ESMQ publication on the use of machine learning to predict transfer fees of soccer players

Background

Permanent transfers are those when a player is permanently engaged by the buying club. The transfer fee for players reflects the financial compensation for this movement of under-contract players, in addition to player wages. The study investigates the relevance of a large set of determinants of players’ transfer fees in the big five European football leagues (men), that is, the English Premier League, the French League 1, the German Bundesliga, the Italian Serie A, and the Spanish La Liga, before and during the pandemic. The determinants include variables of player background characteristics, player performance, selling- and buying-club characteristics, as well as time.

Design and findings

A fully flexible random forest estimator as well as generalized and quantile additive models are used to analyze smooth (non-linear) effects across different quantiles of scraped data (including remaining contract duration) from transfermarkt.de (n = 3,512). While the models are trained with a randomly drawn subsample of before-COVID-19 transfers, the authors compare the prediction accuracy for two subsets of test data, that is, before and during COVID-19. Since the findings suggest several non-linear predictors of transfer fees, moving beyond linearity is insightful and relevant. Moreover, the models trained with before-COVID-19 data significantly underestimate the actual transfer fees paid during COVID-19 particularly for high- and medium-priced players, thus questioning any cooling-off effect of the transfer market.

Implications

The study showcases how moving beyond linearity and modeling quantiles can be revealing for both research and practice. The findings offer some practical implications for a football club’s management. By inserting the corresponding values for player and club characteristics into the model, a club manager might derive a player-specific predicted transfer fee. This estimated fee could be contrasted with the market value of that specific player. Moreover, such procedure might be particularly useful in combination with industry-derived solutions to predict team performances based on the decision whether a player is signed or not.

About the study