Strava just executed one of the most ambitious data purification feats in cycling app history, targeting an issue that’s frustrated cyclists for years. Over recent weeks, the company rolled out three advanced machine learning models designed to scan every major segment leaderboard worldwide. Their goal was straightforward but immense: to excise fake times skewed by misclassified rides, especially e-bikes and vehicle activities masquerading as traditional cycling efforts.

The cleanup results are staggering: 2.3 million e-bike activities scrubbed, 1.6 million vehicle rides removed, and nearly 300,000 athletes restored to their rightful rankings. James, a Strava engineer, shared a detailed explanation on Reddit, describing how the system detects rides recorded on e-bikes but uploaded as regular rides. The process focused on the top 100 times of every segment leaderboard globally, ensuring precision by filtering out activities logged on vehicles, incorrect sport types, and miscategorized e-bikes.

This was no mere software update—it's a game-changer for competitive cyclists worldwide. For years, dedicated riders watched their hard-earned segment times eclipsed by impossible performances in tourist-heavy areas flooded with rental e-bikes. One Reddit user from Sedona emphasized the relief: “I live in Sedona, where every segment was dominated by tourists on e-bikes. I haven’t checked lately, but I LOVE that Strava finally tackled this.”

The technology behind this effort is both fascinating and practical. E-bikes display unique digital signatures: they boost uphill speeds but tend to flatten out on flats and descents, differing from strong road cyclists who generally slow on ascents and exploit gravity downhill. Their power outputs sustain unnaturally long periods, and speed graphs reveal unusual flatness, unaffected by wind or gradient variations.

Heart rate data also plays a crucial role. Authentic KOM efforts nearly always include recorded heart rates showing elevated cardiovascular exertion, whereas e-bike data often lacks such markers or shows mismatched values. One cycling enthusiast noted, "A real KOM effort will always display a higher average heart rate than one achieved on an e-bike."

Importantly, this isn’t a criticism of e-bike riders. Strava already features a dedicated e-bike category, acknowledging them as valid cycling experiences. The problem had been misclassified activities polluting leaderboards reserved for traditional cycling.

Community reaction has been overwhelmingly encouraging. A user summed it up well: "Thanks, James, for the update. I love Strava’s features and it’s reassuring knowing the team listens."

Though this cleanup marks a major milestone, James admits there’s more to come: "Data integrity is never truly ‘done.’ We’ll keep refining and improving."

For cyclists who slogged through years chasing personal bests only to see them erased by dubious data, this cleanup is more than a technical fix—it’s recognition that every effort counts. Leaderboards are finally trustworthy again, leveling the playing field for riders everywhere.