Optimizing Team Sport Training With Multi-Objective Evolutionary Computation

Abstract

Abstract This research introduces a new novel method for mathematically optimizing team sport training models to enhance two measures of athletic performance using an evolutionary computation based approach. A common training load model, consisting of daily training load prescriptions, was optimized using an evolutionary multi-objective algorithm to produce improvements in the mean match-day running intensity across a competitive season. The optimized training model was then compared to real-world observed training and performance data to assess the potential improvements in performance that could be achieved. The results demonstrated that it is possible to increase and maintain a stable level of match-day running performance across a competitive season whilst adhering to model-based and real-world constraints, using an intelligently optimized training design compared a to standard human design, across multiple performance criteria (BF+0 = 5651, BF+0 = 11803). This work demonstrates the value of evolutionary algorithms to design and optimize team sport training models and provides support staff with an effective decision support system to plan and prescribe optimal strategies to enhance in-season athlete performance.