Tennis betting: can statistics beat bookmakers?

We propose a logistic regression model to predict the win probability in a tennis match. The variables included in the model are ATP points and rankings, the players' ages, the home factor and the information derived from bookmaker odds. The model is estimated using data related to 2012 tournaments, and it is then used in an out-of-sample betting experiment where the odds implied by the model are used, following a specific procedure, for betting against bookmakers. The algorithm is applied to all matches of the four Grand Slam Championships 2013, and the whole procedure is evaluated with respect to the global return of the strategy. After 501 matches, the total cumulative return is 15.9\%.