Performance Anomaly Detection in Athletics: A Benchmarking System with Visual Analytics

Researchers Blessed Madukoma and Prasenjit Mitra have introduced a novel system that uses machine learning and visual analytics to detect performance anomalies in athletics, potentially serving as a cost-effective complement to traditional anti-doping testing. The system processes a massive dataset of 1.6 million performances from over 19,000 competitions spanning 2010 to 2025, employing eight distinct detection methods—from statistical rules to advanced trajectory analysis. These methods were rigorously validated against publicly confirmed anti-doping violations to measure their effectiveness in identifying sanctioned athletes. The study found that trajectory-based methods, which compare an athlete's performances to their expected career progression, achieve the best balance between detecting violations and limiting false alarms. However, all methods face challenges from incomplete data and the rarity of confirmed violations.

The system is designed as an interactive interface for expert-driven investigation, emphasizing transparency and human judgment to support, rather than replace, established anti-doping processes. This approach addresses the high cost and limitations of biological testing, which costs over $800 per sample and has short detection windows for many prohibited substances. By analyzing routine competition results, the system can identify suspicious performance patterns in athletes who may not undergo regular testing. While the system shows promise, the researchers note that incomplete data and the low prevalence of confirmed violations remain significant hurdles. This work represents a step forward in leveraging AI for fair play in sports, offering a scalable, data-driven tool to enhance integrity in athletics.