T. Gerster

Abstract—This project aims to develop a Python-based chess program capable of grasping a sufficient understanding of chess dynamics to be able to beat a random-move” player and potentially challenge low-rated opponents. Given the intricate nature of chess, the objective is to find ways for the program to comprehend, even remotely, the chessboard, its different pieces and their possible relationships. This paper proposes the development of evaluation methods to score chessboard positions. Two different approaches have been considered; a translation of the board state through static scoring (ChessBot Battle) and through the training of a Convolutional Neural Networks (ChessBot : a CNN application). The two resulting different Chess Bots, respectively named Harold and Albert, are then tested through simulations.

This paper was developed concurrently with the “Chess Bot : a CNN application” project within the academic realm of a master’s program. Both papers share a common objective but employ distinct methodologies to attain it, thereby overlapping in certain content areas.