An Evolutionary Approach for the Tuning of a Chess Evaluation Function Using Population Dynamics

Using the game of chess, this paper proposes an approach for the tuning of evaluation function parameters based on evolutionary algorithms. We introduce an iterative method for population member selection and show how the resulting win, loss, or draw information from competition can be used in conjunction with the statistical analysis of the population to develop evaluation function parameter values. A population of evaluation function candidates are randomly generated and exposed to the proposed learning techniques. An analysis to the success of learning is given and the undeveloped and developed players are examined through competition against a commercial chess program.