Genetic Algorithms for Evolving Computer Chess Programs

TL;DR

This paper presents a novel approach using genetic algorithms to evolve both evaluation functions and search mechanisms in computer chess programs, resulting in a program that surpasses previous champions and matches top contenders.

Contribution

It introduces a method for evolving chess program components via genetic algorithms, including learning from human grandmaster games and tactical tests, achieving superior performance.

Findings

Evolved program outperforms two-time world champion

Program matches performance of leading chess engines

Uses coevolution to refine evaluation functions

Abstract

This paper demonstrates the use of genetic algorithms for evolving: 1) a grandmaster-level evaluation function, and 2) a search mechanism for a chess program, the parameter values of which are initialized randomly. The evaluation function of the program is evolved by learning from databases of (human) grandmaster games. At first, the organisms are evolved to mimic the behavior of human grandmasters, and then these organisms are further improved upon by means of coevolution. The search mechanism is evolved by learning from tactical test suites. Our results show that the evolved program outperforms a two-time world computer chess champion and is at par with the other leading computer chess programs.