A program that teaches itself to play poker, created almost the best possible strategies for one of the versions of the game. This once again shows how modern technique can help us in finding solutions, for example, related to medicine or any other industry.
The program for two months played with me 24 trillion poker hands. Probably, this number is greater than everything ever played by people, says Michael Bowling, who led the project.
The final strategy will still not be winning in a hundred percent cases due to unsuccessful cards. But in the long run, she will be able to win thousands of games in a row! “We will be able to compete with the best players in the world, and they will still lose,” says Bowling, from the University of Alberta in Edmonton, Canada.
In any case, bowling doubts that the strategy will allow anyone to earn a state on the game. It works only when playing in the limited Texas Hold’em one on one, such a game has quite relaxed in popularity in recent years. Even on the Internet, with such a game there are small bets and, most likely, you will win a very little bit.
In the poker, a mathematical approach to the game has been used for quite a long time ago, called gaming theory and bowling reports that some methods that could be useful and in other situations could be introduced. Developers are studying the ability to help doctors, for example, determining the required dose of insulin for patients with diabetes mellitus. The game theory was also used to plan street patroling and can be useful in other areas, such as developing strategies for negotiation, auctions, cybersecurity, design of drugs and combating epidemics.
In poker on two players, each of them is heard two cards in the onetable and five others are put on the table, the front side up. Players make bets, then laid out, and dreams are distributed. Bar size is fixed. Bowling said his computer strategy is too complicated to someone who could learn, but his university has created a website where people can get advice on the game from the program or even try to play against it.
While scientists have created poker programs for years, the result of bowling is allocated by the fact that we are talking about the proximity to the creation of an optimal strategy. This is a difficult task, since Poker includes imperfection of information about the cards of the enemy.
Tuomas Sandholm from the University of Carnegie Mellon in Pittsburgh, who did not participate in the new job, calls these results of the milestone. He said that it was for the first time to a competitive game with imperfect information, was essentially solved.