Predicting the outcome of a football match is of primary importance in fixed odds betting markets. Dixon and Coles (1997) have proposed a simple bivariate Poisson mixed effects model generalizing Maher’s (1982), showing that the odds of specific match results can be better predicted by their model. Baio and Blangiardo (2010) pro- pose a Bayesian hierarchical model with the additional feature that the teams are grouped into different classes according to their attack/defense performance, leading to a mixture model. We propose a generalization of the approach above, based on the introduction of a Dirichlet Process (DP) Prior on the team performance, allowing for a random number of groups; the application refers to data from the 2010-2011 Italian championship.
A mixture model for predicting football teams' performance / Polettini, Silvia; Francesco De, Icco. - In: QUADERNI DI STATISTICA. - ISSN 1594-3739. - STAMPA. - 14:(2012), pp. 193-196. (Intervento presentato al convegno Methods and Models for Latent Variables tenutosi a Napoli nel 17-19 Maggio 2012).
A mixture model for predicting football teams' performance
POLETTINI, SILVIA;
2012
Abstract
Predicting the outcome of a football match is of primary importance in fixed odds betting markets. Dixon and Coles (1997) have proposed a simple bivariate Poisson mixed effects model generalizing Maher’s (1982), showing that the odds of specific match results can be better predicted by their model. Baio and Blangiardo (2010) pro- pose a Bayesian hierarchical model with the additional feature that the teams are grouped into different classes according to their attack/defense performance, leading to a mixture model. We propose a generalization of the approach above, based on the introduction of a Dirichlet Process (DP) Prior on the team performance, allowing for a random number of groups; the application refers to data from the 2010-2011 Italian championship.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
