Physical performance estimation in practice: A systematic review of advancements in performance prediction and modeling in cycling

Abstract

Physical performance in cycling is commonly evaluated with laboratory-based performance markers. However, these markers are not monitored on a regular basis, mainly due to the high costs of testing equipment, invasive sampling and time-intensive protocols. The use of mathematical modeling offers a promising alternative allowing for consistent performance monitoring, identification of influential variables affecting performance, and facilitation of planning, monitoring, and predictive analysis. Wearable technology, such as physiological and biomechanical sensors, can be integrated with mathematical models to enhance the practicality of performance monitoring and enable real-time feedback and personalized training recommendations. In this systematic review, we attempted to provide an overview of the developments in predicting and modeling of performance in cycling and their respective practical applications. The PRISMA framework yielded 52 studies that met the inclusion criteria. The models were discussed according to their modeling goal: characterizing kinetics, alternatives to the gold-standard, training control, observing training effects, predicting competitive performance and optimizing performance. Field-based models and technological advancements were highlighted as solutions to the limitations of gold-standard testing. Due to the lower accuracies of modeling techniques, the gold-standard laboratory-based methods of testing will not be replaced by mathematical models. However, models do form a more practical alternative for regular monitoring and a powerful tool for training and competition optimization. A modeling technique needs to be individualized to the goal and the person and be as simple as possible to allow regular monitoring. Ideally, the technique would work in the field, uses submaximal exercise intensities and integrates technological advancements such as wearable technology and machine learning to increase the practicality even more.