Strava ML Cleanup Removes 3.9 Million eBike and Vehicle Activities from Leaderboards
eBikes and vehicle activities totaling 3.9 million have been removed from leaderboards.
Strava completed a large-scale leaderboard cleanup using machine learning (ML) in January 2026. This effort resulted in the exclusion of 2.3 million eBike activities and 1.6 million vehicle-based activities, for a total of 3.9 million activities from segment leaderboards. Approximately 293,000 athletes who had their KOM/QOM positions taken due to misclassified activities have regained their rightful Top 10 rankings.
This work involved recalculating all riding and running leaderboards across global segments. It began with the removal of vehicle activities in May 2025 and culminated in the deployment of an eBike detection model at the end of January 2026.
How the ML Model Works, Analyzing 57 Characteristics
Strava's detection system extracts and analyzes 57 individual characteristics (features) from each activity. These include:
Part of the article
For more details, please refer to the link.
Article excerpt:
"If someone sprints faster than world champion sprinter Mark Cavendish on flat ground or a steady incline, we know it's impossible with human power." ㅋ
