Blood sample profile helps to injury forecasting in elite soccer players

Abstract Purpose By analyzing external workloads with machine learning models (ML), it is now possible to predict injuries, but a moderate accuracy. The increment of the prediction ability nowadays mandatory reduce high number false positives. aim this study was investigate if players’ blood sample profiles could increase predictive trained only on training workloads. Method Eighteen elite soccer players competing in Italian league (Serie B) during seasons 2017/2018 and 2018/2019 took part study. Players’ samples parameters (i.e., Hematocrit, Hemoglobin, red cells, ferritin, sideremia) were recorded through two group them into main groups using non-supervised ML algorithm (k-means). Additionally data every or match day GPS device (K-GPS 10 Hz, K-Sport International, Italy), grouping used as predictor for injury risk. goodness tested assess influence profile prediction. Results testosterone, ferritin most important features that allowed analyze response each type player profile. samples’ characteristics permitted personalize decision-making rules based reaching an accuracy 63%. This approach increased about 15% compared take consideration workloads’ features. workload varied accordance physiological demands specific period season. Conclusion Field experts should hence not monitor status players, additional information derived from individuals’ permits have more complete overview well-being. In way, coaches better program maximizing effect minimizing