Poker as a testbed for AI research

For years, games researchers have used chess, checkers and other board games as a testbed for artificial intelligence research. The success of world-championship-caliber programs for these games has resulted in a number of interesting games being overlooked. Specifically, we show that poker can serve as an interesting testbed for machine intelligence research related to decision making problems. Poker is a game of imperfect knowledge, where multiple competing agents must deal with risk management, agent modeling, unreliable information and deception, much like decision-making applications in the real world. The heuristic search and evaluation methods successfully employed in chess are not helpful here. This paper outlines the difficulty of playing strong poker, and describes our first steps towards building a world-class poker-playing program.