Professional Documents
Culture Documents
Abstract
of Poker every year? What are they, the luckiest guys in Las Vegas?”
The popularity of poker has exploded in recent years.1 The premier event, the
World Series of Poker Main Event, which costs $10,000 to enter (buy-in), has increased
from a field of 6 in 1971, to 839 in 2003, and to 5619 in 2005. Broadcasts of poker
tournaments can frequently be found on television stations such as ESPN, Fox Sports, the
Travel Channel, Bravo, and the Game Show Network. These tournaments consistently
Poker has also garnered the attention of many influential economists and served
as a key inspiration in the development of game theory (von Neumann & Morgenstern
(1944), Kuhn (1950), Nash & Shapley (1950)). To economists, poker is a zero-sum game
of incomplete information requiring the use of not just probabilities, but beliefs in order
to play well. Dreef, Borm, & van der Genugten (2003, 2004a, 2004b) compute optimal
mixed-strategy solutions for simple versions of poker and show, theoretically, that skilled
playing for large stakes are able to converge to optimal strategies. If evidence of skill
1
PartyGaming, a leading online poker website, released an IPO in May 2005 which valued the company at
$8.46 billion.
2
“America’s Hottest Sport: Poker.” CNN/Money, July 9, 2004.
3
The authors note that the role of skill in poker has policy implications in that a Dutch gaming act requires
a license to exploit games of chance, but nor games of skill.
2
discrepancies is found among elite poker players, it suggests that not all players are
how many poker tournaments have these “stars” participated in which they did not do
well? Furthermore, even if poker competition among top players were random, we
would expect a few players to get lucky and do well in multiple tournaments. Using data
to properly test for whether skill differentials exist among top-place finishers of high-
stakes tournaments. We find evidence of substantial skill differences among these top
players. To better understand the implications and magnitude of our results, we replicate
our empirical analysis using data on outcomes from professional golf tournaments.
Identifying skill discrepancies among top poker players is complicated by the lack
of precise tournament data. The lists of entrants for large poker tournaments are not
available and typically outcomes are only recorded for players that finish in the final two
or three tables. Thus it is not possible to know the total number of tournaments for which
focus our attention on the individuals who finished in the top 18 in high-stakes
tournaments.4 Thus, while we are unable to identify the number of tournaments that an
4
A full poker table consists of 9 players. Thus, the top 18 represents the final two tables. Almost all recent
high-stakes tournaments record the outcomes of the final two tables.
3
individual has played in, we are able to identify the number of times that a player has
We use data on limit or no-limit Texas Hold’em tournaments that are a part of the
World Series of Poker, World Poker Tour, or World Poker Open.5 Using information
tournaments since 2001 that had at least a $3,000 buy-in.6 A total of 81 separate poker
tournaments fit these criteria. Table 1 presents summary statistics for the players in these
tournaments.
tournaments in 2005. We record the name and final rank of each player that finished in
the top 18 in each tournament.7 Table 1 provides summary statistics for the golf players
in these tournaments.
information to predict the outcome of individuals conditional on being among the final 18
contestants of a given tournament.8 Being able to predict tournament success implies that
outcomes are not random conditional on being in the top 18. We explore two sources of
5
Texas Hold’em is a variant of poker in which all players are given 2 cards; the goal is to make the best 5
card hand from the 2 cards and 5 community cards which apply to all players’ hands. Betting occurs after
each player receives his or her cards, again after three of the five community cards are revealed, again after
the 4th community card, and finally after the 5th community card. In limit Texas Hold’em, the bet amounts
each round are fixed, whereas in no-limit Texas Hold’em a player can wager as many chips as he or she
wants above a set minimum wager.
6
For a small number of tournaments after 2001 (and for all tournaments prior to 2001), the top 18 finishers
were not recorded or not available and thus were not included in the analysis.
7
In golf, there are often ties. We record an average rank for these situations (e.g. if two players tie for 3rd
place, each player is given a rank of 3.5).
8
We do not predict success using future results since the decision of whether to play or not in a tournament
is endogenous to the outcome of previous tournaments.
4
past information that may predict current success: having played in previous tournaments
and performance in the previous tournament played. The basic specifications that we
employ are
where Rank it represents the ending rank (1st through 18th) of player i in tournament t.
the top 18 of another tournament prior to tournament t. One Prev it , Two Prev it , and
Three or More Prev it are dummy variables set equal to one if individual i placed in the top
previous tournament placements serves as the base group). Top7 - 9it , Top4 - 6it , and
Top1 - 3it indicate if individual i placed in the top 7-9, 4-6, or 1-3 in the most recent
tournament for which individual i placed in the top 18 (missing if t is the first top 18
If there are no skill differentials among poker players who made one of the final
two tables in these high-stakes poker tournaments, we would expect all of the coefficients
indicate that, conditional on making it to the final 18, winning is random. However, if
some players are more skilled than others among the top 18 finishers, we would expect to
find negative coefficients in the above specifications (past experience and success should
be associated with a reduction in rank – e.g. from 7th place to 6th place).
5
Table 2 presents the results from specifications (1) – (3) for both the poker and
golf data. Robust standard errors are presented in brackets below the coefficient values.
