Professional Documents
Culture Documents
Poker As A Domain of Expertise: Correspondence: Jussi - Palomaki@helsinki - Fi
Poker As A Domain of Expertise: Correspondence: Jussi - Palomaki@helsinki - Fi
Abstract
Poker is a game of skill and chance involving economic decision-making under uncertainty. It is also a
complex but well-defined real-world environment with a clear rule-structure. As such, poker has strong
potential as a model system for studying high-stakes, high-risk expert performance. Poker has been
increasingly used as a tool to study decision-making and learning, as well as emotion self-regulation. In
this review, we discuss how these studies have begun to inform us about the interaction between
emotions and technical skill, and how expertise develops and depends on these two factors. Expertise in
poker critically requires both mastery of the technical aspects of the game, and proficiency in emotion
regulation; poker thus offers a good environment for studying these skills in controlled experimental
settings of high external validity. We conclude by suggesting ideas for future research on expertise, with
new insights provided by poker.
Keywords
Economic decisions, probabilistic decision-making, risk, expertise, poker
https://www.journalofexpertise.org
Journal of Expertise / June 2020 / vol. 3, no. 2
Palomäki et al. (2020) Poker as Domain of Expertise
creates a tension: Restricted tasks abstract away under risk and uncertainty. Poker is a well-
much of real-world domain complexity, structured game played in a social setting with
ambiguity, and the “world knowledge” that many different game variants involving
experts1 bring to bear on the task. This allows randomness and probabilistic economic
one to arbitrate more definitively among decision-making. These aspects of poker make
competing mechanistic hypotheses, but also it attractive for various scientific disciplines
raises the question of whether those putative interested in economic or rational decision-
mechanisms are a factor in more realistic making at the individual and social levels. Poker
settings. also comes with a very large online community
Laboratory tasks are meant to be analogues of players generating big datasets and powerful
of real-world environments, but whether the data-gathering opportunities (e.g., Eil & Lien,
tasks actually have relevance outside the 2014; Siler, 2010) similar to many electronic
laboratory has to be taken on faith (that is, sports (Esports) games (e.g., Thompson,
researchers’ intuition for how similar their McColeman, Stepanova, & Blair, 2017). In
simplified, abstract decision-making task really general, games that have gone online provide
is to a real-world task setting). Traditional enormous research opportunities—poker in
decision-making tasks are thus designed for particular, given its long history of defined
laboratory convenience—often presented in text analytic structure, game theoretical analysis, as
or numerical form using novice subjects and/or well as large player base.
with domain-general problems. This makes We will argue that poker also offers a novel
them particularly limited in terms of shedding look into expertise, since the concept of poker
light on skilled decision-making processes in skill is more complex than the much-studied
rich and more natural contexts. technical skill in other well-studied game
The study of games has been a valuable domains such as chess. This is due to the
route for cognitive scientists and can offer some element of chance in the game: Skilled poker
middle ground between experimental control players need to have emotional tolerance of
and ecological validity. Most everyday natural outcome variability—that is, to be successful
decisions—such as choosing ingredients for they need to able to control and reflect on their
cooking a meal or deciding on what to wear to a negative emotions when even right choices can
party—cannot be given comprehensive lead to catastrophically bad outcomes merely
mathematical definitions, nor are there often due to chance. Compared with chess, poker is
clear normative criteria on the goodness of a also typically played with a larger group of
decision. However, many games are everyday people, making emotion regulation particularly
tasks with definable rules that can be compactly important.
represented. Also, gameplay offers means to So far, this element of poker has not been
design recurring situations that can be used to thoroughly studied, despite the potentially
present decision-making tasks that have both significant benefits for decision- and cognitive
experimental control and high ecological sciences. Therefore, poker has strong, but as yet
validity (such as choosing the next move in untapped, potential for research on social- and
chess). Game decisions can, moreover, often be cognitive psychology, decision-making, and
varied in terms of task difficulty and complexity expert performance.
to suit particular participants or experimental Overall, while poker has received a lot of
questions. Finally, mathematical analysis of attention outside academia2, up until recently
games has in many cases provided normative much of the research on poker has been
standards whereby decision quality is evaluated. clinically motivated; for example, evaluating
In this review we show how these desirable how pathological gambling behavior manifests
characteristics apply to the game of poker, in poker, or theoretically focused on using poker
which can serve as a valuable model system for as a testbed for artificial intelligence (see Brown
studying expert economic decision-making & Sandholm, 2019; Moreau, Chabrol, &
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
Chauchard, 2016; and Rubin & Watson, 2011). technical (mathematical, statistical and game
We argue that games of economic decisions theoretical) and emotion regulation (“mental
such as poker can and should be used more in game”) components, and how various social
basic behavioral research on decision-making elements of the game can bias players’ decision-
and expertise under risk and uncertainty. The making. (2) How do poker skills develop into
element of randomness may make mastering expertise, and how does poker allow study of
poker different compared to mastering many expertise? We link the components and
deterministic games—however, as articulated development of poker skill to previous work on
by Siler (2011), it is precisely this stochastic expertise, deliberate practice, and skilled
nature of poker that makes it a much more intuition and show that poker offers a novel way
realistic task environment reflecting the vagaries to look at expertise and expert performance due
and uncertainties of many real-life phenomena to its emotion regulatory skill aspects. (3) How
such as financial decisions. can future studies on expertise and decision-
In our review we first provide a section making make use of poker? We conclude our
describing the technical aspects of poker and review by detailing how future studies can draw
explain the basic structures of the game and how insights from poker to examine skilled decision-
the element of chance influences skill making under emotional and social constraints.
