Market Value Estimation of Football Players: A Metaheuristic Machine Learning Approach Incorporating Performance Metrics, Injuries and Club Stature

Football has witnessed unprecedented monetary growth over the years, enablingclubs to have higher purchasing power, which along with rapidly rising number oftalented players, has given birth to a booming interest in player evaluation. Marketvalue fluctuates depending on player performance and enables clubs to establish aclearer monetary threshold for any player when they aim at attempting a transfer.Crowd/forum-based estimates, formulated through discussions have been used as areference point for long, but machine learning predictive models are on the rise dueto the subjectivity and potential biasedness of the crowd estimates. This paper isaimed at predicting market values of players in Europe’s top 8 Leagues across thespan of their career till date. The novelty of this paper stems from extensive usageof statistic-based performance metrics along with novel injury metrics as features inML models employed which comprised of eXtreme Gradient Boosting (XGBoost),Random Forest Regression and Support Vector Regression (SVR). The modelswere enhanced with 4 different metaheuristic algorithms, Genetic Algorithm (GA),Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), DwarfMongoose Optimization (DMO) for feature selection, with the DMO-XGB modelsyielding MAPE values of below 2%, indicative of its high predictive accuracy.