Poker-playing computer program could defeat best human players

Science : For the first time, a computer algorithm has been developed to solve a complex game that involves hidden information and uncertainty. Michael Bowling of the University of Alberta and colleagues have designed a program to play a version of poker called heads-up limit Texas Hold ‘em. Unlike simpler games in which players have full knowledge of the state of the game, such as tic-tac-toe and checkers, in poker the players can’t see each other’s cards. Further complexity is introduced by the possibility of drawing additional cards, betting, and bluffing. Through a series of training rounds played against itself, the computer learned from its mistakes and improved its strategy to the point that it would be virtually unbeatable by a human opponent. Besides helping poker professionals to improve their game, the algorithm could have practical applications in fields in which people must make decisions based on limited information, such as airport security and health care.