development. Then we address the following Table 1 illustrates the features of poker
review research questions (RQs): (1) What are reviewed in this paper and their relevance to
the components of poker skill? We describe how research on expertise and decision-making.
the concept of poker skill comprises both
Table 1: Features of poker and their relevance for decision-making and expertise researchers
Poker Feature Research Relevance; Poker allows study of the following:
Incomplete information Microcosm of naturalistic financial decision-making
Interplay of skill and chance Skill perception and biases in decision-making: in the short
term, bad players may win (inflated skill perception), and good
players may lose (obfuscation of true skill)
Male-dominated social environment Masculinity and gender stereotypes in a competitive setting,
gender biased decision-making
Technical and emotional aspects of skill Interplay between emotion regulation ability and decision-
making accuracy
Multiple sources of both public and private “Game theory optimization” strategies, how skilled players
information avoid exploitation
Skilled intuition Ecologically valid skilled intuition in a “medium validity” (as
opposed to “high validity”; e.g., chess) environment
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
information on one’s strategy, further pressuring similar results). Skill has a demonstrably
players to control their behavior. Poker is also a significant role also in real-world poker success.
dynamic environment, as the “game state” Professional players are consistently more
changes even when the agent does nothing. The successful than amateurs at the World Series of
social pressure and the monetary stakes Poker (Croson et al., 2008; Levitt & Miles, 2014).
involved create additional cognitive and One way to illustrate the role of chance in
emotional load—for professional players the poker is through simulations of outcome
rewards can reach millions of US dollars. variability. Players’ levels of skill are reflected in
In the technical examples that follow, we their win rate, which is the average amount of
will focus on the most popular variant of poker profit over some number of played hands (usually
called No Limit Texas Hold’Em (NLHE). In 100; van Loon et al., 2015). The standard
NLHE, each player is first dealt two cards, and deviation of a player’s win rate (a measure of
the goal is to form the best five-card outcome variability) can be 20 times higher than
combination from one’s own two cards (not the win rate itself (Billingham et al., 2013). To
seen by the other players) plus cards dealt on the illustrate, we will compare two equally skilled
table (shared with all other players). There are hypothetical players playing 200,000 hands each.
up to four rounds of betting, during which the By assuming both players have somewhat low
number of publicly shared cards increases, win rates (on the statistical edge of making long-
starting with no shared cards and ending with, at term profit), we might observe the situation
most, five. Between each round the players can presented in Figure 1: One player could be
make investment decisions on whether to keep winning substantially (> 15 000 €), and the other
playing, how much to invest in the pot, or give clearly losing (-5000 €). Outcome variability is
up (i.e., fold)4. The pot will go to the winner (or thus a highly significant factor, masking a player’s
split between winners in case of ties), who is the “true” skill as defined by the expected long-run
player with the best card combination, or the winnings (dashed line in Figure 1). This means
only one not to have folded. that while poker differs from games of pure
chance (such as roulette) or games of skill and
Skill and Chance chance where long-term profit is unattainable
Generally, poker is viewed as a game of both skill (e.g., blackjack; Bjerg, 2010), outcome variability
and chance, but the extent to which one or the still makes it challenging to empirically estimate
other dominates is debated (Croson, Fishman, & the actual skill level of any individual player from
Pope, 2008; Dedonno & Detterman, 2008; Fiedler naturalistic play data5.
& Rock, 2009; Levitt & Miles, 2014; Meyer, von However, player skill can also be estimated
Meduna, & Brosowski, 2013). Anecdotal experimentally, by using representative decision-
evidence supports the view of poker as a game making tasks, with known normative solutions:
where one’s skills can constantly be improved more (technically) skilled players should
(Brunson, 2005; Sklansky & Malmuth, 1999; consistently reach that solution more quickly
Tendler, 2011). The consensus view in academic and/or reliably. In two laboratory studies (Linnet
discussion is that although chance plays a role in et al., 2010; 2012), those who had played poker at
short-term results, with enough skill poker can be least once a week for at least a year were better at
played profitably in the long run. Empirical estimating betting outcomes than less experienced
support for this view comes from an analysis of ones. Two online studies with simplified poker
456 million online poker hands (van Loon, van tasks showed that the amount of poker experience
den Assem, & van Dolder, 2015). Van Loon et al. was strongly and positively associated with
(2015) created a simulation based on these data, making mathematically appropriate poker
comparing the best players with the worst ones, decisions (Laakasuo, Palomäki, & Salmela, 2015;
and found that skill starts to dominate chance Palomäki, Laakasuo, & Salmela, 2013a). Thus,
when performance is assessed over about 1,500 or components of poker skill can be isolated and
more hands of play (see Fiedler & Rock, 2009, for studied both “in the wild” and in the laboratory.
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
Figure 1. A simulation of 900 NLHE “poker players” with equal win rates. Win rates are based
on “big blinds”’; that is, the minimum bet size allowed by the rules. These win rates are
calculated based on 3 big blinds—in this case, euros—per 100 hands played, with a standard
deviation (SD) of 80 for the win rates (typical win rate SDs in NLHE are 70-90; Billingham et
al., 2013). Note that the “players” are simulated processes based on two parameters (win rate and
SDwin rate) and thus independent of one another. The figure depicts only the highest and lowest
earning players (top and bottom curve, respectively), and the expected value of earnings for all
900 players (dashed line). (Translated into English from Laakasuo, Palomäki, & Lappi, 2015).
