A cognitive model for human player bad moves

Abstract

To create a chess AI that selects human-like moves, I created a model of the cognitive process when humans make bad moves by extracting the situations in which humans make bad moves. I checked each game played by intermediate players and extracted the situations where mistakes occurred. I then created multiple classification categories based on the situation of the position and the cause of the mistake, categorized the mistakes, and evaluated the characteristics of each. First, I manually considered about few games and determined the mistake type and feature of positions for each situation. Then the mistake type vector and feature vector for whole mistake situations were generated, analyzed, and classified.