Professional Documents
Culture Documents
INTRODUCTION
01. INTRODUCTION
You are watching a football match, possession is being circulated
around the middle in unspectacular fashion. Then all of a sudden,
your central midfielder (known for his vision) gets a moment of
time and hits a great pass. You know it's a great pass, it's
something different, yeah it's not normal, it's..... unique
Further Reading
Statsbomb Resource Centre
Sign-up to their public data user agreement, get access to the data &
get more info on the dataset
Accounts to Follow
StatsBombR Package
Ted Knutson Mike Goodman
The R package that will help you get your hands on that data super fast Founder / CEO Managing Editor
and super easy @mixedknuts @TheM_L_G
The NNS will calculate the dissimilarity of pass n with other • pass.length: distance from origin to destination
passes as the Euclidean distance within the multi-dimensional • pass.angle: the angle of the pass measured in radians
space of the data • pass.height.id: 3 grade variable for pass height (1) ground
pass (2) mid-height (3) high-ball
Instead of creating clusters, the ‘uniqueness’ metric will be the
• x: the vertical location of the pass origin
sum of the distance of a certain number of passes. To make
this analysis scalable for more data, 1% of the total passes is • y: the horizontal location of the pass origin
used as the comparison scope (569 passes in this case) • x.destination: the vertical location of the pass destination
• y.destination: the horizontal location of the pass
The higher the distance the more ‘unique’ the pass is deemed
destination
UNIQUENESS PASSING MODEL 05. DOES IT MATCH?
neighbour search is doing it’s job and finding the 50 most similar Pass Origin - Circle denoting the origin of the pass
passes. Does it match? Yes, it’s doing a very good job at matching Pass Path - Line denoting the path of the pass
similar passes, so let’s delve a little deeper into the results Similar Passes - Lower Opacity for matched passes
UNIQUENESS PASSING MODEL 06. HISTOGRAM OF UNIQUENESS
Quantiles
0 % 70.4
25 % 100.2
Frequency
50 % 119.4
75 % 168.8
100 % 608.9
y 40
curved dip. Uniqueness Left Side Right Side uniqueness aside
x location of the reaching high from a high
200 pass origin does 20
values in at the values being seen
not seem to be wings in forward passes
0 25 50 75 100 125
overly descriptive 0
200 400 600
x Uniqueness
Backwards
Uniqueness
400
than a mild y 40 than a longer the pass
Uniqueness
400
curved dip and a Uniqueness the higher the
distinct uplift in 20
reaching high uniqueness rating
200 very advanced values in at the 200
receives. 0
wings
200 400 600
0 25 100 125 0 30 60 90
50 75
xx Uniqueness Pass Length (m)
UNIQUENESS PASSING MODEL 08. LOOKING FOR PATTERNS
2
2
1. An increase at the edges of the pitch 1. An increase at the edges of the pitch
2. An increase in the defensive box 2. An increase in the offensive box
3. An increase in wide areas around the box 3. An interest deep lying play-maker spot?
UNIQUENESS PASSING MODEL 09. WATCH THE VIDEO!
The shortlists largely pass the eye-test. Toni Kroos tops the pile, high Yasir Al Shahrani Saudi Arabia Al-Hilal Riyadh 7.6
possession teams like Germany and Spain show up and and Red Bull Marcelo Brazil Real Madrid 6.6
Leipzig features twice. Midfielders
good knowledge of. Yet when Mesut Özil Germany Arsenal FC 8.9
reviewing their footage I can Tarek Hamed Egypt Zamalek SC 8.6
understand why they are scoring João Mário Portugal Inter Milan 8.2
highly. For example ——> Forwards
13. BEST XI
06. BEST XI
NK
3
P
RA
TO
OP
OM
-T
FR
XI
K
IC
T
ES
-P
-B
XI
S
T
SE
Another way of eye-balling the
ES
S
-B
PA
ISCO ISCO
S
NG
SE
findings to is form a Best XI from DZYUBA WERNER
KI
S
11 11
PA
AC
T
NG
AT
9 9
KI
UE
AC
IQ
T
UN
AT
position or in the top 3 per
UE
ÖZIL ÖZIL
IQ
UN
position (to account for
10 10
BORGES MODRIĆ
8 8
S. RAMOS S. RAMOS
Although the Top 4 4
completion analysis.
completion probability
within the mid 200s to
a low of 0.47
expected completion 15
20 - 30 Unique Attacking Passes
0
-40% 0% -40% -80%
xCompletion.Rate
UNIQUENESS PASSING MODEL 16. xCOMPLETION RATES - PER POSITION
Henrik Dalsgaard Denmark -12.5 % Dele Alli England -16.6 % Antoine Griezmann France -22.3 %
UNIQUENESS PASSING MODEL 17. PLOTTING THE RESULTS - DEFENDERS