Analysis and Comparison of Chess Algorithms

Abstract

In this paper we analyze the results of three different algorithms programmed for playing chess – genetic algorithm, Monte Carlo, and Minimax. The algorithms are implemented in Python through 5 players that play chess against the Stockfish engine, each for 10 games, after which their Elo rating, game evaluation, game status, and time per move are compared. The results show that the algorithms cannot compare to an extensively trained and optimized chess engine such as Stockfish, and only 2 games of 50 total were won by the Minimax algorithm. There were no draws. The genetic algorithm is very fast, with less than a second needed for each move, while the other two are much slower, with times sometimes reaching over a minute. The Minimax algorithm’s speed decreases over time, while the Monte Carlo algorithms’ speeds increase over time.