Can Horse Racing AI Beat the Odds? LightGBM Doubles Winning Probability Using 35 Years of JRA Data

This fascinating experiment brilliantly demonstrates the power of machine learning by applying it to the unpredictable world of horse racing. By strictly excluding betting odds and relying solely on pre-race features like jockey and trainer statistics, the author created a highly effective predictive model. The resulting 15.62% win rate for the AI's top picks is an exciting proof-of-concept that showcases how pure data analysis can significantly outperform basic statistical baselines.