Intelligent combat learning sample auto-generator

In the field of artificial intelligence, research on computer chess games has predominantly focused on defeating human players, with the rapid and efficient generation of learning samples being a key factor in enhancing AI board game performance. This study designed and implemented an intelligent combat learning sample auto-generator, employing a self-play learning mode to autonomously produce learning samples without prior knowledge. A comprehensive sample size metric, S, was introduced, which effectively evaluates the scale and quality of the sample set by quantifying the metric y and the repetition rate x of samples. Experimental results demonstrate a positive correlation between the S value and the win rate of the white side, indicating its validity as an evaluation metric for the quality of learning samples. This method optimizes sample size control, reduces learning costs, and provides an innovative tool for training AI in the game of checkers.