An Academic Approach to Poker (Gets Dumbed Down)
Noticed an item in yesterday’s USA Today about online poker, a reference to a newly-published study about online poker called “Social and Psychological Challenges of Poker” by Kyle Siler, a doctoral student in sociology at Cornell University. As usually happens with these articles that try to summarize a discipline-specific study for a wide audience, the USA Today piece boils Siler’s article down to one simple, easy-to-digest claim, essentially announcing that it shows “Poker wins often lead to bigger losses.” In other words, the USA Today article makes it sounds as though Siler’s exhaustive study of a large sample of online poker hands proves that players who win a little tend to lose it back and then some -- confirming, in a way, the fears of those who object to poker and/or gambling as an inevitable road to ruin, regardless of one’s short-term successes.
The USA Today article is accompanied by a picture of 2009 World Series of Poker champion Joe Cada, who does not actually figure in the piece. I suppose there could be some implication here being advanced about the possibility of Cada’s not holding onto his winnings, but I think it more likely this was the first available poker-related photo that came up following a quick search.
I was curious to read the study, especially because the way it was presented in the USA Today article seemed more than a little sketchy. Took a little bit of work to get a look at it, but I did manage get a copy and have now read it through.
Siler speaks knowledgeably of poker and the online game, and as far as I can tell seems to be operating within accepted expectations for sociological research and argumentation. He also shows an understanding of economic theory and applies some of those ideas when appropriate. Siler additionally brings in many references to “poker literature” -- both to strategy texts and to narratives -- which help ground the study within conversations about theory and practice with which we poker players who have read those books are versed. Those references to people like Sklansky, Caro, Harrington, Brunson, and others also make the article more fun for poker players to read than probably would be the case with most dry, academic treatises.
All in all, I think Siler’s article is smart and interesting, and while its findings mostly confirm ideas we already had about what strategies are most successful his study is nonetheless useful for its having found support for those ideas in the data. I also think it is obvious that the USA Today writer probably only skimmed the study, coming away with a vague, uncertain understanding of its findings.
Here, let me take a shot at summarizing this sucker a little more carefully...
“Social and Psychological Challenges of Poker” by Kyle Siler appears in the latest issue of the Journal of Gambling Studies. Using Poker Tracker, Siler examines roughly 26.9 million hands of online poker played over a five-month period at different stakes in order to try to determine “which strategies are conducive to winning at these different levels,” and therefore perhaps draw some conclusions about the “social and psychological challenges” the game presents. The game on which Siler focuses is short-handed (6-max.) no-limit hold’em, and the hands he’s looking at come from games played at NL50, NL200, and NL1000. In all, he was able to gather and analyze stats on around 295,000 different players.
After crunching the numbers with Poker Tracker, Siler reaches a few conclusions which he does a good job explaining, also using charts and graphs to help him further illustrate what he has found. Those conclusions include his having discovered that
It is this last item that I think caused the USA Today writer to stumble a bit. The point there is that in no-limit hold’em winning a lot of pots doesn’t mean one wins a lot of money, and, in fact, when one looks at millions of hands like Siler did, one discovers that the big winners tend to win fewer (though bigger) pots relative to the rest of the player pool. That correlates with the first finding, namely, that the tight-aggressive strategy has the best return.“tight and aggressive strategies have the best return across all levels”; one finds “an increased proportion of aggressive players as one moves up stakes,” where also “the number of passive players decreases”; there is an “overrepresentation of loose and aggressive players” among the biggest winners and the biggest losers at all stakes; “None of the biggest winners at any of the levels were even close to being in the top hundred win rates,” having made their money via higher volume (the “grinders”); “a high win percentage (i.e. the percentage of total hands won by a player) is negatively correlated with win rate.”
The USA Today writer took that point and mangled it into a declaration that “players who played a lot of hands and often won ended up losing more money than others.” He then quotes from Siler’s study in a way that makes it sound like Siler is agreeing with that somewhat vague claim, but he’s misrepresented Siler’s findings.
