Using the PageRank Algorithm to Rank Football Players in a Game

Abstract

There are many methods to rank football players based on their performance in a game or series of games. However, since most methods are subjective, this paper proposes the PageRank algorithm as an objective method to rank players in a football team, where players can be considered as the nodes, and the passes made between them as the edges of a graph. To achieve this, we consider weighting functions, which are based on parameters which consider the number and quality of passes as well as the actions of individual players in the game. In this paper, the game chosen for implementing the rankings is the 2018 World Cup Final between France and Croatia. The weighting functions are then combined in multiple ways to create different models, which hare implemented in Python to compute the rankings. The models are compared with the official rankings of players during the game with the help of the Kendall's Tau Correlation Coefficient in order to find the distance between the two ranking vectors. While the results may not be highly accurate for the models tested in this paper, a number of additional factors influencing player performance, which official rankings account for, can be considered through more weighting functions. This would lead to more accurate results, thus making the PageRank algorithm a promising and objective tool for ranking football players in a game.