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
relative to the opponents’ assumed hand and novice poker players in a simulated poker
strengths—as further cards are dealt (see the task during which participants had to “think out
Appendix for detailed examples). Calculations loud.” Proficient players outperformed both
of hand strength and hand potential require intermediate and novice players (in terms of
knowledge of poker betting odds in a given profit), and self-reported the highest amount of
situation, mathematical aptitude, and working thought processing and attention to relevant
memory capacity (e.g., DeDonno, 2016; Meinz technical aspects of the task—such as betting
et al., 2012). patterns, estimated opponent ability, future
Betting strategy, bluffing, and unpredictability opponent actions and “tells,” and hand selection
refer to knowledge of when and how much to bet and strength. Practicing these skill elements
or raise (or fold) in a sufficiently unpredictable leads to better performance: DeDonno and
manner to maximize one’s profit and protect Detterman (2008) conducted a laboratory
oneself from exploitation. Betting strategy refers experiment where naïve poker players
for instance to the decision to bluff with a fixed systematically practiced technical poker
frequency or not, as a player might decide a concepts which lead to improved success in the
priori to bluff a given number of times in a game. The players were given information and
game. These skill elements require players to feedback about (1) when and why to pay
apply (either explicitly or implicitly) the attention to the other players’ decisions; (2) the
concepts of game theory, such as Nash concept of playing fewer hands, and how to play
equilibrium6, in their own decision-making. them; and (3) hand ranking strategy with quality
Opponent modeling refers to estimating the values for the initial hand. These correspond to
full range of an opponent’s possible hands. opponent modeling, betting strategy, and
Specifically, opponent modeling relates to how evaluating hand strength and hand potential,
various behavioral and social opponent respectively.
characteristics, such as betting patterns, physical Poker is a knowledge-rich domain, with
“tells”, or gender, influence the way (or what complex demands on both technical and
range of possible hands) one’s opponents are strategic skills. These demands present
predicted to play—and, consequently, how they information processing challenges requiring the
should be played against to maximize profit. player to go beyond the information embodied
This generally requires interpreting concealed in the cards and explicit in the rules. We have
social signals, reading covert facial expressions, provided a detailed poker task analysis in the
and detecting deception in general. Appendix, which illustrates the complexities
Bankroll management is the knowledge of involved in poker decision-making.
how much money is needed for playing, in Poker is also well-suited to facilitate study
relation to the stakes played, to avoid going of players’ information processing because the
broke. That is, how much capital is needed to relationship between different forms of
withstand outcome variance and avoid “going information is relatively clear and understood.
broke due to merely bad luck.” Good bankroll The above aspects of technical poker skill
management skills are typically associated with embody different challenges of information
a good understanding of the concepts of manipulation (Salen & Zimmerman, 2004, p.
statistical variance and risk of ruin (Browne, 148), in the sense of the information embodied
1989; Palomäki et al., 2013a). by the cards as defined by the rules of poker.
The depth of the technical aspects of poker This information can take multiple forms,
is evidenced by clear differences in technical including: (1) Information known to all players;
skill between proficient and novice poker i.e., the five “community cards” shown on the
players. For example, in a laboratory experiment table; (2) Information known to only one player;
St. Germain and Tenenbaum (2011) compared i.e. the two “hole” cards of each player; (3)
the performance of proficient players, with Information known to the game only; e.g., the
significant tournament success, to intermediate unused cards in the deck; and (4) Randomly
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
generated information; i.e., the shuffling of the technical skill elements, the concept of poker
deck. Part of the technical skill in poker is to skill encompasses an emotion regulatory aspect.
know which form of information the game Emotion regulation skills refer to the ability to
embodies at any given time—and to keep track withstand the arduous, yet inevitable, losing
of and predict how the game’s information streaks without having them affect the quality of
moves between these forms. one’s decisions (Boujou et al., 2013; Palomäki
For example, a certain amount of the et al., 2013a). These skills may be conscious
information known only to one player (the hole processes explicitly controlling one’s impulses
cards), can be leaked to other players due to that by willpower or positive self-talk, or they could
player’s response to the turn, flop, or river (see be more unconscious processes, which might be
the Appendix for explanation of these terms). In termed trait emotional stability or “character”
fact, in a recent study, Frey, Albino, and developed by surviving previous encounters.
Williams (2018) analyzed 1.75 million poker McCormack and Griffiths (2012) interviewed
hands and found that winning (skilled) poker professional and recreational poker players and
players were better than losing (unskilled) found that professional players were not only
players at integrative information processing— more likely to be logical and controlled in their
creating new information based on the behavior, but also took fewer risks and were less
interaction between their own hole cards and likely to chase after losses (i.e., keep playing in
their opponents’ betting patterns. This made the an attempt to win back their losses). Conversely,
winning players’ decision-making less exploitable recreational players showed more signs of
and harder for others to reverse engineer (Frey et losing control, taking unnecessary risks and
al., 2018). playing under the influence of intoxicants. In
To recap, technical poker skills consist of correlational online studies, poker experience
knowledge of hand strength and hand potential, has been found to be negatively associated with
betting and bluffing strategy, ability to avoid the psychological traits of emotionality
exploitation (playing unpredictably) and to (Laakasuo, Palomäki, & Salmela, 2014), self-
exploit others, and bankroll management; all of rumination (Laakasuo, Palomäki, & Salmela,
which can be viewed in terms of challenges for 2016; Palomäki et al., 2013a) and emotional
information manipulation. However, we note sensitivity to losses (Laakasuo et al., 2016;
that empirical research on technical poker skill Palomäki et al., 2014). Experienced players are
development in terms of information thus less emotional, dwell less on negative
manipulation strategies is still relatively scarce. thoughts, and report higher emotional tolerance
of poker losses than do inexperienced players.
