A Graph Theoretic Model of Human Cognition in Chess

Abstract

Although advances in computing power have greatly improved computer chess playing, human chess players still rival their computer counter-parts. Computer algorithms typically use a strategy of exhaustive search, which is unlikely to be used by human players. We hypothesized that human chess players recognize higher order properties of the game and use these properties to limit their need for exhaustive move searching. We used graph theoretic modeling to quantitatively determine three possible higher order properties. We then conducted an experiment by using the higher order properties to preselect moves for a typical exhaustive search chess engine. We played the enhanced chess engine against its unenhanced version in six games. The enhanced version won all six games, regardless of color played, suggesting that pre-selection of moves based on higher order properties of the game is indeed a viable strategy.