A Machine Learning and Linear Regression Based Study of Momentum Changes and Impacts

This paper explores the changes and effects of momentum in tennis matches, and introduces indicators such as "leverage", "turning point" and "leverage momentum" to construct a classification model. Three key models were developed: first, Logistic Regression, XGBoost and CatBoost models were used to analyze the effects of key features on momentum; second, hypothesis testing was used to verify the effects of momentum on match performance; and third, a turning point model was developed to reveal the key factors of momentum fluctuation moments. The results show the predictive accuracy of the model and a deep understanding of momentum changes, providing decision support for players and coaches. Through a data science approach, this study quantifies the impact of momentum, reveals its importance in the game, and provides new perspectives for controlling game momentum.