Emotion Regulation Skills Moreover, Palomäki and colleagues (2013a)
Due to statistical variance in the game, even report that in an online setting with simplified
technically skilled poker players regularly poker tasks, experienced players—but not
encounter losing streaks and “bad beats” — inexperienced—made mathematically better
losing money in situations where losing is poker decisions when they had a strong
objectively unlikely, and not the result of tendency for self-reflection. Self-reflection is a
normatively poor decision-making. Losing large trait related to analyzing one’s past mistakes in a
sums of money often elicits negative emotions, cool and detached manner. Consistent with
which, in turn, can have detrimental effects on these results, Leonard and Williams (2015)
upcoming decisions. For example, in a bout of employed a measure of technical poker skills
anger an experienced and otherwise technically and betting strategy and found that proficient
skilled player might forgo sound betting or players were less susceptible to gambling
bankroll-management strategies, ending up fallacies and had higher emotional tolerance for
playing with too high stakes and betting financial risk and better social information
erratically despite factually knowing it is processing skills.
mathematically inadvisable. Thus, in addition to
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
Tilt: Intense Moral Anger Revealed in Poker the negative emotional states leading to tilting
The importance of emotion regulation skills and (Browne, 1989). The sense of injustice is
aversion to loss (via pursuing risk) in poker is particularly interesting, as it makes tilting a
underscored by the phenomenon known as form of moral emotion: Individuals (sampled in
tilting, which refers to losing control due to Palomäki et al., 2013b) who tilt reported feeling
negative emotions—typically related to bad personally insulted, and that they “unfairly” lost
beats or prolonged losing streaks—and making money for which that had worked diligently.
strategically weak decisions and losing They viewed variance as “bad luck,” took it
significantly more money than otherwise personally, and started pouring their earnings
(Palomäki et al., 2014; Moreau, Delieuvin, into the game and chasing their “fair chance.”
Chabrol, & Chauchard, 2017). Extreme cases of The authors postulated (Palomäki et al., 2013b)
tilting have even led to losing entire life savings that the psychology of tilting could be viewed as
within minutes, and to self-reported memory moral anger: Losing due to bad luck is
losses of the preceding events (Palomäki et al., perceived as “cosmically” unjust, which
2013b; Tendler, 2011). motivates an overly aggressive yet ineffective
Poker communities seem to agree that tilting retaliation strategy of excessive betting. In the
is a significant phenomenon: in an online study, aftermath of tilting, the players reported being
88% of poker players reported having tilted disappointed in themselves and that they were
severely at least once within their last 6 months, ruminating over lost resources.
43% more than five times, and 24% more than Experienced players, however, differ from
10 times (Palomäki et al., 2014). Hence, this inexperienced ones in their reporting of better
form of “mental disarray” occurs with a skills for regulating negative game-induced
substantial frequency, leading to substantial emotions. Some experienced players in
costs for those involved. These findings are in (Palomäki et al., 2013b) reported that their
line with the studies by Smith, Levere, and general emotion regulation skills had improved
Kurtzman (2009), as well as by Eil and Lien through playing poker. These players also thought
(2014), who used big data on millions of played that a clear understanding of mathematical
online poker hands and found that players tend concepts, such as variance, is related to a mature
to pursue risk when losing, but play cautiously disposition towards encountering ”bad luck”
when winning. This effect is possibly driven by (“luck doesn’t exist, only variance does”
emotional aversion to loss. [Palomäki et al., 2013b]), which suggests that in
Social cues may also interact with emotional poker, emotion regulation skills and technical
reactions during poker decision-making: In an skills are tightly intertwined.
online experiment employing a poker decision-
making task with mathematically defined optimal Social Nature of the Game
choices, inducing the feeling of anger (via reading In poker, the dynamics of social interaction—
emotional stories) reduced decision-making such as opponent characteristics or gender
accuracy. However, this effect was driven by a effects—are crucial in understanding decision-
social cue: displaying a pair of human eyes that making quality. The social setting of the game
“followed” the participants’ mouse cursor during also plays a significant role in biasing poker
the task (Laakasuo, Palomäki, & Salmela, 2015). decisions on the one hand, and on the other
What leads to such costly lapses in provides players with potentially accurate
judgment? In a qualitative study on poker information in the form of behavioral “tells”
players’ experiences of losing significant (Caro, 2003).
amounts of money, tilting was characterized by Schlicht, Shimojo, Camerer, Battaglia, &
feelings of anger, frustration, and significantly, Nakayama (2010) employed a simplified poker
injustice (Palomäki et al., 2013b; see also task and found that opponents whose facial
Barrault et al., 2014). Social elements such as expressions displayed more trust were more
unfriendly comments by other players often fuel often folded (given up) against. The authors
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
argued that by betting the opponent is implicitly simplified poker tasks, as previously discussed.
