CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1032
Title: Plus-minus a couple of millions: a model for transfer fee evaluation Authors:  Arne Maes - BNP Paribas Fortis (Belgium) [presenting]
Christophe Ley - University of Luxembourg (Luxembourg)
Dries Goossens - Ghent University (Belgium)
Lars Magnus Hvattum - Molde University College (Norway)
Senthil Murugan - University of Luxemburg (Luxembourg)
Abstract: Soccer players contribute to their respective teams both on and off the field, to the extent that clubs are willing to pay a significant fee to purchase the contractual rights of a player from another club. Tough negotiations usually take place about the size of this fee, but what is a good price? Our models estimate transfer fees for players using detailed information on the player, his remaining contract duration, and the selling and buying club. We believe these models can give practitioners a better estimate of the expected transfer fee a player would command. As such, it can be used to guide negotiations, and it allows for more sophisticated planning of investment decisions. Finally, it may serve as a tool for transfer clearing houses in their battle against fraud, allowing them to focus their attention on those transfers where the fee differs substantially from what was predicted. We improve on the existing literature in three ways. First of all, we include domain knowledge to identify niches of the transfer market where a model-based approach can add most value. Secondly, we combine various data sources better to capture the financial and sportive aspects of each transfer. Lastly, we consider several feature-selection (Baruto, Barutoshap, and Statistical FS) and modelling techniques (Decision Tree, Linear Regression, XGBoost, and RF) in a structural fashion to arrive at the most suitable specified models.