Previsões de resultados em partidas do campeonato brasileiro de futebol

Silva, Moacyr Alvim Horta Barbosa da
Título da Revista
ISSN da Revista
Título de Volume

Predicting football (soccer) results is a problem that has been explored for decades. The results can be seen from two points of view, predict the score or just to predict the result: win, draw or defeat. When we modeling the number of goals from each team, both points of view can be contemplated, score and result. Since 1950, many approaches have been proposed in order to model the number of goals scored by each team in a match. One of the most explored approaches considers the number of goals scored by each team as a variable following a Poisson distribution. From the first works, a underlying hypothesis was that the number of goals scored by the home team and away team was independent. However, some authors have used approaches that consider correlation in the score of the two teams, either through the use of Bivariate Poisson or the adaptation of the independent model. However, the vast majority of these works were limited to the data about the teams playing the matches and the number of goals scored and concede only. This thesis aims to explore the predictive capacity of different Poisson models proposed in the literature to predict the number of goals scored by each of the teams in a match, in addition to making use of more explanatory variables, such as number of shots, number of shots on target, tackles, all those variables coming from Cartola FC. Each one of the explored models was analyzed from the point of view to correct the true scoreboard of the game, as well as to correct the true result of the match, win, draw or defeat.

Área do Conhecimento