Modelagem Matemática para Previsão de Jogos de Futebol

Abstract

This study proposes to construct a method to calculate soccer probabilities and compare the results with the UFMG Mathematics Department model. In the presented methodology, we assume that the number of goals scored by the teams in a match are independent and follow a Poisson distribution, where the average reflects the attack factor or defense factor. Applied in 250 games of Series A of the 2018 Brazilian Championship, the methodology studied presented results consistent with the UFMG model. It was possible to predict with good accuracy the champion and obtained satisfactory results for the qualifiers for the Copa Libertadores of America, as well as for the relegated teams. The DeFinetti measurement average was 0.568 and approximately 57% of the results are below 2/3. The proposed methodology was implemented in the software R.