Performance Metrics in High-Speed Sports: Fundamentals, Examples & Analysis

Sports are in many ways ideal for an expert performance approach. The skills needed are rich and complex, yet displayed in tightly constrained tasks that can be isolated and simplified for experimental control and analysis. Well-defined goals enable objective and precise performance evaluation, and participants are already highly practiced on representative tasks from their domain, and motivated to perform well. Modern technology allows collection of large quantities of rich physiological and behavioral data, even in naturalistic settings. Indeed, the challenge often lies more in interpretation of the results and avoiding information overload, than the technical challenge of obtaining accurate measurements. Performance metrics that can summarize large quantities of data are therefore of both theoretical and applied interest; to the basic researcher they parametrize complex behavior into more elementary and interpretable components, to the practitioner they pinpoint aspects of technique to evaluate and improve. This paper discusses some foundational issues in the development of such metrics for high-speed sports, where human performance can be observed at the limit of physiological and information processing capacity. We first define high-speed sports, make a logical distinction between criterion differentiator and proxy metrics, and discuss identifying expert performance directly in terms of performance parameters. We then illustrate three general problems - and how to deal with them - using concrete examples from the openly available IMSRace data set:(1) Extreme events. How to deal with fat tailed performance distributions, a characteristic of high-speed sport even at the expert level? (2) Expert superiority. How good does one need to be, to be an "expert"? We suggest that a Magnitude Based (physical) rather than Deviation based (statistical) approach is to be preferred in high-speed sport.(3) Dissecting performance. How to find predictors for performance for individuals with vastly different levels of skill? We discuss distribution asymmetries and ceiling effects.The topic is of interest to experimental researchers on high-speed sports expertise, and expertise in related fields. Coaches, performance engineers and other practitioners may also gain insights from a more fundamental approach to performance data and its analysis.