Using ML Models to Predict Points in Fantasy Premier League

Abstract

Fantasy Premier League is an ever-growing game, with millions of people playing the game. To outperform the rest, it is imperative for the players to accurately predict the expected points the footballer would earn over the course of the match. However, doing so is not easy as there are several aspects to consider as well as the human bias towards the players’ favourite footballers and teams. This paper attempts to build and compare three machine learning models to accurately predict the number of points that each footballer would earn over the course of the season. For doing so, the Linear Regression, Decision Tree, and Random Forest algorithms have been leveraged. Features such as fixture difficulty, form of the two teams, creativity, and threat of the footballer have been considered. This would help the players of this game to make more informed decisions while making their respective teams.