Bayesian framework for longitudinal profiling of athlete performances

Abstract

Anti-doping remains a central priority in sports to ensure fairness, particularly as the estimated prevalence of doping exceeds the rate of positive tests. This highlights the need to enhance detection strategies to allow more effective screening of doping and more efficient allocation of resources. This study proposes a Bayesian longitudinal implementation of the athlete performance passport as a risk assessment tool, informed by data-driven reference populations and designed to account for both temporal dynamics and data availability. By modeling clean performances, the framework constructs expectations for future results and detects potentially suspicious performances through probabilistic inference. The results demonstrate that the approach effectively identifies atypical performances with good discrimination capability, while emphasizing the importance of temporal information and sample size in modeling athlete performance. This approach shows potential for supporting risk-based testing strategies in anti-doping.