Strava Deletes 2.3 Million e-bike Entries

In one of the largest data-cleaning operations since the app launched, Strava used three new machine learning models to scan every major segment leaderboard on the platform. The mission was straightforward. scope: Remove the fake times that have plagued competitive cyclists for years.

The process deleted  2.3 million e-bike activities and removed 1.6 million vehicle activities. With these changes, 293,000 athletes restored to their rightful spots in the top 10 in the Strava rankings.

The system now catches activities recorded on an e-bike but uploaded as normal rides. The cleanup focused on the top 100 activities for every single ride segment leaderboard globally, targeting vehicles, incorrect sport types, and e-bikes misclassified as traditional bike rides.

The technology behind the cleanup is sophisticated. E-bikes leave distinctive digital fingerprints in the data. They go faster uphill but slower on flats and downhills compared to strong road cyclists. They sustain high power output for far longer than any human could reasonably. The speed graph stays unnaturally flat, unaffected by wind or subtle gradients that would slow down a regular rider.

Heart rate data also plays a role. Legitimate KOM attempts almost always include heart rate monitors showing maximum effort. E-bike riders rarely record heart rate, and when they do, the numbers tell a different story. As one user noted: "A legitimate KOM will without a doubt have a higher HR than one obtained with an e-bike."