Automated Player Selection for Sports Team using Competitive Neural Networks

Abstract

The use of data analytics to constitute a winning team for the least cost has become the standard modus operandi in club leagues, beginning from Sabermetrics for the game of basketball. Our motivation is to implement this phenomenon in other sports as well, and for the purpose of this work we present a model for football, for which to the best of our knowledge, previous work does not exist. The main objective is to pick the best possible squad from an available pool of players. This will help decide which team of 11 football players is best to play against a particular opponent, perform prediction of future matches and helps team management in preparing the team for the future. We argue in favour of a semi-supervised learning approach in order to quantify and predict player performance from team data with mutual influence among players, and report win accuracies of around 60 %.