Predicting the outcome of professional darts tournaments

Abstract

Sports statistics are one of the oldest applications of data analysis techniques in order to visualize and understand the outcome of sports events. One particularly interesting application is the analysis of darts data as darts has become tremendously popular and, hence, professional in recent years. In this paper we analyse a data-set consisting of more than 800 professional darts matches. We propose and test various player specific and match specific performance measures in order to identify variables that allow us to assess the players’ winning probabilities in a (single-set) darts match. We find that in about 80% of the analysed single-set matches the winner could be identified correctly by the fitted generalized linear regression models. Afterwards we show how these results can be used in a negative Poisson binomial model to assess the winning probabilities for a multi-set match. Finally, the results of multi-set matches are used in a tournament framework in order to predict the probabilities of all potential outcomes of a darts tournament. The approach is illustrated by analysing the latest World Darts Championship.