Scholarlink

This study uses a deep-learning-driven approach to predict dynamic fatigue in football players, which is crucial for optimizing performance and reducing injury risks. Traditional methods for measuring fatigue often fall short by not capturing its multifaceted nature. This paper aims to define fatigue metrics, build a predictive model for fatigue measurement, and evaluate the model’s effectiveness in supporting tactical decisions and injury prevention. The study uses a 2023/24 PKO BP Ekstraklasa league season dataset, integrating event, tracking, and momentum data. A deep learning model, consisting of embedding layers and fully connected layers with normalization techniques, is used to process the data. The model’s architecture includes embedding layers for categorical features and a deep sequential network with GELU activation functions. The model can pinpoint moments when a player’s fatigue level reaches a critical threshold, aligning with real-life substitution patterns observed in professional football. The study bridges sports science and artificial intelligence, offering practical applications in performance optimization.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dynamic Fatigue Prediction in Football: A Deep Learning-Driven Approach to Optimize Performance

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Paweł Dopierała,
  • Wiktor Leszczyński,
  • Paweł Ła̧czkowski,
  • Tomasz Piłka,
  • Tomasz Górecki

摘要

This study uses a deep-learning-driven approach to predict dynamic fatigue in football players, which is crucial for optimizing performance and reducing injury risks. Traditional methods for measuring fatigue often fall short by not capturing its multifaceted nature. This paper aims to define fatigue metrics, build a predictive model for fatigue measurement, and evaluate the model’s effectiveness in supporting tactical decisions and injury prevention. The study uses a 2023/24 PKO BP Ekstraklasa league season dataset, integrating event, tracking, and momentum data. A deep learning model, consisting of embedding layers and fully connected layers with normalization techniques, is used to process the data. The model’s architecture includes embedding layers for categorical features and a deep sequential network with GELU activation functions. The model can pinpoint moments when a player’s fatigue level reaches a critical threshold, aligning with real-life substitution patterns observed in professional football. The study bridges sports science and artificial intelligence, offering practical applications in performance optimization.