The project evaluates how well methods of statistical machine learning apply to predicting an elite-athlete's readiness to return to normal training and competitions after an injury, based on analysing signal data from a wearable sensor. These evaluations report prediction correctness in order to systematically compare helpfulness of different signal representations as features for learning. This  project is part of a larger interdisciplinary collaboration that aims to support physiotherapists in assessing when their elite-athlete patients are recovered after an injury. This addresses the more general purpose of improving human performance in sports and minimising the risk of new injuries. The study focuses on the development of a software system that covers the entire workflow from sensor data collection through its computational analysis to result visualisation as training session maps and recovery trends. The system should provide a predictive second-opinion to a physiotherapist on the athlete's recovery percentage after a given session as well as cumulative trends of these percentages in time along the physiotherapy sessions.