O impacto do desempenho no valor de mercado dos jogadores de futebol: uma abordagem em data science

Abstract

This study developed a player valuation system using machine learning techniques, aiming to provide a more objective and accurate approach to estimating the market value of football players. The research was justified by the need to overcome the traditional subjective methods used in sports negotiations and by the economic relevance of the sector, which moves billions of dollars annually. The focus was the analysis and comparison of regression models, specifically Linear Regression, Random Forest, and XGBoost, applied to a public dataset of player transfers. The CRISP-DM methodology guided all stages of the project, from business understanding and data preparation to modeling and evaluation of the results. The analyses included the rigorous application of statistical metrics such as R2, MAE, and RMSE, as well as cross-validation to ensure the robustness and generalization capacity of the models. The results demonstrated the predictive potential of these algorithms in the sports context, with the Linear Regression model emerging as the winner, achieving a coefficient of determination R2 of 0.8634 and a Root Mean Square Error (RMSE) of 0.4774. These metrics provide a solid foundation for decision-making and validate the application of data science in measuring value within the football market.