“sending a message” that he has a strong hand, The element of chance obfuscates empirical
and, because he looks trustworthy, the message assessment based on earnings and calls for very
is believed. In another study with a poker task long observational histories. It would be better if
involving repeated decisions against the same the probabilistic “goodness” of individual
opponent, participants’ decisions were more players’ decisions (e.g., the expected value in
strongly influenced by their opponents’ prior terms of monetary winnings) could be evaluated
actions when the opponents were represented as based on a reasonable number of hands played.
humans rather than as computers (Carter et al., The expected value of poker decisions can
2012). be evaluated in simplified scenarios (see
In both of these studies, the human Laakasuo, Palomäki, & Salmela, 2015; Leonard
opponents were presented as males. Poker & Williams, 2015). However, evaluating the
players seeking long-term engagement with the expected value of complex poker decisions “in
game value masculine identities and player traits the wild” is extremely difficult, given all the
(Wolkomir, 2012), and the vast majority (90- aforementioned cues potentially affecting (or
95%) of poker players are male (Palomäki et al., biasing) the players’ decisions and the element
2014; see also Abarbanel & Bernhard, 2012). of chance. One way to tackle this problem is by
Also, poker decision-making itself seems to be benchmarking poker players’ decisions against
gender-biased: In an experiment using realistic the best artificial intelligence (AI) poker
online poker tasks where opponents were programs. Somewhat recently, an NLHE AI not
represented as either male or female avatars, only won the 2016 Annual Computer Poker
participants (of whom 93% were male) bluffed Competition, but in 2017 defeated four highly
6% more frequently at online tables with skilled professional poker players in heads-up
female-only avatars compared with male-only (one versus one) matches for about $1.8 million
tables amounting to a significant difference over over 120 thousand hands7. Poker AI has thus
time (Palomäki et al., 2016). A majority of the been benchmarked against the highest human
participants also reported that the gender of their standard and proven sophisticated enough to
opponents did not influence their decisions to beat the very highest-performing human players.
bluff, which suggests an implicit (unconscious) Therefore, these programs can act as a
bias in bluffing female opponents, who might normative reference whereby the performance
have been perceived as “easier” targets than of sub-elite players at least can be evaluated.
males. This would be based on how well their decisions
Together, these results highlight the notion correspond to the consensual decisions of the
that the social nature of poker is a key element best AI. To our knowledge, such efforts have
in fully understanding decision-making quality not been made yet, highlighting a potential
and biases in the game. But turning it around, avenue for future research.
poker is a tool to study decision-making and
socially driven decisions in a market Development of Poker Expertise (RQ2)
environment-like scenario, which, to date, has Our second research question asked how poker
received relatively little attention in research. skills develop into expertise and how poker
allows for studying expertise. The complexity of
Measuring Poker Skill requisite technical knowledge in poker (as
Time, speed, or distance measures can be used explained in “Technical Skills,” above) is
in many sports for objective quantification of evident even in a simplified poker decision-
performance; and in chess—the game most making task, which we have provided in the
studied in cognitive science—Elo points provide Appendix. Poker also lends itself well to be
a high-validity measure of performance. In examined under theories of expertise. Due to
poker, however, skill-level is often assessed having a chance component embedded in a
indirectly by self-reported experience or well-defined rule structure, poker even helps
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
be made. Such walk-throughs may also isolate directions of causality between poker skill
subtasks, such as focusing on different stages of development and various psychological
the hand (e.g., decisions on the flop, turn, or characteristics could be fully evaluated only by
river). Second, there should be clear goals for employing a longitudinal study design, where
the players who aim for self-improvement. For poker players’ behavior is measured over
example, the feedback generated by playing the extended periods of time. To our knowledge, no
game might be positive for bad decisions such study currently exists and would thus offer
(winning money despite making a decision with a fruitful line for future research.
negative expected value), but posting such There is, however, a rich corpus of poker
situations online for scrutiny helps players self-coaching textbooks that focus on improving
discover the actual goodness of their decisions. one’s mental game skills. The authors of these
The proximal goal for players who seek books typically recognize emotion control as a
feedback on their decisions is often not merely highly significant element in poker skill
to enjoy poker or winning since they also post development (e.g., Angelo, 2007; Taylor &
hands where they have won but are uncertain Hilger, 2007; Tendler, 2011). Similar anecdotal
whether they played correctly9. Finally, there is evidence has emerged from Esports, where
the element of diligent repetition, as players professional teams and individual players have
who strive to get better keep posting poker been significantly more successful in tournaments
hands for scrutiny, which, in turn, helps them after hiring sports psychologists specializing in tilt-
increase their skills. management (theScore esports, 2019).
It should also be noted that because poker is Tendler (2011) draws from his experience as
a competitive game, skilled players might have a clinical psychologist working extensively with
an incentive not to help novice players to poker players and offers detailed guidelines and
improve—or even an incentive to hinder their techniques for players to improve their tilt
progress. Novice players aspiring to get better control. He views poor tilt control in poker as an
thus sometimes need to discern between misleading issue of consistency in individual performance
and accurate information disseminated in online level. Players perform better on some days than
poker forums (Talberg, 2019), as aspect of social on others—and the overall distribution of
skill learning. performance level forms a bell curve around the
average performance level for each player. For
Emotion Regulation Skill Acquisition players with poor tilt control, this distribution is
The consensus is that technical poker skills can wide, reflecting a large difference in
be learned via practicing and studying the game. performance level between their best and worst
However, studying poker alone is probably not possible performance. Players with proficient
enough to learn and improve one’s emotion tilt control, in turn, have narrower performance
regulation skills, because it is not easy to level distributions. In other words, their
“simulate in training” the loss of significant performance is more constant across time—they
amounts of money. play almost as well on their “worst day” as they
Traits such as low emotionality and low do on their “best day.”
