Development, application, and limitation of a stochastic Markov model in explaining championship squash performance.

This study reports a stochastic (Markov) model for squash which uses empirical data to transit event states on a shot by shot basis. This offers more information than traditional models with regard to how points were won or lost and the potential for predicting future athletic performance from a priori observation. The predictive capacity of the model, however, is presently restricted because the observed behaviors (shots) and associated outcomes (winners, errors and lets) are statistically variant (p < .25). A player does not produce a consistent athletic response to the same preceding condition when competing against different opponents, although it is unclear at present whether this observation is a function of the particular analysis employed. Nevertheless, the modeling of athletic behavior is a way to search for critical data which underpin competitive sport performance.