Earlier this month, a team of researchers at the University of Alberta announced that they have developed a computer program that can “play an essentially perfect game of poker.” The program, named Cepheus, was developed by Michael Bowling, Neil Burch, Michael Johanson, and Oskari Tammelin.

Unlike other attempts to build a poker playing bot, Cepheus was not aided by expert players. Instead, the programming team simply gave the program the rules of the game and let it play against itself. After each training round, Cepheus would analyze its decision-making, calculating something called “regret” (essentially how much it lost in a given hand for making the wrong move). It adjusted its strategy over time, learning from every session, until the regret measure was whittled down to nearly zero.

The regret could not get all the way to exactly zero because of the nature of poker, a game in which players have imperfect information, but because of how close to zero Cepheus was able to get that regret figure, the research team told Science Magazinethat there was no way a person could beat the machine over the course of a lifetime. Certainly, the program could lose in the short-term, but over the long-haul, the Alberta team sees Cepheus as completely unbeatable.

The amount of poker Cepheus played to get to this point is astounding: a billion billion hands. To put this in a little perspective, PokerStars dealt its 100 billionth cash game hand in 2013. That’s all players over the life of the site. Additionally, researchers used the example of a “human lifetime” of hands as 200 hands an hour, 12 hours a day, for 70 years, and that only gets to a bit over 60 million hands. A billion billion is just gross.

“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 4,000 CPUs each, considering over six billion hands every second. This is more poker than has been played by the entire human race,” Bowling wrote in the team’s published report.

Interestingly, Cepheus only learned how to play a very specific type of poker: Heads-Up Fixed-Limit Texas Hold’em. Poker is so complicated that trying to perfect Full Ring or No Limit would be exponentially more difficult. Even so, Cepheus learned how to make complex decisions such as bluffing.

For all of its decisions, the computer runs the math and comes to the solution “always fold,” “always raise,” or “raise or fold according to an optimized percentage.” It will occasionally just call, but that is almost always when the pot is capped. Rarely will it just call an initial raise.

Cepheus is the successor to Polaris, another poker program developed by the University of Alberta team last decade. It was very advanced, but not to the level of Cepheus. In July 2007, professional poker players Phil “The Unabomber” Laak and Ali Eslami took on Polaris heads-up, beating it in two 500-hand matches, losing one and tying one.

The following year, an improved version of Polaris played six sessions, totaling 6,000 hands, against six professional players. The computer won three matches, lost two, and tied one. It won 195 big blinds across all 6,000 hands.

Want the latest poker headlines and interviews? Follow PocketFives on Twitterand Like PocketFives on Facebook. You can also subscribe to our RSS feed.