Siler does conclude the study with some thoughts about how moving up in stakes presents players with new challenges, and does a nifty job relating how the stress of adapting one’s style -- necessary to succeeding at the higher levels -- presents especially difficult “social and psychological challenges” to poker players. He ends with the point that “the biggest opponent for many players is themselves,” an idea familiar to any poker player who has struggled to improve his or her game.
Like I say, a smart study that I recommend if you can somehow get access to a copy. And I’m sorry to see it somewhat misrepresented in USA Today the way it was -- that is, as seeming uncritically to confirm fears about poker as just another unhealthy avenue to degeneracy and self-destruction -- and thereby soliciting further misinformed, unrelated comments like “this is the reason why the house always win[s]” and “that is why they call it gambling.”
Labels: *the rumble, Kyle Siler, USA Today
posted by Short-Stacked Shamus at 1/13/2010 09:36:00 AM
11 Comments:
Time's version of the Siler report:
www.time.com/time/health/article/0,8599,1953205,00.html
Good post, but Siler's wrong in his explanation why players that win the most pots lose money. It's not necessarily because they lose a few bigger pots than better players, it's because they play hands a better player would have folded and some of those hands end up winning, either by bluffing out their opponents or winning a showdown.
So what else is new? A journalist takes a statement out of its intended context in order to support a pre-determined, often reductivistic conclusion.
I haven't read the original paper but from the quotes I would say that Siler's writing is itself to blame for the reaction. It seems like his explanation wasn't clear and there was too much focus on this one particular aspect of his research.
That people who win more hands lose money is only paradoxical to a beginning or amateur player. To a game theoretician, even one that knows nothing about poker, it is obvious. There's a maximally exploitative strategy against any particular set of opponents, which will result in a particular percentage of hands won. If you win a higher percentage of hands, then you're making mistakes and not winning as much money as a perfect player. The same thing is true if you're winning too few hands, although you may still be a winning player if you're playing against players who are too loose. If you were to graph profit against percentage of hands won, you'd find a peak somewhere.
Maybe Siler provided more sophisticated statistical analysis that backs up his hypothesis but I would think that the data is inconclusive. Many of the best players win by stealing lots of small pots which more than makes up for losing more large pots.
The Time article is wrong in applying Siler's hypothesis to investing. It's not true that "The more small returns you get from your small investments in stocks, the likelier you are to make — and lose — a big investment." It is true that often people downplay the risks of large losses, and a steady stream of small gains makes this more likely, but many small returns does not make a big loss more likely. Even when there are big losses after a stream of steady small gains, that doesn't mean the bet was negative expectation.
I have personally accumulated a fair amount of anecdotal evidence to suggest that losing players are losers because they play bad.
Seriously, the Time and USA Today pieces make it sound as if all these losers would be winners if they would just surrender some random pots here and there.
"Given the huge role of luck in delivering big payoffs and big losses, the best poker players must learn to keep winning or losing in stride"
Oh fuck you USA today. I really can't stand their so called "journalism" and it really seems like their newspaper in specific has helped lead to a dumbing down of our culture.
Love the blog by the way, Feel free to check mine out at masterpiecepoker.blogspot.com. I'll have a link up to yours soon.
I absolutely love this type of analysis of poker and related theory. As Andy points out, there are game theory errors in the conclusions made by the paper. However, what continues to be the wild card, as Andy also points out, is that a player's mind cannot be accounted for in empirical studies. And furthermore, the reasons each of the us play poker have a huge effect on our performance during that particular session. I examine the reasons why we all play poker in my blog.
Stay focused and stay the course.
Jonathan
http://WhyDoIPlayPoker.net
@ Shamus: I just stumbled upon your write-up of my research, and I have to thank you for engaging my work thoughtfully and in-depth. Watering things down is unfortunately often a necessary evil with working with the media. However, one of the reasons I was willing to endure that is because I was hoping to reach the poker community, which would have been much more difficult by leaving the article in an academic journal. The veracity of this research is as much driven by poker as it is by sociology, psychology et al. Kudos on an excellent blog as well; I’m glad I found it!
