A novel methodological framework for analyzing the momentum effect in tennis singles

Abstract

This paper proposes a novel methodological framework for analyzing momentum effects in tennis singles. To statistically substantiate the existence of momentum, we employ Chi-squared independence tests for the contingency table. Assuming momentum is present, we develop a momentum metric based on the entropy weight method. Subsequently, we apply a CUSUM control chart to detect change points within the derived momentum series and define a relative distance measure to quantify the intensity of momentum shifts. Furthermore, we construct a predictive model utilizing a Back Propagation neural network (BP) optimized by a Particle Swarm Optimization (PSO) algorithm. The importance of predictive features is analyzed via SHAP values. An empirical analysis applying this framework to data from the 2023 Wimbledon Men's Singles yields the following key findings: (1) Statistical evidence signifficantly supports the existence of momentum in tennis singles. (2) Incorporating momentum characteristics substantially enhances point outcome prediction performance. (3) The BP+PSO model demonstrates competitive advantages over alternative machine learning algorithms, including Random Forest, Support Vector Machines, and logistic regression. (4) SHAP value analysis identifies an athlete's unforced errors, winning shots, the momentum metric, and the momentum shift intensity as the four most critical features for predicting point outcomes.