Statistical GGP Game Decomposition.

This paper presents a statistical approach for the decomposition of games in the General Game Playing framework. General game players can drastically decrease game search cost if they hold a decomposed version of the game. Previous works on decomposition rely on syntactical structures, which can be missing from the game description, or on the disjunctive normal form of the rules, which is very costly to compute. We offer an approach to decompose single or multi-player games which can handle the different classes of compound games described in Game Description Language (parallel games, serial games, multiple games). Our method is based on a statistical analysis of relations between actions and fluents. We tested our program on 597 games. Given a timeout of 1 h and few playouts (1k), our method successfully provides an expert-like decomposition for 521 of them. With a 1 min timeout and 5k playouts, it provides a decomposition for 434 of them.