Quantifying biomechanical fatigue for athlete health modeling

A recent priority in professional sports has focused on data analytics to reduce athlete injuries. To model the athlete’s risk of injury, a broad set of indicators is needed to completely understand the scope of the problem. The current state-of-the-art injury models still have low accuracy, and our athlete health modeling framework suggests that this is likely due to a common lack of indicators pertaining to the athlete’s biomechanical movement quality. Many biomechanical tests already exist, but cannot be measured on a regular basis because of compliance issues within team environments (e.g. time, cost, discomfort).

This dissertation adapts two traditional biomechanical tests, which can be covertly collected in a team environment on a regular basis. 1) A model that predicts the athlete’s maximal squat weight (1RM) with barbell sensor data from a single-repetition submaximal squat. 2) A model that automatically estimates important vertical jump indicators for on-field vertical jump testing, using motion tracking micro-sensors already worn on-field by many athletes.

Even though athlete biomechanics is a broad category, our two indicators provide largely independent information at opposite ends of the biomechanics spectrum. The squat 1RM functions as a maximal strength biomechanical indicator and the vertical jump test serves as a maximal speed biomechanical indicator. Overall, the combination of using wearable sport technology with complex modeling methods was essential to minimize team compliance issues and still obtain exceptional accuracy compared to the test’s gold standard.