Beginning with the poker results, Column (1) indicates that having placed in the top 18 in
a prior tournament is associated with a significant decrease in rank of .78 (this represents
an improvement in rank – e.g. moving from 3rd place to 2nd place is a decrease in rank).
Column (2) indicates that the more tournaments for which an individual has placed in the
top 18, the better the predicted rank for the current tournament. For example, individuals
who placed in the top 18 in three or more previous tournaments are predicted to finish
one rank better than individuals who never placed in the top 18 in a previous tournament.
Column (3) attempts to predict current rank based on the rank that a player placed in a
previous tournament.9 The coefficients suggest that ranking in the top 1-3, top 4-6, or top
7-9 in the previous tournament predicts a better rank in this tournament relative to
individuals who placed 10th-18th in the previous tournament. While the coefficients are
negative and increasing with last tournament rank, only top 1-3 is statistically significant;
individuals who placed in the top 3 in the last tournament on average end 2.5 spots higher
in the current tournament (relative to individuals who placed 10th -18th in the last
tournament). The results in Column (3) are of particular interest because the sample is
reduced to individuals who have placed in the top 18 in at least two high-stakes
tournaments. These individuals should all be experienced players and yet still can be
Columns (4) – (6) present the analogous results using data on professional golfers.
The results are qualitatively similar to the poker results. Indicators of past experience in
9
By “previous tournament,” we are recording the last tournament in which the player participated and
finished in the top 18.
6
columns (1) and (2) are approximately twice as predictive of current golf success relative
to the same indicators of past experience for current poker success. Last tournament
ranking is less predictive of current golf success relative to last tournament ranking of
the final two tables in high-stakes poker tournaments. The results are similar in size to
regarding the cause of these discrepancies. Our main hypothesis is that even though
poker is simple to learn and play, optimal strategies are very complex and require
discrepancies is that some players are better at exploiting unintended signals made by the
opponent. In classic poker lore, this is referred to as “reading” another player’s “tells.”
The skills required by poker players such as the understanding of probabilities and
beliefs are similar to those needed by agents in real markets such as price setters and
litigation negotiators. Our results suggest that executing optimal game theoretic solutions
in situations like these may be very difficult for people, even those who are experienced
and select into playing these types of games for high stakes.
10
We consider this explanation to be unlikely due to the fact that all players face the exact same incentives
structure, which provides very large incentives to achieve the highest rank possible rather than make a
“respectable showing.” Both winnings (and fame) grow exponentially as a player places higher up in the
rankings.
7
References
Dreef, M., P. Borm, and B. van der Genugten. “A new relative skill measure for games
with chance elements.” Managerial and Decision Economics, 2004a.
Dreef, M., P. Borm, and B. van der Genugten. “On Strategy and Relative Skill in Poker.”
International Game Theory Review, 2003.
Dreef, M., P. Borm, and B. van der Genugten. “Measuring skill in games: several
approaches discussed.” Mathematical Methods of Operations Research, 2004b.
Kuhn, H.W. "A Simplified Two-Person Poker." Contributions to the Theory of Games,
I. H.W. Kuhn and A.W. Tucker, eds. Annals of mathematics Studies, Number 24.
Princeton, New Jersey: Princeton University Press. 1950, pp. 97-103.
Nash, J.F. and L.S. Shapley. "A Simple Three-Person Poker Game." Contributions to
the Theory of Games, I. H.W. Kuhn and A.W. Tucker, eds. Annals of mathematics
Studies, Number 24. Princeton, New Jersey: Princeton University Press, 1950, pp. 105-
116.
8
Table 1. Summary Statistics
Poker Golf
% with 1 top 18 finish 70.1 24.3
% with 2 top 18 finishes 14.7 14.7
% with 3 top 18 finishes 6.9 18.9
% with 4 or more top 18 finishes 8.3 42.2
Number of tournaments 81 48
9
Table 2. Predicting Poker and Golf Success
Dependent Variable: Rank (1st - 18th)
Poker Golf
(1) (2) (3) (4) (5) (6)
Experience -0.781 -1.420
[.274]*** [.384]***
1 Prev -0.558 -1.323
[.361] [.500]***
2 Prev -0.920 -0.565
[.474]* [.548]
3 or More Prev -1.000 -1.865
[.415]** [.438]***
Top7-9 -0.137 0.107
[.598] [.582]
Top4-6 -0.552 1.170
[.595] [.697]*
Top1-3 -2.510 -0.440
[.554]*** [.595]
Constant 9.810 9.810 9.660 10.270 10.268 8.770
[.172]*** [.173]*** [.310]*** [.323]*** [.323]*** [.271]***
R-Squared 0.005 0.006 0.035 0.016 0.024 0.007
Observations 1494 1494 595 811 811 593
st th
Notes: Columns (1) - (6) present coefficients from the regressions with finishing rank (1 – 18 ) as the
dependent variable. Experience is an indicator that equals one if the player had previously finished in the
top 18 of a tournament in our sample. 1 Prev, 2 Prev, and 3 or More Prev indicate the exact number of
previous tournament finishes in our sample. Top 7-9, Top 4-6, and Top 1-3 are indicators that equal one if
the player finished in one of those spots in the previous tournament (missing if no previous tournament in
our sample).
* significant at 10%; ** significant at 5%; *** significant at 1%
10