When interpretable machine learning meets the beautiful game: a predictive analytics approach to soccer player valuation in the transfer market

Purpose The purpose of our research is to ascertain the key drivers of professional soccer player valuation in the transfer market. Design/methodology/approach Drawing on sports economics, finance and management literature, we connect data-driven approaches to player valuation in the context of organizational decision-making. We evaluate the performance of four predictive models using over 800 real-world transfer fee records and extensive features in the “Big Five” leagues from seasons 2017–2018 to 2019–2020. Subsequently, we leverage Shapley Additive Explanations (SHAP) values, an interpretable machine learning (ML) technique, to identify important features and quantify their contributions to transfer fees. Findings A few fundamental human capital factors (e.g. age) and labor market variables (e.g. contract remaining) emerge as the key value drivers, outweighing technical capabilities (e.g. goal-scoring). Sport-general features (e.g. composure and reaction) hold greater predictive power than soccer-specific skills (e.g. dribbling). Originality/value Our research enhances the explainability and transparency of a reasonably accurate player valuation model in two ways. First, we utilize a rich set of interpretable, fine-grained player features. Second and more importantly, SHAP values allow us to deconstruct player valuation and provide economic interpretations of feature importance at both individual and aggregate levels. We also outline the practical implications of adopting interpretable ML in sports organization decision-making.