Analysing serves in para table tennis: Modelling advantage and clustering points

In racquet sports, the serve is considered an important aspect of performance and tactical match play. This is also the case in Para table tennis, yet the advantage of the server and how it differs across impairment classes has not been established. This study aimed to model serve advantage in Para table tennis, as well as provide a framework for identifying serve-return tactics. A convenience sample of shot-level data from 71 Para table tennis matches completed during the 2021–2024 Paralympic cycle were analysed. The win probability of the server, conditioned on shot turn and shot number, was modelled using Bayesian hierarchical logistic regression. A two-step clustering process was used to identify recurring point characteristics using the bounce locations of the first four shots of a rally. The estimated advantage within a point alternated such that the point win probability was lower for whichever players turn it was to play a shot. The exception was the serve, meaning that the server had a higher point win probability for the first two shots in a rally before the alternating pattern began. There were no clear differences in the serve advantage model between impairment classes. Point cluster frequencies differed by class in the sample of matches analysed. These findings suggest that, in the sample of athletes considered, the magnitude of serve advantage does not differ between impairment classes, despite potential differences in shot location characteristics.