December 19, 2017
4:30 PM
CORE, b-135
Optimal Practice Processes for Performance
Guillaume Roels, INSEAD
Throughout their lifetime, people engage in many activities to learn new skills such as learning a new language, taking an online course, or training for a marathon. Although physical training and cognitive learning have their own idiosyncrasies, they can both be viewed as processes of repeated practice to increase performance. Yet, there exist few guidelines to optimize such processes. Building upon research in physiology and psychology, this paper builds an analytical model of a practice process and optimizes it to maximize performance on a predefined date. We model each practice session as having both a positive impact on performance (by contributing to a stock of ``fitness'') and a negative impact on performance (by contributing to a stock of ``fatigue''), the extent of which is moderated by a reference point. During rest between sessions, fitness and fatigue decay at certain rates. We validate the model using a runner's training data and we demonstrate the optimality of periodization or spacing, which alternates practice sessions of high intensity with sessions of low intensity, and of tapering, which gradually reduces the intensity of practice before a performance assessment. Our model is parameterizable to individualize training or learning processes.