The infamous KOM (King of the Mountain) chasers try to beat the best time on a Strava segment in order to climb to the top of the rankings. Achieving the fastest time on a particular section is a motivation that actually drives many cyclists. This makes it all the more annoying when the crown goes to an e-bike rider who has deliberately or accidentally misdeclared their activity.
The fitness tracking platform Strava has now carried out a global clean-up of its segment leaderboards, removing 2.3 million e-bike activities that had been incorrectly uploaded as normal bike rides. This was revealed by a Strava employee on Reddit.
The comprehensive measure addresses a problem that has long been criticised in the Strava community: anomalous activities that distort the leaderboards. In addition to the e-bike activities, 1.6 million motor vehicle activities were also removed from the leaderboards. This often happens when users forget to stop the recording and drive home by car.
This clean-up enabled 293,000 athletes to regain their rightful places in the top 10 leaderboards. The initiative is based on three new machine learning models that have been specially developed for the detection of e-bikes, vehicles and incorrectly categorised activities.
The clean-up included the top 100 activities on all global segment leaderboards for cycling and thus represents one of the most comprehensive interventions in the data integrity of the platform.
New recognition technology at Strava
At the centre of the clean-up operation is a new machine learning model that has been specially trained to identify activities that were recorded on an e-bike but uploaded as normal bike rides. This technology analyses speed profiles, power data and other parameters to identify motor-assisted rides.
In addition, Strava has implemented two further AI models: one to recognise vehicle activity and another that identifies when an activity uploaded as a run was actually a bike ride.
The latter is intended to prevent cyclists from accidentally or deliberately distorting the run leaderboards. According to the description on Reddit, the development of these recognition technologies began in November 2023 and led to a fundamental redesign of the system.
Scope of the clean-up on Strava
The scale of the clean-up illustrates how widespread the problem of incorrectly categorised activities on Strava was. With 2.3 million e-bike activities and 1.6 million vehicle activities removed, a total of almost 4 million entries were removed from the leaderboards. This represents a significant amount of data that compromised the integrity of the leaderboards.
The cleansing was not limited to the segment leaderboards, but was also extended to challenge leaderboards. This comprehensive data cleansing shows that the company is responding to criticism from the community and considers data integrity to be an important aspect of the platform.
Effects for the users
For the Strava community, this clean-up means a significant improvement in the fairness and accuracy of the leaderboards. Particularly noteworthy is the restoration of the rightful positions for 293,000 athletes in the top 10 leaderboards. This should bring joy to many users who will now see their actual performances properly recognised.
The clean-up addresses a key concern of the community: the integrity of the competitions on the platform. By removing motor-assisted and vehicle activities, the performances of those who are travelling exclusively by muscle power are once again displayed correctly. This should strengthen trust in the platform and increase motivation to participate in segment challenges.
Future developments at Strava
Strava emphasises that the work on data integrity is never really finished and that further improvements are planned. The machine learning models that have now been implemented form a basis on which the company can build to recognise and remove anomalous activity even better.
According to the Strava engineer, the feedback from the community, especially from forums such as Reddit, has helped to uncover and address long-term problems. The company is encouraged that the results of this work are now visible in various areas of the platform.
The continuous improvement of detection technologies and the regular cleansing of data should continue to be an important part of Strava's strategy to maintain the integrity of the platform.
Kristian Bauer
Editor
