Survival Analysis of Offensive Performance in Football Players from South American Leagues

Bernuy Rey J.D.; Granda Zapata G.P.; Dios Luna J.B. · Proceedings of 8th International Conference on Systems Engineering - Cybersecurity and AI: Building a reliable digital future, CIIS 2025 (2025) · DOI: 10.1145/3771678.3771688

Resumen

Survival analysis consists of a set of techniques aimed at studying the time until the occurrence of a given event. Its application in the sports field is valuable for identifying patterns in player performance evolution and supporting strategic decision-making related to talent management. In this study, a dataset was constructed containing performance statistics of football players from the first divisions of the Argentine, Brazilian, Chilean, and Peruvian leagues between the 2018 and 2024 seasons. An exploratory data analysis was performed, followed by variable selection using Lasso penalization, and the application of different survival analysis models, including Cox Proportional Hazards, Weibull AFT, Random Survival Forest, and DeepSurv. As a result, 96 key variables were selected for survival prediction, and the DeepSurv model achieved the best performance, reaching a concordance index of 0.91 and the lowest integrated brier score, outperforming both traditional and ensemble models in the evaluated metrics.

Instituciones: Universidad de Lima