Bayesian statistics approach to chess engines optimization

TL;DR

This paper introduces a Bayesian statistics-based stochastic optimization method for tuning chess engine parameters, applicable to any scenario with indirect, non-analytical gain/loss functions, and compares it with SPSA.

Contribution

The paper presents a novel Bayesian approach for stochastic optimization, demonstrating its effectiveness in optimizing chess engine parameters and general applicability.

Findings

The Bayesian method outperforms SPSA in certain optimization tasks.

The approach effectively handles non-analytical gain/loss functions.

Experimental results show improved parameter tuning accuracy.

Abstract

We develop a new method for stochastic optimization using the Bayesian statistics approach. More precisely, we optimize parameters of chess engines as those data are available to us, but the method should apply to all situations where we want to optimize a certain gain/loss function which has no analytical form and thus cannot be measured directly but only by comparison of two parameter sets. We also experimentally compare the new method with the famous SPSA method.