Physiologically Interpretable Models of Large-Scale Human Performance

Physiologically Interpretable Models of Large-Scale Human Performance Data

Project Team: David Clarke (Biomedical Physiology and Kinesiology, SFU), Oliver Schutte (Computer Science, SFU), Ming-Chang Tsai (Canadian Sport Institute Pacific)

Players of field-based team sports, such as soccer and rugby, routinely wear GPS devices during practices and games. When multiplied by each player and by all practices and games over the course of a season, these data qualify as bona fide big data. A major challenge in sports analytics is to develop methods for synthesizing such data into models for predicting and optimizing training and tactics.

Most big data models tend to be empirical and descriptive in nature. A big question in the big data field is how models can be made more broadly predictive and inferential. The answer lies, in part, by using models that incorporate physiological mechanisms when studying human health and performance. Through this research, the team will compare physiologically interpretable models with traditional empirical models in their abilities to derive insights from performance data collected using GPS-based monitoring of elite athletes. Specifically, they will apply an extended version of the critical velocity model, which features parameters representing aerobic and anaerobic fitness, to velocity-duration data collected from elite soccer players. The results of this Next Big Question Fund project will motivate deeper studies that will be attractive to non-profit agencies and companies in the sport and health sectors.