Multi-Output Regression for the Prediction of World-Class Performances in Women’s Handball

Abstract

The purpose of this study was to investigate the relationship between workload and in-gametechnical and athletic performance. To achieve this, multi-output regression models were used to predict7 performance indicators based on previous training and game athletic workloads measured by InertialMeasurement Units (IMU) indicators, previous in-game actions annotated by staff members and gamecontextual factors. We compared 4 single-output models (kNN, regression tree, random forest and neuralnetworks), their multi-output counterparts and a dummy baseline (predicting the average performance ofeach player over the last month) in terms of average RMSE (aRMSE) during a chronological evaluationwhere previous trainings and games data are used to train models to predict the next game performances.Overall, the use of multi-output regression models enabled a decrease of the average predictive error(aRMSE = 4.23) in regards to their single-output counterparts (aRMSE = 4.35) while providing a significantdecrease of average computation times (4.75 to 0.82 seconds). Among the 4 multi-output models, only thekNN (aRMSE = 3.852) and random forest (aRMSE = 3.888) performed better than the dummy regressor(aRMSE = 3.944). These results point towards that physical training may have a limited impact on gameperformance.