A Study of Race Momentum Based on Machine Learning Methods and Pearson's Correlation Coefficient

In this study, we developed two models to deeply investigate and predict momentum effects in tennis matches. First, a momentum assessment system model was constructed to categorise momentum effects into past momentum, serving situations, psychological and physical factors, and to identify the effects through data processing and correlation analysis, defining the momentum function and applying it to the 2023 Wimbledon tournament; second, we carried out hypothesis testing to demonstrate the significant effects of momentum on match progression and player scoring. Additionally, we developed the Match Swing Prediction Model, which uses a machine learning approach to predict lead and swing points in a match, providing practical advice to players. By exploring the model's extension to women's tennis and table tennis matches, we found that the momentum effect for women's matches was slightly less significant than for men's matches, but that the model performed well in predicting women's matches. These findings provide useful insights into understanding and predicting momentum changes in tennis matches and provide practical references for players and coaches.