Performance Monitoring for evaluating risk in anti-doping

Code: 20D03JH

The use of information technology within sport has significantly increased over recent years. The availability of data for longitudinal tracking of athletes over the course of their careers is critical to identify an individual’s performance progress. As such, a key component of this longitudinal monitoring is to be able to differentiate between “normal” increases in performance caused by maturation and training, from an “unnatural” improvement caused by doping. Our initial work demonstrates that it is possible to identify specific performance profiles characteristics for athletes with previous ADRVs, which appear to be consistent across the track and field disciplines we have investigated (100-800m, jumps & throws). However, there is a need to extend the current model and translate it into a useful tool for anti-doping authorities. Therefore, the purpose of this study is to refine and further develop our existing Bayesian modelling approach. Specifically, this project will enable us Page 2/8 to explore confounding factors for modeling performance data, and at the same time explore the effects of these confounders on the probability level needed to flag a suspicious individual, balancing false positives (clean athletes identified as doping) and false negatives (doping athletes which are not flagged). We will therefore develop a “risk” score for suspicious performance profiles. A key aspect of the method will be its impact on decisions made via the traditional Athlete Biological Passport (ABP). Working in conjunction with the AIU, we will retrospectively explore whether performance data adds value to decisions made by expert witnesses in historical passport cases. Using this approach, we will evaluate whether adding performance data to existing ABP data is more effective than current methods for identifying athletes who demonstrate may present adverse, or uncertain passport profiles findings.