
Modeling and simulating durations of men’s professional tennis matches by resampling match features
Author(s) -
Francesco Lisi,
Matteo Grigoletto
Publication year - 2021
Publication title -
journal of sports analytics
Language(s) - English
Resource type - Journals
eISSN - 2215-0218
pISSN - 2215-020X
DOI - 10.3233/jsa-200455
Subject(s) - resampling , duration (music) , set (abstract data type) , computer science , goodness of fit , data set , statistics , simulation , econometrics , artificial intelligence , machine learning , mathematics , art , literature , programming language
In this paper we analyze the factors impacting on the length of a men’s professional tennis match and propose a model to simulate matches’ durations. Two distinctive features of the model are that i) it considers all kinds of events that impact on the duration of a match and ii) it is based only on publicly available data. Once built, the model allows to analyze the impact of different formats or rule changes on matches’ duration. The model is built and validated using a dataset including 19,961 matches played in the period January 2011 – December 2018. The simulated and observed distributions of the durations are compared with an in-depth goodness-of-fit analysis. This points out that the model provides a good description of the real distribution both in the central part and in the tails. We also show that our model improves similar models present in the literature. Finally, several case studies are analyzed: the effect of abolishing the first service or the advantages or both; the new tie-break format at Wimbledon; and the introduction of fifth set tie-break at Roland Garros.