A Study of Tennis Momentum Analysis Based on Random Forest and Logistic Regression Models

The fluctuations exhibited in sports games exhibit a dynamic unpredictability that is often attributed to the elusive concept of "momentum". In this context, the research in this paper attempts to analyze player performance and game fluctuations through a versatile pace-of-play model. Based on real match data, this paper first uses a random forest classifier to capture the importance of player features and constructs a time series model to capture the matching process. Then, run tests and Monte Carlo simulations are used to evaluate the randomness of match fluctuations and player victories to prove whether the campaigns work. Furthermore, this paper defines game turning points and momentum turning points, and through logistic regression analysis, we validate the relative accuracy of the metrics to provide coaches and players with a more accurate basis for decision making. Finally, we verified the predictive accuracy of the assessment model using a variety of model assessment metrics and conducted a sensitivity analysis to further demonstrate the utility and generalizability of the model for future games.