Aprendizado de máquina aplicado à previsão da efetividade de substituições de jogadores no Campeonato Brasileiro de Futebol série A

Abstract

Substitutions of players are determining resources for the outcome of a football match. Due to the relevance and limitation of substitutions in official matches, several studies have been conducted to propose an optimal way to substitute a player. That is the best moment or the best strategy. This work proposes to apply and compare machine learning algorithms to classify the second and third substitution of the visiting team as effective or not, through the creation of two distinct models. As the input data set, we use data from four years of the Brazilian Soccer Championship (2015-2018). Using 30% of the data set to test the models, the results show that it is possible to predict the effectiveness of the second substitution with 78.39% accuracy and the third with 86.93% accuracy.