The ChessAIThon project (2025-1-ES01-KA220-VET-000354329) is co-funded by the European Union. The views and opinions expressed in this publication are those of the author(s) only and do not necessarily reflect those of the European Union or the Spanish Service for the Internationalisation of Education (SEPIE). Neither the European Union nor the National Agency SEPIE can be held responsible for them.
Table of Contents
Chess has historically been used by Artificial Intelligence (AI) to solve the complex problem of predicting the optimal move. This module guides technical teachers so they know how to explain the essential concepts and procedures of modern AI, applying them to decision-making in chess. The methodology focuses on Deep Learning (DL) and Reinforcement Learning (RL) approaches, which have revolutionized the domain.
It will be understood how architectures based on Convolutional Neural Networks (CNNs) , a key component of DL that allows for the recognition of spatial patterns, are used to evaluate positions and predict promising moves, overcoming the limitation of traditional brute-force algorithms. While brute force (used by engines like Stockfish) is based on exhaustive search and manual function evaluation, the modern approach (inspired by AlphaZero) is based on selective learning and neural network intuition. We will be inspired by the RL architecture of AlphaZero, but adapted to train with human games, which will make it viable in an educational context with limited resources.
Teachers will learn to guide students in modeling the complexity of chess through the transformation of data into advanced numerical representations, such as the 77x8x8 matrix. Finally, we will explore the integration of search algorithms like Monte Carlo Tree Search (MCTS) to provide the AI with strategic depth, necessary to develop a competitive AI within the framework of the ChessAIthon competition.