On Learning How To Play

It is believed that chess masters use a pattern--based knowledge to analyze a position, followed by a pattern--based controlled search to verify or correct the analysis. There has been, however, very little attention on how to learn both patterns and pattern--based strategies to play games. This paper describes Pal--2, a system that incrementally learns both patterns and pattern--based strategy rules from traces of games. Pal--2 starts with general purpose chess knowledge and after a few games, it is able to play a simple endgame. Experiments in the King and Rook against King endgame show how Pal--2 is able to learn how to checkmate the opponent. 1 Introduction It is believed that chess players use stored patterns to analyze positions, and verify or correct their analysis through a pattern--based controlled search [Cha77, dG65]. Some attempts have been made to use a pattern--based approach by hand--coding the required patterns in some "simple" endgames (e.g., [Bra80, Bra82]) or "tacti..