Machine Learning in Tennis Match Analysis: Predicting Score Point Victor and Momentum Shift

In tennis and other competitive sports, momentum plays a pivotal role, where actions like consecutive scoring can enhance a player's momentum and increase his chances of winning the match. Quantifying momentum and predicting match outcomes in real-time are critical for both scientific training and strategic decision-making in the sports industry. However, traditional forecasting approaches often overlook the cumulative nature of time-series data and lack quantitative analysis of the dynamic properties influencing match outcomes. In this study, we compare seven machine learning models across two key tasks: predicting the winner of the next point and identifying the critical moment of momentum transition. We introduce a novel momentum formula to quantify the win/loss trends throughout a match. Our experiments using data from the 2023 Wimbledon Championships, demonstrate up to 70% accuracy in predicting point winners using logistic regression (LR) and up to 97% accuracy in detecting momentum shifts via a light gradient boosting machine (LGBM). These findings underscore the significance of momentum in tennis and offer valuable insights for performance optimization and competitive strategy development.