tendency to self-ruminate are largely (possibly It is important to note that we do not claim
innate) predispositions that enable some people emotion regulation is an important “sub-skill”
to become good players; namely, those who can only in the game of poker. It probably has wide
endure the stressful learning period as well as relevance across a range of domains, especially
the unavoidable losing streaks. Personality, IQ, those dealing with risk and uncertainty. We are,
and other psychological traits, when measured however, proposing that the role of emotion
with standard psychometric instruments, are to a regulation becomes more pronounced in poker
large extent stable across time, and may be than most domains that have been used in
difficult to alter systematically through practice. cognitive science to study the nature and
However, the malleability of such traits, and the development of expertise. In other fields such as
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
Figure 2. Hypothetical role of emotion regulation (ER) skills in Music, Chess, and Poker among players
with high technical skills (“on their best day”) in their respective games. The lines depict the hypothetical
within individual variation across time in level of performance (i.e., consistency in level of performance)
for individuals with (1) low ER skills in Music, Chess, and Poker (dashed lines), and (2) high ER skills in
any field. For example, a technically proficient Poker player with low ER skills might sometimes perform
as well as those with high ER skills, but due to high variability in their performance level, they sometimes
perform as poorly as an amateur. This is not the case for technically proficient individuals with low ER
skills in Chess or Music, where performance level almost never drops significantly low. Note that for
simplicity, we assume that for good ER skills the profile of performance variability is the same across all
fields. We also note that this is a conceptual model; the level of individual performance with respect to
technical and ER skills in real life is likely somewhat more complex and nuanced.
Thus, some poker players who have unclear and an important venue for future
acquired a high level of technical skills (e.g., research.
through years of DP) might still struggle with
having highly inconsistent performance levels Skilled Intuition as the Interplay of Technical and
Emotional Skill
(dashed line for Poker in Figure 2). For these
A corpus of anecdotal evidence suggests that
players, technical skills alone are not enough to
since the poker environment is complex and fast
reach a high average performance level. The
paced, players need to trust their intuitions or
extent to which emotion regulation skills can be
“gut feelings” when making a decision (e.g.,
learned, and if DP would be a suitable
Brunson, 2005; Tendler, 2011). These feelings
framework to understand learning them, is
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
can also be called affective heuristics (Finucane articulated in more detail. In cognitive terms,
et al., 2000) —that is, “unconscious” processing skilled intuition can be viewed as a hallmark of
of task-relevant information experienced expert decision-makers across many domains,
phenomenologically as good or bad “feelings” but it can reliably exist only in environments
about a situation. It has been empirically with stable relationships or regularities between
established that chess masters, too, often rely on identifiable cues and specific events, actions and
an intuitive “feel” for different moves and outcomes, such as chess (Kahneman & Klein,
assessment of the “board as a whole,” especially 2009). These kinds of regular environments and
in the mid-game where options for various games are known as high-validity environments.
moves are astronomical (e.g., Chassy & Gobet, The opposite are low-validity environments,
2011; Gobet & Chassy, 2009) and it is futile to such as changes in political climates, where
attempt to work through the alternatives step- predictability of long-term outcomes from past
by-step in working memory. performance is limited, and any intuition-based
Cognitive modeling work suggests that chess judgment is likely to be flawed or biased
masters’ intuition relies on pattern recognition (Kahneman & Klein, 2009). The higher the
(“chunking”). Their intuition is a cognitive process validity of the environment, the better the
involving implicit memory and fast and automatic chances are for acquiring skilled intuition in that
procedural knowledge: the present board environment (e.g., chess, bike riding, or the
configuration is compared to a vast knowledge- game of djenga). DP may be seen as a means to
base of mid-game positions encountered over increase the validity of (some aspects of) the
uncounted chess matches, analysis of chess environment.
literature, and thousands of hours of playing chess Is poker a high-validity environment? Given
and solving chess problems (Gobet & Chassy, the strong element of randomness, specific
2009). This implicit information processing seems decisions consistently lead to specific outcomes
to also be accurate enough to assist in complex only over the long run. Learning poker strategy
decision-making in familiar domains. What would is therefore challenging because the process is
this type of “skilled intuition,” or more specifically masked by outcome variability (Figure 1). In the
implicit domain memory retrieval that benefits task short run, players will often receive positive
performance, look like in poker? We illustrate this feedback from bad decisions, which may elicit
with a quote from a two-time World Series of an illusion of skill (Bjerg, 2010), and vice versa,
Poker main event champion, Doyle Brunson obfuscating actual skill. Even after many hours
(Brunson, 2005, p. 542): of practice and play, players might have an
Whenever I . . . “feel” . . . I recall erroneous conception of their true skill, and the
soundness of their choices. Indeed, MacKay and
something that happened previously.
colleagues (2014) found that increased
Even though I might not consciously do
frequency and duration of poker play was more
it, I can often recall if this same play . . .
strongly associated with online poker players’
came up in the past, and what the player
perceived skill than with their objectively
did or what somebody else did. So,
measured skill. There are tools to measure one’s
many times I get a feeling that he’s
level of skill objectively in chess (Elo-ratings)
bluffing or that I can make a play and
but not in poker; due to this poker players are
get the pot. [My] subconscious mind is
also more biased in predicting their individual
reasoning it all out.