@Andy: I'm trained as sociologist, so my inclination is to look at the general demographics of the population. Revealing and examining the “demographics of decision-making” was the intent of my study; not LAG and "high % of hands won" strategies are generally more challenging (but potentially lucrative) to play profitably. On the whole, more players at NL50, NL200 and NL1K lose money with "high win%" strategies. However, this does not imply that "high win% = losing $$" is an invariable law for all players, or that there do not exist profitable niches for other strategies, if played well. Somewhat analogously, while LAG and Semi-LAG strategies *on the whole* were losing strategies, most of the highest win-rates were derived from these strategies. While this could in part be explained by higher variance, and the fact that such players play more hands dealt in the sample, I also figured this had to do with the fact that a few LAG players are skilled and are capable of earning high win-rates with this difficult strategy. As you rightly pointed out, many of the best players are winners via stealing lots of small pots, while perhaps surrendering the occasional home run. However, these data suggest that for every LAG or high win% strategy “shark”, there’s probably three or four “fish” enacting such strategies in an unskilled manner.
I would also add that the strength of the negative relationship between win% and win-rate significantly declines as one moves up from NL50 to NL1K. I did not have data above NL1K, but I would not be surprised if the negative relationship completely attenuates, or even reverses to become positive at higher-limits, and perhaps even earlier in heads-up games. Relatedly, I was recently asked by a local journalist reporting Brian Hastings' recent $4.2 million winning session, and asked if the “high win% lose money” causal relationship that diffused through the media was applicable to the story, and I replied that I strongly doubted it, given the trends between NL50 and NL1K in my research, in addition to qualitative accounts of how players win money in the nosebleeds.
For the average research subject in a psychology lab, the incentive structure of “frequent small gains/occasional large losses” is one that is difficult for most to handle profitably. Off the top of my head, this has been shown in psychological research using Kahneman and Tversky’s prospect theory, and various incarnations of the Iowa gambling task. My results suggest for that for most players (particularly unskilled ones), the poker table isn't much different than the psychology lab. Part of the point of my article was to argue that good poker players are able to break through and exploit these typical hardwired shortcomings in human reasoning. Since most humans have problems identifying profit and losses accurately with these sorts of frames, I think it is reasonable to suggest that this is one explanatory factor for why some losing players bleed money to savvier players.
(continued below...)
I’d imagine you have a lot of data available at Full Tilt, and you may do demographic analyses of your clientele, and know things about the poker ecology of a site that a researcher like myself cannot surmise at this point. On the whole, I think the general notion of “high win% -> less money” works for the average player. However, winners in poker, and success stories in society tend lie outside of the median demographic, so to speak. Hence, for skilled players, there are almost certainly profitable equilibria and niches outside of the general strategic trends. Of course, as mentioned, I argue the exploitation of these trends and tendencies in human reasoning are part of being a successful poker player, underpin and augment the nuts and bolts of the strategic card game.
I’m grateful to engage with serious poker players on a more nuanced and academic level, particularly because poker is such a nuanced and theoretical game, which the media and academics don’t always entirely grasp. I have plans to do more work on poker, gambling and games in the future, so I am very interested in how I can improve my theories and methods.
Kyle Siler
Department of Sociology
Cornell University
ksiler[at]gmail[dot]com
I have read the paper with great attention and would like to make a comment about one of your conclusions:
The finding that the more hands a player wins, the more money the player loses is very interesting. However your conclusion is flawed (“players who win large proportions of hands lose money, because they lose a few hands of great value. In other words, these players overweight frequent small gains vis-à-vis occasional large losses.”)
The issue lies in the statistic that you use. The win percentage statistic is highly correlated with the percentage of hands a player invests in. As losing players tend to invest in a higher percentage of hands, the likelihood of winning a hand (i.e. win percentage) increases.
While your conclusion makes a lot of sense, it is not backed up by any data. It could also be true that losing players win and lose pots of the same size, but lose pots more often.
Regards,
Patrick Koch
Post a Comment
<< Home