Research on dynamic analysis and prediction model of tennis match based on Bayesian probability and analytic Hierarchy process

TL;DR

This paper develops a Bayesian and AHP-based dynamic analysis and prediction model for tennis matches, integrating performance, momentum, and psychological factors to enhance match outcome predictions.

Contribution

It introduces a multi-faceted prediction framework combining logistic regression, AHP, and trend analysis, with validation across men's and women's tennis data.

Findings

Logistic regression model achieved 67.34% probability for first serve win.

AHP-based momentum score strongly correlates with winning probability.

Psychological factors significantly influence match outcomes.

Abstract

In the 2023 Wimbledon Gentlemen's final, Carlos Alcaraz defeated Novak Djokovic. This study aims to predict athletes' performance through five key aspects: first, a mul-ti-classification model based on logistic regression was established, yielding probabilities of winning and losing the first serve at 0.6734 and 0.3266, respectively, and validated with match 1701. Second, an unsupervised "momentum" evaluation model using AHP analytic hierarchy process showed a strong correlation between "momentum" score and winning rate. Third, a trend analysis model identified psychological factors as significantly impacting re-sults. Fourth, the model's generalization was tested with additional competition data, in-cluding women's tennis matches. Finally, data analysis suggested that coaches should focus on improving mental toughness and serve-receive skills, as these significantly affect mo-mentum.