Title: Weighted ELO rating predictions in tennis
Authors: Vincenzo Candila - University of Salerno (Italy) [presenting]
Luca De Angelis - University of Bologna (Italy)
Giovanni Angelini - University of Bologna (Italy)
Abstract: Several methods are available in literature for estimating the probability of winning in tennis, such as the regression-based, point-based and paired-comparison approaches, for instance. Among these latter, the ELO rating method plays a prominent role. Originally applied to tennis by the data journalists of FiveThirtyEight.com, the ELO rating method estimates the strength of each player on the basis of the last match in order to predict the probability of winning for the upcoming match. Notwithstanding its widely recognized merits in terms of ease of reproducibility and good performances, the ELO rating system does not take into account the number of games won by each player in the last match(es). The aim is to investigate the profitability of a variant of the standard ELO rating method, where also the games of the last match(es) concur to define the rating of each player.