Previsão probabilística dos resultados da copa do mundo de 2022 usando uma abordagem bayesiana

Abstract

This work aims to build and evaluate a predictive statistical model for football matches. To achieve this, we consider a Bayesian approach in a new model inspired by Lee (1997). The inspiration comes from two points mentioned by the author: the fact that he considers the number of goals scored by both teams as independent variables and with Poisson distribution, and that the average number of goals by teams in a match are explained by the offensive and defensive effects of the teams. In this work, we propose model these effects for the teams participating in the 2022 World Cup, through of players' scores in the FIFA 23 video game. These effects are considered together with the odds of sports betting sites, which are represented by parameters of a prior with Gamma distribution. The Bayesian procedure is applied to obtain of the marginal distributions of the data (number of goals for each team), allowing thus calculating the probabilities associated with the matches.