The poker world will never be the same as before. Researchers at the University of Alberta in Edmonton created Cepheus, the program poker game that … can not lose. This is the result of over a decade of work with the developers of Computer Poker Research Group, in collaboration with the Finnish computer scientist, Oskari Tammelin. He writes about it in the latest issue of the journal “Science”.
Cepheus game in double version of the most popular poker games Texas Hold’em. Implements the strategy, according to which in the long run can not lose money. Though its creators admit that is not fully perfect, its imperfection is so small that even playing against him for life (60 million hands), we would not be able to see her.
Computer programs have long surpassed human capabilities in games such as chess, checkers, or Othello. These games are different, however, from the poker that’s where all players have complete information about the situation on the board, or the board and on its basis can make decisions. In poker is obviously different, key card information rivals players do not have. The computer also.
Cepheus must therefore take risks, and in order to achieve mastery, learn from mistakes. The program was built so that it can cope without the help of man, after the rules of the game, played with himself unimaginably large number of hands and storing the results, improving its capabilities. Canadians reported that Cepheus was playing with himself over a billion billion times, and gained from this experience compress the disk in the form of 11 terabytes of data. In other words, Cepheus poker game played more than humanity throughout its history. Importantly, the program can also bluff and fairly precise estimates when he was worth.
Cepheus is unlikely to reduce the interest in poker, the possibilities are even for experienced players completely inaccessible and can therefore help in training. On the website of its creators, anyone can try. Moreover, in real competitions often involved more than two players, which makes it difficult to analyze such games. The authors not only count on a successful card, the algorithms can also assist you in making other decisions, for which the knowledge is not complete. How do they think appropriate elements of the program can be used, for example, when investing, auctions, medicine, and even with … securing air traffic.