It’s no secret that artificial intelligence improves every day. Centuries from now, after humanity has done its part to make Earth uninhabitable for all but the heartiest flora and fauna, we will all be dust and the machines will remain. For now, however, we still have poker: Over the course of two weeks and 40,000 hands, a team of top poker pros have out-earned Claudico, the artificial intelligence developed by scientists at Carnegie Mellon University.
The catch? The victory is so slim that it’s not statistically significant, so it’s technically a tie.
“We knew Claudico was the strongest computer poker program in the world, but we had no idea before this competition how it would fare against four Top 10 poker players,” said Tuomas Sandholm, the CMU professor of computer science who directed development of Claudico. “It would have been no shame for Claudico to lose to a set of such talented pros, so even pulling off a statistical tie with them is a tremendous achievement.”
By the end of the tournament, human player Bjorn Li led the way with $529,033 in chips, followed by Doug Polk with $213,671 and Dong Kim with $70,049. Jason Les brought up the rear, trailing Claudico by over $80,000. The games were played in “heads up” one-on-one format, with Claudico playing 20,000 hands against each of the four pros. All four players are ranked in the world’s top 10 for that style of play. The pros raked in a collective $732,713 – more than Claudico, but not enough to be statistically significant.
Poker is unique in that it’s an incomplete information game – in Texas Hold’em, players can only know the two cards they hold, plus the five communal cards that are eventually turned over. They must base their decisions on the probability of other players having better hands than they do. Making those kinds of semi-blind decisions is difficult for humans and computers alike, which is why the Carnegie Mellon team aimed to create a program to help.
“Beating humans isn’t really our goal; it’s just a milestone along the way,” Sandholm said. “What we want to do is create an artificial intelligence that can help humans negotiate or make decisions in situations where they can’t know all of the facts.”
In popular culture, seasoned poker players deliver platitudes about poker is about “playing the man across from you,” and other pseudo-mystic ideas. In reality, poker is more about science than art: The best poker players fold more hands than they play, sticking to the odds and mentally calculating the statistical odds of other players having better hands with the available cards. Even with incomplete information, it doesn’t get much better than a computer when it comes to calculating odds in a 52 card deck.
The human element, bluffing and things of that nature, are more relevant in the “heads-up” style of play used in the experiment, when players must be more aggressive. Whether Claudico can successfully mimic that behavior remains to be seen. It’s strategy is determined by algorithms (versus pre-programed inputs), and sometimes it works, and sometimes it doesn’t.
“There are spots where it plays well and others where I just don’t understand it,” he added. Some of its bets, for instance, were highly unusual, in Polk’s estimation. Where a human might place a bet worth half or three-quarters of the pot, Claudico would sometimes bet a miserly 10 percent or an over-the-top 1,000 percent. “Betting $19,000 to win a $700 pot just isn’t something that a person would do,” he observed.
Overall, the team considers Claudico’s performance a resounding success. The human players received appearance fees out of a $100,000 prize purse. Claudico left researchers with over two terabytes of data, far more than they could ever pour over themselves.