Are points in tennis independent and identically distributed? Evidence from a dynamic binary panel data model

This article tests whether points in tennis are independent and identically distributed (iid). We model the probability of winning a point on service and show that points are neither independent nor identically distributed: winning the previous point has a positive effect on winning the current point, and at "important" points it is more difficult for the server to win the point than at less important points. Furthermore, the weaker a player, the stronger are these effects. Deviations from lid are small, however, and hence the lid hypothesis will still provide a good approximation in many cases. The results are based on a large panel of matches played at Wimbledon 1992-1995, in total almost 90,000 points. Our panel data model takes into account the binary character of the dependent variable, uses random effects to capture the unobserved part of a player's quality, and includes dynamic explanatory variables.