Prediction of next-day subjective fatigue in professional soccer players: a new key for periodisation?

Predicting daily fatigue in soccer players can help coaches tailor training loads and optimise performance. This study aimed to identify, through a big data analytics approach, the key variables that predict subjective daily fatigue using cost-effective and easily applicable monitoring tools. Six professional teams from the Italian third division (Serie C) were recruited. Each morning, players self-reported fatigue, sleep quality, muscle soreness, stress, and mood. After each training session or match, the session Rating of Perceived Exertion (sRPE) was multiplied by session duration to compute Training Load (TL). Contextual factors, such as the time interval from the previous and to the next match, were also included. The dataset comprised 30,211 observations collected over an entire competitive season. Four machine learning models were tested to predict next-day fatigue (NDF). All models showed strong performance, with accuracies between 79% and 84%. The previous day’s fatigue rating was the most influential predictor. Mediation analysis indicated that the effect of TL on NDF was partly mediated by stress, sleep quality and muscle soreness. Practically, the model can identify players at increased risk of next-day fatigue using routinely collected wellness and sRPE data, allowing coaches to proactively adjust training volume or intensity.