Unleashing the Power of Momentum: Analyzing Tennis Match Flow through Data Analysis

Abstract

This paper aims to analyze the impact of momentum in the Wimbledon Men's Singles Final, uncovering its influence on the game's trajectory and providing strategic advice. To analyze player performance and game pace in tennis, we identified key variables within the dimensions of Consumption, Serving, Accuracy, and Difference. Initially, we developed a momentum prediction model in tennis using logistic regression, with an accuracy of over 73%. In line with the actual match, our visualizations provide valuable insights into the flow of play and momentum shifts in tennis matches. Secondly, we conducted a randomness analysis using the Run test on two metrics: CW (swings in play) and DPW (runs of success by one player), indicated that neither CW nor DPW occurs randomly. Furthermore, we employed Vector Auto-regressive Models (VAR) and found that momentum significantly affects the states of players and their winning streaks. Thirdly, we created the Momentum Swing Prediction Model using the Random Forest approach, introducing a dependent variable, TP, to capture the swings in momentum during a match. By analyzing the variable, we have identified four key factors: win/loss of the serve, rally length, serve speed, and distance. Moreover, we offer three practical recommendations for players entering a new match. Finally, we tested the Momentum Swings Prediction model on matches that were not included in the 2023 Wimbledon Men's model, and proposed incorporating player strength and injury information into future models to enhance the model's performance in different scenarios. Through testing on the 2022 Wimbledon Women's matches, we conclude that our model exhibits generalization across various types of matches. In conclusion, these findings highlight the significance of momentum and how it can impact the flow of play. By understanding these dynamics, coaches can better prepare players to respond to events that impact the match's trajectory.