Measuring clutch performance in professional tennis

Abstract

Not all points in tennis are of equal importance, yet current performance evaluation ignores the variable importance of points in a match. This paper introduces ‘clutch averaging’, a general method for evaluating player performance on important points. Clutch averages are statistical summaries that are weighted by the probabilistic importance of points. Using point-level data of men’s and women’s Grand Slam matches,
we compare the discriminatory ability and predictive accuracy of clutch averaging and simple averages on a set of 11 serve and return performance measures. We find that clutch averages generally improved prediction accuracy and discrimination compared to simple averages. This general and easy-to-use method for accounting for importance in summaries of tennis match performance will be a useful tool for understanding match outcomes and the impact of pressure on performance in elite tennis.