613517600

Sampaio, J. y Maçãs, V. (2012). Measuring tactical behaviour in football. International

Journal of Sports Medicine, 33(5), 395–401.

Sibani, P. y Dall, J. (2003). Log-Poisson statistics and full aging in glassy systems. EPL
(Europhysics Letters), 64(1), 8. doi:10.1209/epl/ i2003-00109-0

Silva, P. (2014). Shaping Tactical Behaviours in Football : An Ecological Dynamics

Approach. (Tesis doctoral, Universidade do Porto, Oporto, Portugal). Recuperado de:
http://www.fade.up.pt/cifi2d/files/tese-pedro-silva.pdf

Silva, P., Chung, D., Carvalho, T., Cardoso, T., Davids, K., Araújo, D. y Garganta, J.
(2016). Practice effects on intra-team synergies in football teams. Human Movement
Science, 46, 39–51. doi:10.1016/j.humov.2015.11.017

Silva, P., Duarte, R., Sampaio, J., Aguiar, P., Davids, K., Araújo, D. y Garganta, J. (2014).
Field dimension and skill level constrain team tactical behaviours in small-sided and
conditioned games in football. Journal of Sports Sciences, 32(20), 1888-1896.
doi:10.1080/02640414.2014.961950

Silva, P., Esteves, P., Correia, V., Davids, K., Araújo, D. y Garganta, J. (2015). Effects of
manipulations of player numbers vs. field dimensions on inter-individual coordination
during small-sided games in youth football. International Journal of Performance
Analysis in Sport, 15(2), 641-659.

Silva, P., Garganta, J., Araújo, D., Davids, K. y Aguiar, P. (2013). Shared knowledge or
shared affordances? insights from an ecological dynamics approach to team
coordination in sports. Sports Medicine, 43(9), 765–772. doi:10.1007/s40279-013-
0070-9

114
Silva, P., Travassos, B., Vilar, L., Aguiar, P., Davids, K., Araújo, D. y Garganta, J. (2014).
Numerical relations and skill level constrain co-adaptive behaviors of agents in sports
teams. PloS One, 9(9), e107112. doi:10.1371/journal.pone.0107112

Silva, P., Vilar, L., Davids, K., Araújo, D. y Garganta, J. (2016). Sports teams as complex
adaptive systems: manipulating player numbers shapes behaviours during football
small-sided games. SpringerPlus, 5(1), 191. doi:10.1186/s40064-016-1813-5

Smith, D., Ric, A., Goncalves, B., Torrents, C., Sampaio, J. y Hristovski, R. (2017).
Exploring coordination in soccer players’ movement using reginme switching state
space models. En Vth World Congress on Science & Football, Rennes (France), 31st
May – 2nd June, 2017.

Stergiou, N., Buzzi, U., Kurz, M. y Heidel, J. (2004). Nonlinear Tools in Human
Movement. In N. Stergiou (Ed.), Innovative analyses of human movement (pp. 63–90).
Champaign, IL, Human Kinetics.

Torrents, C., Hristovski, R., Coterón, J. y Ric, A. (2016). Interpersonal coordination in

contact imporvisation dance. En P. Passos, K. Davids y J. Y. Chow (Eds.),
Interpersonal Coordination and Performance in Social Systems (pp. 94–108). London:
Routledge.

Torrents, C., Ric, A. y Hristovski, R. (2015). Creativity and emergence of specific dance
movements using instructional constraints. Psychology of Aesthetics, Creativity, and
the Arts, 9(1), 65–74. doi:10.1037/a0038706

Travassos, B., Araújo, D., Duarte, R. y McGarry, T. (2012). Spatiotemporal coordination

behaviors in futsal (indoor football) are guided by informational game constraints.
Hum Mov Sci, 31. doi:10.1016/j.humov.2011.10.004

Travassos, B., Araújo, D., Vilar, L. y McGarry, T. (2011). Interpersonal coordination and
ball dynamics in futsal (indoor football). Human Movement Science, 30, 1245–1259.
doi:10.1016/j.humov.2011.04.003

Travassos, B., Gonçalves, B., Marcelino, R., Monteiro, R. y Sampaio, J. (2014). How

115
 perceiving additional targets modifies teams’ tactical behavior during football small-
sided games. Human Movement Science, 38, 241–50.
doi:10.1016/j.humov.2014.10.005

Travassos, B., Vilar, L., Araújo, D. y McGarry, T. (2014). Tactical performance changes
with equal vs unequal numbers of players in small-sided football games. International
Journal of Performance Analysis in Sport, 14(2), 594–605.

Turvey, M. T. (1992). Affordances and Prospective Control: An Outline of the Ontology.

Ecological Psychology, 4(3), 173–187. doi:10.1207/s15326969eco0403_3

Vázquez, P. (2017). Unified framework for the study of sport-related behavior. Universidad
de Barcelona, Barcelona.

Vilar, L., Araújo, D., Davids, K. y Bar-Yam, Y. (2013). Science of winning soccer:
Emergent pattern-forming dynamics in association football. Journal of Systems
Science and Complexity, 26(1), 73–84. doi:10.1007/s11424-013-2286-z

Vilar, L., Araújo, D., Davids, K., Correia, V. y Esteves, P. T. (2013). Spatial-temporal
constraints on decision-making during shooting performance in the team sport of
futsal. Journal of Sports Sciences, 31(8), 840–846.
doi:10.1080/02640414.2012.753155

Vilar, L., Araújo, D., Davids, K., Travassos, B., Duarte, R. y Parreira, J. (2014).
Interpersonal coordination tendencies supporting the creation/prevention of goal
scoring opportunities in futsal. European Journal of Sport Science, 14(1), 28–35.
doi:10.1080/17461391.2012.725103

Vilar, L., Duarte, R., Silva, P., Chow, J. Y. y Davids, K. (2014). The influence of pitch
dimensions on performance during small-sided and conditioned soccer games. Journal
of Sports Sciences, 32(19), 1751–1759. doi:10.1080/02640414.2014.918640

Warren, W. H. (1990). The Perception-Action Coupling. In H. Bloch & B. I. Bertenthal

(Eds.), Sensory-Motor Organizations and Development in Infancy and Early
Childhood: Proceedings of the NATO Advanced Research Workshop on Sensory-

116
 Motor Organizations and Development in Infancy and Early Childhood Chateu de
Rosey, France (pp. 23–37). Dordrecht: Springer Netherlands. doi:10.1007/978-94-
009-2071-2_2

Westerhuis, J. A., Kourti, T. y MacGregor, J. F. (1998). Analysis of multiblock and

hierarchical PCA and PLS models. Journal of Chemometrics, 12(5), 301–321.

Withagen, R., Poel, H. J. d., Araújo, D. y Pepping, G.-. J. (2012). Affordances can invite
behaviour: reconsidering the relationship between affordances and agency. New Ideas
Psychol, 30(2), 250–258. doi:10.1016/j.newideapsych.2011.12.003

117