Using an Evolutionary Algorithm for the Tuning of a Chess Evaluation Function Based on a Dynamic Boundary Strategy

Abstract

One of the effective ways of optimising the evaluation function of a chess game is by tuning each of its parameters. Evolutionary algorithms have become an appropriate choice as optimisers. In the past works related to this domain, the values of the parameters are within a fixed boundary which means that no matter how the recombination and mutation operators are applied, the value of a given parameter cannot go beyond its corresponding interval. In this paper, we propose a new strategy called "dynamic boundary strategy" where the boundaries of the interval of each parameter are dynamic. A real-coded evolutionary algorithm that incorporates this strategy and uses the polynomial mutation as its main exploitative tool is implemented. The effectiveness of the proposed strategy is tested by competing our program against a popular commercial chess software. Our chess program has shown an autonomous improvement in performance after learning for hundreds of generations