Learning to Play Chess Selectively by Acquiring Move Patterns

Abstract

Human chess players do not perceive a position as a static entity, but as a collection of potential actions. Moreover, they seem to be able to follow promising moves without considering all the alternatives. This contribution investigates the possibility of incorporating such capabilities into chess programs. We describe a methodology and a language for representing move patterns. A move pattern is a structure consisting of a board pattern and a move that can be applied in that pattern. Move patterns are used for selecting promising branches of a search tree, hence allowing for a narrower and therefore deeper search tree. Move patterns are learned during training games and are stored in a hierarchical structure to enable fast retrieval. The paper describes algorithms for learning, storing, retrieving, and using the move patterns.