Quantifying momentum and influencing factors of tennis players using the XGBoost model
摘要
Momentum can directly or indirectly affect a tennis player’s mentality and the trajectory of the game, thereby changing the outcome of the match. The article provides a clear quantitative description of the concept of momentum in tennis, analyzing the impact of momentum fluctuations on the trajectory of the game and the athlete’s scoring; a secondary indicator system is established, and the weights of each indicator are determined through expert analysis method, CRITIC weighting method, and hierarchical analysis method. The final value of momentum is randomly tested with a random walk model. The results show that momentum is not random but influenced by specific factors, indicating that there is a certain correlation between the fluctuation of momentum and the success of the player. Due to the non-normal distribution of momentum, using the XGBoost model and Shap feature importance analysis can determine the significant influence of factors such as the distance run by athletes and the speed at which they hit the tennis ball. Moreover, the model demonstrated excellent performance in five randomly selected matches, with low error indicators and high fitting degrees, and an R2 coefficient of up to 0.9814, showing the model’s high precision and strong generalization capability.