Poker bots are getting better

Not long ago, but merely 2 years we were introduced to a poker bot named Cepheus, which was developed by the Computer Poker Research Group (CPRG) at the University of Alberta. The team, which was led by Dr. Michael Bowling and included Neil Burch, Michael Johanson, and Oskari Tammelin, published a paper “Heads-up limit hold’em poker is solved” in Science, where you can read all the details about it.

“While Cepheus does not play perfectly, it is so close to perfect that even after playing an entire lifetime of poker against it — over 60 million hands — it is impossible to tell the difference. It is essentially perfect.”

“Cepheus accomplished this goal with no human expert help, only being given the rules of the game. It was trained against itself, playing the equivalent of more than a billion, billion hands of poker. With each hand it improved its play, refining itself closer and closer to the perfect solution. The program was trained for two months using more than four thousand CPUs each considering over six billion hands every second! This is more poker than has been played by the entire human race.”

You can also play Cepheus on their website and examine his strategy in detail.

But can poker bots defeat pro players in NL Texas Hold’em?

Well I can quite understand solving pot limit texas hold’em head’s-up, as the number of choices is considerably limited, but recently we are hearing about new poker bots being trained in no limit texas hold’em.

A team of researchers recently published a paper called “DeepStack: Expert-Level Artificial Intelligence in No-Limit Poker”, where you can read:

“There has been some AI success in heads-up limit

Texas hold’em, the smallest variant of poker played by humans. However, heads-up limit Texas hold’em has

fewer than 10^14 decision points. By comparison, computers have exceeded expert human per-

formance in go a perfect information game with approximately 10^170 decision points. The comparable variant of poker is heads-up no-limit Texas hold’em (HUNL). In HUNL, the two players can bet any number of chips, resulting in over 10^160 decision points”

“We evaluated DeepStack by playing it against a pool of professional poker players recruited

by the International Federation of Poker(…)

In total 44,852 games were played by the thirty-three players with 11 players completing

the requested 3,000 games. Over all games played, DeepStack won 492 mbb/g(…)

In the poker community, it is common to use big blinds per one hundred games (bb/100) to measure win rates, where 10 mbb/g equals 1 bb/100.”