Study on Momentum Assessment in Tennis Matches Based on Dynamic Weights and Machine Learning

Qinghua Liu; Boshen Zhu; Jiran Wang; Tongtong Sun

doi:10.62051/06nbe423

Authors

Qinghua Liu
Boshen Zhu
Jiran Wang
Tongtong Sun

DOI:

https://doi.org/10.62051/06nbe423

Keywords:

Momentum evaluation Model; XGBoost Model; Machine Learning; SHAP model; decision tree.

Abstract

In recent years, with the popularization and technological progress of tennis, momentum has been an important indicator of changes in the players' state. Traditional evaluation methods cannot fully reflect their dynamic changes and multi-factor effects, so it is essential to establish a more accurate momentum evaluation model. In this study, seven key indicators, such as serve advantage, breakpoints, etc., were selected to measure momentum fluctuations in matches, and weight was assigned to each indicator to reflect their importance in the overall momentum. Among them, for serving advantage, the dynamic weight Wdy-serve is newly introduced in this study. Moreover, the momentum fluctuation of players in the tiebreaker between Carlos Alcaras and Novak Djokovic is further analyzed, and its effectiveness is verified using the match results. Further, to evaluate index influence on momentum, this study conducted an XGBoost and decision tree algorithm performance, based on analyzing the interpretability displayed shapes. Finally, the XGBoost model with better performance is selected for deep training and iteration in this study. To better understand the influence of each momentum index on the prediction results, this study uses the SHAP model to analyze the correlation of the training set data. Among them, serve advantage and breakpoints are the two most influential indicators, and they contribute the most to the prediction of momentum fluctuations. The dynamic weight Wdy-serve is particularly helpful in capturing the fluctuations in a player's serve performance during a match. The analysis results of this study show that the change of momentum is not only closely related to the technical performance of the players but also significantly related to psychological factors.

Downloads

Download data is not yet available.

References

[1] Briki W. Rethinking the relationship between momentum and sports performance: Toward an integrative perspective [J]. Psychology of Sport and Exercise, 2017, 30: 38-44.

[2] Dietl H M N C. Momentum in tennis: controlling the match[J]. International Journal of Sport Psychology, 2017, 48(4).

[3] Jayapradha J, Abdulsahib G M, Khalaf O I, Prakash M, Uddin M, Abdelhaq M, Alsaqour R. Cluster-based anonymity model and algorithm for 1:1 dataset with a single sensitive attribute using machine learning technique[J]. Egyptian Informatics Journal, 2024, 100485.

[4] Crespo M, Gallego M R ,Filipcic A .Determining the tactical and technical level of competitive tennis players using a competency model: a systematic review[J].Frontiers in Sports and Active Living,2024,61406846-1406846.

[5] Bret J J ,Gorce P .Kinematics characteristics of key point of interest during tennis serve among tennis players: a systematic review and meta-analysis[J].Frontiers in Sports and Active Living,2024,61432030-1432030.

[6] Song Y ,Lin H,Wang B, et al.Factors and Measures Affecting Players' Psychology in Tennis[J].Journal of Social Science Humanities and Literature,2024,7(3):

[7] Sampaio T ,Oliveira P J ,Marinho A D , et al.Applications of Machine Learning to Optimize Tennis Performance: A Systematic Review[J].Applied Sciences,2024,14(13):5517-5517.

[8] Gao X, Yu J, Hu Z, Hu Y. Feature Sorting Algorithm Based on XGBoost and MIC Combination Model[J]. International Journal of Advanced Network, Monitoring and Controls, 2021, 4: 79-88.

[9] Zhang H, Qiu D, Wu R, Deng Y, Ji D, Li T. Novel framework for image attribute annotation with gene selection XGBoost algorithm and relative attribute model[J]. Applied Soft Computing Journal, 2019: 57-79.

[10] Lai J, Liu C, Lin B, Miao D. An improved decision tree algorithm based on hierarchical neighborhood dependence[J]. Soft Computing, 2024, 1-11.

[11] Wang Z, Pang J, Tao P, Ji Z, Chen J, Meng L, Ming D. Locomotion transition prediction at Anticipatory Locomotor Adjustment phase with SHAP feature selection[J]. Biomedical Signal Processing and Control, 2024, 106105.

[12] Ergün S. Explaining XGBoost predictions with SHAP value: a comprehensive guide to interpreting decision tree-based models[J]. New Trends in Computer Sciences, 2023, 1(1): 19-31.

[13] Moss B, O’Donoghue P. Momentum in US Open men’s singles tennis[J]. International Journal of Performance Analysis in Sport, 2015, 15(3): 884-896.

[14] Chen T, Guestrin C. Xgboost: A scalable tree boosting system[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2016: 785-794.

[15] Li Z ,Yin Y .A Quantitative Assessment Model for Momentum in Tennis Based on Exponential Decay and EWM-TOPSIS Method[J].Academic Journal of Computing & Information Science,2024,7(5):

[16] Zhu X ,Wang S ,Wu Z .Prediction of the Swings in Plays Based on Momentum of Tennis Matches[J].Frontiers in Sport Research,2024,6(3):

[17] Tang Y, Ma X, Wang S. Interpretable XGBoost-SHAP model predicts the nanoparticles delivery and reveals its interaction with tumor genomic profiles[J]. bioRxiv, 2022.

[18] Mitchell R, Frank E, Holmes G. GPUTreeShap: massively parallel exact calculation of SHAP scores for tree ensembles[J]. PeerJ Computer Science, 2022, e880.