success in tournaments (Park & Santos-Pinto,
In this quote Brunson clearly alludes to what 2010). On the other hand, poker is based on a
can be called skilled intuition in the domain of deterministic system of rules, such that it is
poker, manifesting in episodic memory recall or predictable at some scale. Thus, poker—or any
gut feelings. The “feel” is, presumably, a other similar game where the goodness of
subconscious recollection of something that has decisions is defined only over the long run—
happened in the past, which cannot be
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
biases or errors in economic decision-making. about the “swings” of their fortune in poker
Research on this topic is prima facie interesting (Palomäki et al. 2013b), similar to the skill of
and relevant to understanding of, for example, experienced investors like Warren Buffet. Can we
stock market trading or risky decision-making in find in other domains similar results regarding
general. If poker for some players is about emotion-regulation skill: Namely that self-
pursuing “male glory” (Palomäki et al., 2016; reflection, emotional stability, and understanding
Wolkomir, 2012), is it also about “manly risk “variance in life” (a que sera, sera-type stoic
taking”? Do poker players with stronger mentality) protects against destructive emotions?
masculine identities make riskier decisions, and Players could be taught emotion regulation skills
when are riskier decisions better decisions? Since via, for example, mindfulness meditation (which
poker involves risk, it would also be useful to see has shown promise in improving emotion
how risk-taking, masculine identity and emotional regulation in a gambling context [de Lisle et al.,
volatility react with and possibly hinder rational 2012]); meditation-based intervention might
decision-making. Whether or not people make improve poker players’ decision-making.
risky decisions in economics and sports for Pinning down and measuring the elements
reasons of fame and glory is interesting for several comprising poker skills would also be informative
reasons. Do we want egoistically-motivated stock in the study of skilled intuitions and their
traders, or leaders who take risks to boost their acquisition. The role of skilled intuition, or gut
own self-image, if these motivations make them feelings, in poker decision-making has not been
blind to disastrous outcomes? empirically investigated. At what level of skill do
Poker research seems to have uncovered a gut feelings start being accurate enough to aid in
previously unstudied emotional state called “tilt” decision-making—or in other words, when will
—a specific type of moral anger—which could poker players start profiting from listening to their
possibly also be found in other areas of decision- intuition instead of losing because of it? In poker,
making and expertise, such as sports or finance. it is difficult to accurately estimate one’s own
What is going on in other domains of action when skill, since the observed outcomes of playing are
people lose control; for example, in stock market masked by variance. This creates extra pressure on
trading, golf, tennis, or racing (Wei et al., 2016)? players to deliberately self-reflect on their session-
The term “tilt” has also been adopted into by-session decisions without focusing too much
common use in the world of online gaming and on the actual results. These questions offer a
Esports (theScore esports, 2019). Is tilting a fertile and significant area for study that will serve
uniform phenomenon across of these domains; if to further integrate research on emotions and
so, how much of expertise within these domains decision-making. Table 2 presents our
depends on emotion regulation skills? conceptualization of poker skill and its sub-
There seems to be a zen-like quality in top components.
poker players who report not getting anxious
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
players in estimation bias and decision bias. Experience, Self-Rumination and Self-
Journal of Gambling Issues, 86-100. Reflection as Determinants of Decision-
Macnamara, B. N., Hambrick, D. Z., & Oswald, Making in On-Line Poker. Journal of
F. L. (2014). Deliberate practice and gambling studies, 29(3), 491-505.
performance in music, games, sports, Palomäki, J., Laakasuo, M., & Salmela, M.
education, and professions a meta-analysis. (2013b). “This is just so unfair!”: A
Psychological Science, 25(8), 1608-1618. qualitative analysis of loss-induced emotions
Macnamara, B. N., Moreau, D., & Hambrick, D. and tilting in on-line poker. International
Z. (2016). The relationship between Gambling Studies, 13(2), 255-270.
deliberate practice and performance in sports: Palomäki, J., Laakasuo, M., & Salmela, M.
A meta-analysis. Perspectives on (2014). Losing more by losing it: Poker
Psychological Science, 11(3), 333-350. experience, sensitivity to losses and tilting
Malmuth, M. (2012). Fundamentals of poker -- severity. Journal of Gambling Studies, 30(1),
General guidelines. TwoPlusTwo Magazine, 187-200.
8(6). Available at http://www.twoplustwo.com/ Palomäki, J., Yan, J., Modic, D., & Laakasuo,
poker-fundamentals/general-guidelines/poker- M. (2016). “To bluff like a man or fold like a
etiquette.php girl?”: Gender biased deceptive behavior in
McCormack, A., & Griffiths, M. D. (2012). online poker. PloS one, 11(7), e0157838.
What differentiates professional poker Parke, A., & Griffiths, M. D. (2011). Poker
players from recreational poker players? A gambling virtual communities: The use of
qualitative interview study. International computer-mediated communication to
Journal of Mental Health and Addiction, develop cognitive poker gambling skills.
10(2), 243-257. International Journal of Cyber Behavior,
Meinz, E. J., Hambrick, D. Z., Hawkins, C. B., Psychology and Learning, 1(2), 31-45.
Gillings, A. K., Meyer, B. E., & Schneider, J. Rubin, J., & Watson, I. (2011). Computer poker:
L. (2012). Roles of domain knowledge and A review. Artificial Intelligence, 175(5), 958-
working memory capacity in components of 987.
skill in Texas Hold’Em poker. Journal of Rhodes, M. (2010). Randy ‘‘nanonoko’’ Lew
Applied Research in Memory and Cognition, and his rise to multi-table poker fame.
