University researchers' robot playing Texas Hold'Em

The poker robot makes decisions after analysing poker rules

Computer scientists have moved beyond figuring out how to beat computerised chess systems and are now tackling automated Texas Hold'Em programs.

Carnegie Mellon University researchers have created a robot that uses knowledge of game theory, not poker smarts, to beat online Texas Hold'Em programs.

The GS1 poker robot, which makes decisions after analysing poker rules, was created by Tuomas Sandholm, director of Carnegie Mellon's Agent-Mediated Electronic Marketplaces Lab and graduate student Andrew Gilpin. Sandholm says the challenge of developing a poker robot is greater than that of trying to beat a computerised chess program because unlike chess, poker involves making decisions with incomplete information (you know what pieces an opposing chess player has, but don't know the hand of a competing poker player).

An algorithm used to accommodate such uncertainties to play poker might have applications in e-commerce, such as in auctions, says Sandholm, who has done significant amounts of research on e-commerce. He is chairman and chief scientist of CombineNet, a company that helps large organisations save money and time on procurement.

A new version of Sandholm's poker robot, dubbed GS2, will partake in the Computer Poker Competition during the National Conference on Artificial Intelligence July 16-20 in Boston.