Evaluating the effectiveness of different player rating systems in predicting the results of professional snooker matches

This paper is the first to consider different methods for rating professional snooker players based on their past performance. The official World Rankings (based on prize money earned) and a player's win percentage are considered, along with paired comparison approaches in the form of Bradley-Terry and Elo models. The models are assessed through their ability to predict the results of subsequent matches, with relatively little to choose between them. Subsets of matches are then analysed to identify relative strengths and weaknesses of the models and potential improvements. The models tended to under-estimate the performance of ‘new’ players and this is the main limitation of using the World Rankings to predict performance. Accounting for the strength of opposition faced by the highest-ranked players is shown to be relevant; although this is less true for lower-ranked players. Models based on two years of results out-perform those based on a single year but there is some indication that accounting for a recent improvement in form may be beneficial.