1(1), 34-40. Resource document. OnlinePoker.net.
Meyer, G., von Meduna, M., Brosowski, T., & http://www.onlinepoker.net/poker-
Hayer, T. (2013). Is poker a game of skill or news/poker-pros-news/randy-nanonoko-lew-
chance? A quasi-experimental study. Journal rise-multitable-poker-fame/7327. Accessed
of Gambling Studies, 29(3), 535-550. October 15, 2019.
Moreau, A., Delieuvin, J., Chabrol, H., & Salen, K., & Zimmerman, E. (2004). Rules of
Chauchard, E. (2017). Online Poker Tilt Scale play : game design fundamentals (Vol. 1st).
(OPTS): creation and validation of a tilt London: MIT.
assessment in a French population. International Schlicht, E. J., Shimojo, S., Camerer, C. F.,
Gambling Studies, 17(2), 205-218. Battaglia, P., & Nakayama, K. (2010).
Moreau, A., Chabrol, H., & Chauchard, E. Human wagering behavior depends on
(2016). Psychopathology of online poker opponents' faces. PloS one, 5(7), e11663.
players: Review of literature. Journal of Siler, K. (2010). Social and psychological
Behavioral Addictions, 5(2), 155-168. challenges of poker. Journal of Gambling
O’Leary, K., & Carroll, C. (2013). The online Studies, 26(3), 401-420.
poker sub-culture: Dialogues, interactions Sklansky, D., & Malmuth, M. (1999). Hold’em
and networks. Journal of Gambling Studies, poker for advanced players, 21st century
29(4), 613-630. edition. Henderson, NV: Two Plus Two
Palomäki, J., Laakasuo, M., & Salmela, M. Publishing.
(2013a). “Don’t Worry, It’s Just Poker!”:
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
Smith, G., Levere, M., & Kurtzman, R. (2009). Thompson, J. J., McColeman, C. M., Stepanova,
Poker player behavior after big wins and big E. R., & Blair, M. R. (2017). Using video
losses. Management Science, 55(9), 1547-1555. game telemetry data to research motor
Talberg, N. (2019). Learning Poker in Different chunking, action latencies, and complex
Communities of Practice: A Qualitative cognitive‐motor skill learning. Topics in
Analysis of Poker Players’ Learning Cognitive Science, 9(2), 467-484.\
Processes and the Norms in Different van Loon, R. J. P., van den Assem, M. J., & van
Learning Communities. Journal of Gambling Dolder, D. (2015). Beyond chance? The
Issues, 42, 8-41. persistence of performance in online poker.
Taylor, I., & Hilger, M. (2007). The poker PloS one, 10(3), e0115479.
mindset: Essential attitudes for poker von Neumann, J., & Morgenstern, O. (1944/2007).
success. Dimat Enterprises Incorporated. Theory of games and economic behavior.
Tendler, J. (2011). The mental game of poker: Princeton: Princeton University Press.
Proven strategies for improving tilt Wright, R. (2001) Nonzero: The logic of human
control,confidence, motivation, coping with destiny. New York: Vintage.
variance, and more. Self-published. Wolkomir M. (2012). “You fold like a little
theScore esports (2019, March 24). What is girl:”(Hetero) Gender framing and
Tilt? The Gentle Art of Not Losing Your competitive strategies of men and women in
Sh*t in Esports [Video file]. Retrieved from no limit Texas hold em poker games.
https://www.youtube.com/watch?v=OvaPVB Qualitative Sociolology, 35 (4), 407–426.
FVUWQ .
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2
Palomäki et al. (2020) Poker as Domain of Expertise
given the community cards and the Opponent’s The analysis above is an oversimplification,
betting actions previously (and body language in and merely illustrates the complexities in poker
live poker; or chat comments in online poker, and decision-making. You as a player in the game
so on). should also consider bluffing on the river, if your
In the situation outlined in Figure A.1, you hand does not improve. Also, you could decide
can estimate how “strong” your own five-card to bet initially, or raise the Opponent’s initial bet
hand is against the “average strength” of the after checking. These would entail new
Opponent’s hand range. This estimation is probabilistic dependencies, which we have
sometimes done quickly and implicitly, because omitted. While this task analysis is hypothetical,
time pressure alone often prevents explicit it is an empirical question how explicitly
detailed calculations – skilled players sometimes analytical (or intuitive) players’ cognitive
play on multiple tables online, some on as many processes are in similar situations. Determining
as 24 at a time (e.g., Rhodes, 2010). opponents’ hand ranges in various situations is
discussed across poker communities (O’Leary &
Carroll, 2013).
Figure A.1. An online NLHE table (adapted from Palomäki et al., 2016). A: Opponents 1, 3 and 4 have folded
(given up) and are no longer contesting the pot. B: The total amount currently in the pot, which represents all the
money that has been previously waged during the current hand. C: The amount of money the player has remaining
in their stack, which represent the total amount they will be able to wage during any particular hand. D: The “hole
cards” of the Player, not visible to the opponents. E: The “hole cards” of the remaining opponent. F-H: The
“community cards” shared by the player and the opponent. F: The flop (three first “community” cards). G: The
turn (the fourth community card). H: The river (the fifth and last community card to be dealt, at this point
unknown).
https://www.journalofexpertise.org
Journal of Expertise / March 2020 / vol. 3, no.2