"Strava is using "Machine Learning" (ML) models to help clean up logged riding errors, intentional or not. James, an engineer at Strava, explained in the post that the recent crackdowns were threefold. First, they introduced a new ML model, specifically aimed at catching e-bikes. This "Enhanced E-Bike Detection" flags and removes activities logged as a normal ride, but that were clearly recorded with electric assist."
"Strava accomplished this by reprocessing "the top 100 activities on every global ride segment." They claim this move helps confirm that the leaderboards for these segments are free of vehicle and e-bike use and also of "incorrect sport types," such as logging a ride as a run. Lastly, Strava introduced ML that focuses on its run leaderboard, better indicating when a logged "run" is actually completed on a bike."
Strava deployed machine learning models to detect and remove activities that misrepresent electric-assist or vehicle use. The approach targeted three areas: an Enhanced E-Bike Detection model to flag rides recorded with pedal assist, reprocessing the top 100 activities on every global ride segment to eliminate vehicle use and incorrect sport types from leaderboards, and ML tuned to the run leaderboard to identify runs completed on bikes. These actions removed 2.3 million apparent e-bike rides and 1.6 million vehicle activities, restoring 293,000 athletes to rightful top-10 positions. The models aim to correct both intentional cheating and accidental logging errors to improve fairness across segments globally.
Read at Bikerumor
Unable to calculate read time
Collection
[
|
...
]