A stochastic analysis of some two-person sports

We consider two-person sports where each rally is initiated by a server, the other player (the receiver) becoming the server when he/she wins a rally. Historically, these sports used a scoring based on the side-out scoring system, in which points are only scored by the server. Recently, however, some federations have switched to the rally-point scoring system in which a point is scored on every rally. As various authors before us, we study how much this change affects the game. Our approach is based on a rally-level analysis of the process through which, besides the well-known probability distribution of the scores, we also obtain the distribution of the number of rallies. This yields a com-prehensive knowledge of the process at hand, and allows for an in-depth comparison of both scoring systems. In particular, our results help to explain why the transition from one scoring system to the other has more important implications than those predicted from game-winning probabilities alone. Some of our findings are quite surprising, and unattainable through Monte Carlo experiments. Our results are of high practical relevance to international federations and local tourna-ment organizers alike, and also open the way to efficient estimation of the rally-winning probabilities, which should have a significant impact on the quality of ranking procedures