********************************************************************* M.Sc. in STOCHASTICS and DATA SCIENCE http://www.master-sds.unito.it/do/home.pl

UNIVERSITA' DEGLI STUDI DI TORINO *********************************************************************

Martedì 9 Maggio 2017 alle ore 11:30 in Aula 12 (Corso Unione Sovietica 218/bis, School of Management and Economics), nell'ambito delle attività previste per il corso di Statistics for Stocahstic Processes

il Prof. Christophe Ley (Ghent University, Belgio)

terrà un seminario dal titolo

An alternative ranking for national soccer teams based on strength parameters

Abstract: The FIFA ranking of national (male) soccer teams suffers from various drawbacks that cause heated debates, all-the-more as this ranking governs the seating in the international tournaments and the associated qualification rounds. In the present paper, we propose five alternative models based on strength parameters that are determined by maximum likelihood estimation. We conduct a comparison of these five statistical models, to which we add the ELO model used for the women's FIFA ranking, on grounds of their predictive performance. Indeed, contrary to regular league rankings, the FIFA ranking is designed to measure the current strength of teams, hence should allow accurate predictions of future matches. Our comparison is based on fifteen English Premier League seasons. The best performing model is then used to build the alternative FIFA ranking. We furthermore show its predictive power by analyzing the 2016 European Championship and comparing our predictions to the bookmaker ratings, usually considered as golden standard.

Tutti gli interessati sono invitati a partecipare.