Poker Far From Being A Solved Game

Home » Poker News » Poker Far From Being A Solved Game

Wednesday, June 24th, 2015 | Written by Adam Brown

In 1997, supercomputer Deep Blue defeated Grandmaster Garry Kasparov in the second of their celebrated chess matches, and in so doing inspired other scientists to develop more and more sophisticated artificial intelligence (AI) programs to beat their human opponents. It wasn’t long before computer scientists subsequently tried devising programs to solve incomplete information games such as poker, and this year the University of Alberta Computer Poker Research Group (UACPRG) surprised everyone by announcing its program “Cepheus” had essentially solved heads-up limit Texas hold’em.

Recently, the Carnegie Mellon University School of Computer Science then sent its new program called “Claudico” to do battle against four top poker professionals at heads-up no limit Texas hold’em, and while thus far the humans still have the upper edge, studies such as these raise the question of whether all formats of poker games will eventually be solved, and what will such progress mean for the average poker player.

Heads-up Limit Hold’em Poker Solved

While simple board games such as tic-tac-toe and checkers have already been solved, the more complicated game of chess has only been partially solved as it would require a supercomputer beyond out current capabilities to accomplish the task. A game of incomplete information, such as poker, requires still more work, which is why the gaming community reacted with such surprise when UACPRG announced it had solved ‘heads-up limit’ poker. The computer known as Cepheus had spent two months playing 24 trillion hands of poker per second for two months, and at the end of the study project manager Professor Michael Bowling, said:

“Even if you played 60 million hands of poker for 70 years, 12 hours a day, and never made any mistakes, you still wouldn’t be able to say with statistical confidence you were better than this program.”

Nevertheless, limit heads-up Texas Hold’em is the simplest of all poker variants to compute, and a far more sophisticated program is needed if poker games involving more opponents and no limit betting is ever to be solved.

Heads-up No-Limit Hold’em More Complicated

In April and May 2015, the Carnegie Mellon University School of Computer pitted its poker program “Claudico” against Doug Polk, Dong Kim, Bjorn Li and Jason Les in a contest held at the Rivers Casino in Pittsburgh called “Brains vs. Artificial Intelligence”. The Heads-Up No Limit Texas Hold’em contest was held over 80,000 hands, with each pro playing 1,500 hands per day, or roughly 12 hours, and a $100,000 reward awaited the poker pros should they win. By the mid-way point, however, the human opponents had pulled away to an almost unassailable lead and professor Tuomas Sandholm and his team were already plotting a rematch, once some much needed tweaks had been made. of course. Despite their winning ways, however, the poker pros seemed to acknowledge that the time would arrive when computers will ultimately surpass their human creators, and as Doug Polk, 26, commented at the time:

“I know computers will eventually be able to beat humans. But I hope we can make them go a few more rounds after this before they do, like Kasparov did.”

Should Players Be Worried?

Having read this far, readers may be worried that they could soon find themselves playing against unbeatable poker bots online without knowing it. So far, that worry is unfounded as not only do poker operators go to great lengths to detect and remove bots from their sites, but a solution to no-limit texas hold’em, even with only two players at the table, is still far, far away, let alone solving a game with multiple opponents.

Moreover, as long people are playing poker then the game will remain infinitely complex and difficult to master as human nature is far from being solved. As humans, we all have areas of vulnerability and often play in ways which are far from optimal, or alternatively we can draw upon intuition and experience to make inspired decisions, all of which is difficult for a computer to predict. In other words, while programs have learned to play perfect poker against other bots, against a human opponent computer programs may never be able to resolve the conundrum which is human nature and unpredictability.