Predicting player transfer market value with ensemble machine learning models

Predicting the transfer market value of professional football players is a complex problem in sports economics, given the uncertainty of transfer markets, non-linear associations between performance metrics, and temporal career dynamics. Current machine learning models have demonstrated their effectiveness; however, many lack temporal integration of players’ careers and interpretability for decision-making. This research introduces a Temporal-Aware Heterogeneous Stacking Ensemble (TAHSE) framework that combines gradient boosting techniques (XGBoost, CatBoost, and LightGBM), attention-based Long Short-Term Memory (LSTM) neural networks, and Random Forest regressors in a two-stage stacking ensemble framework, controlled by an adaptive XGBoost meta-learner. The model includes a Position-Specific Feature Engineering module to account for positional differences and a MHTE to capture career dynamics across seasons. We performed experiments on 18,547 players from the top five European leagues between the 2018 and 2024 seasons. The model’s performance was evaluated using cross-validation, resulting in an \(\:{R}^{2}\:\) of 0.9487 and a Mean Absolute Percentage Error (MAPE) of 8.73%, with further bootstrap analysis to confirm the stability of the predictions. The model’s interpretability is improved through a compound explainability framework, involving hierarchical SHAP analysis and counterfactual explanations, which provide global feature importance and individual instance explanations. Although the framework has superior predictive capability, the analysis uses Transfermarkt market value estimates as a surrogate for actual transfer values and is only applicable to major European football leagues, which can potentially limit the transferability of these findings to other markets or competitions. This research indicates that temporal ensemble learning and interpretable AI methods can be used to improve the transparency and data-driven nature of player valuation analysis in football.