Recently, the world's largest social fitness platform Strava completedthe largest data-cleanup operation in its history.

Theyre-examined major segment leaderboards using three advanced machine-learning models（ Segment Leaderboards ）. The result was that 2.3 million e-bike activities were removed (mainly targeting e-bike activities that were uploaded as regular bike rides), and 1.6 million vehicleactivities (for examplesuch as car trips uploaded as ride data) were removed, restoring 293,000 athletes to the top 10 of the segment leaderboards.

For a long time, Strava's segment leaderboardshave frequently shown anomalous data, which has frustrated many users.Strava is nowusingartificial intelligenceto improve the accuracy and fairness of the segment leaderboards.

In fact, many domestic sports apps also face the problem of "data cheating," and the methods used to "cheat" take many different forms.

In the realm of fitness apps, which aim for authenticity and health, "cheating" may seem contradictory, but behind it lies a complex web woven from personal psychology, social pressure, and real-world temptations.

In simple terms, the motivations for cheating can be summed up in three points: avoiding pain (or laziness), craving recognition, and gaining benefits.

Many people use "step shakers" simply to complete a self-set or externally imposed "check-in" task—such as a 10,000-steps-a-day goal—without doing real exercise. It's a low-cost "placebo" for the mind.

In familiar social circles like WeChat Sports, some people cheat simply to avoid having the lowest step count or to maintain a basic image of "I'm still active," rather than to genuinely compete.

Ranking high on various leaderboards brings a huge sense of virtual honor and achievement. When real ability can't reach that level, cheating becomes a shortcut to this “instant gratification.” It's like using hacks in a game—the thrill comes from crushing others.

After cultivating an image of a “fitness guru” or “disciplined person” on social media, maintaining that persona with impressive data becomes a pressure. Once you relax, you may resort to cheating to fill the data gap and prevent a “persona collapse.”

Many platforms or companies partner with apps to launch activities like “steps-for-cash/coupons/gifts” and “charitable step donations.”This is a powerful driver unique to domestic apps.Cheating here is directly equivalent to “making money” or “boosting one's charity credits.” Alipay's “Walk-to-Donate” has long been troubled by step shakers.

In some companies or schools, step counts or exercise duration are set as mandatory health-assessment indicators, tied to performance reviews, commendations, or even bonuses. In such cases, cheating is no longer entertainment, but a "countermeasure" to unreasonable demands.

When we see abnormal data on a fitness app, it may not just be a technical glitch; it could well be a wrong turn taken by the person behind it while battling inertia, seeking recognition, or chasing benefits. A platform's anti-cheating technology is essentially confronting the complex weaknesses of human nature.

A relatively basic method: using a 'step shaker' attached to a bicycle or treadmill wheel to trick the phone's sensors into thinking the user is moving.

A more advanced method is to use virtual location software (for example, Fakelocation) to fake GPS tracks and simulate any running route.

An even more sophisticated approach: for example, packet interception and data modification — intercept and alter the communications between the app and its server to directly change distance, membership information, etc. Another method is using modding tools, such as modules for the Xposed framework, to directly modify step counts or other in-app data.

Exploiting platform features and loopholes. For example, data file import: simulating runs using GPX/FIT track files and importing them to backfill the platform's records. Another example is account boosting/group ghost-running: multiple people sharing an account or running on behalf of someone, even paying others to run for them.

Like Strava, most platforms rely on technical detection to prevent "cheating." Platforms use algorithms to identify suspicious data such as abnormal speeds, implausible tracks (e.g., long straight lines), or overly smooth signal waveforms.

For example, through sensor pattern recognition. Take Keep as an example: it checks device gyroscope data and combines features like abnormally steady cadence, no GPS movement, and no heart-rate changes to make a comprehensive determination.

For example, validating the plausibility of GPS data. The system checks whether the track is smooth, whether speed changes are within human limits, whether signal loss is abnormal, whether altitude data shows sudden jumps, and so on.

For example, strengthening communication security. Data transmitted between the app and the server is encrypted and signed; if altered, the server will reject it. The system also monitors whether third‑party packet‑capturing tools are running.

And strictly restricting data import and backfill. For example, for manual uploads or third‑party data import features, strict review rules are applied (e.g., prohibiting duplicate uploads of the same route on the same day).

Also,establish a user-reporting channel and reset the leaderboard eligibility thresholds (such as exercise distance, frequency, etc.).

Although technology keeps improving, anti-cheat still faces challenges. Cheating methods are also “evolving” — for example, more sophisticated physical simulation devices and more realistic trajectory-simulation software — which forces platforms to continually update their algorithms. And if the rules are too strict they can mistakenly penalize genuine extreme-sport data (such as high-speed downhill cycling), so platforms need to find a balance between enforcement and tolerance.

Which platform do you think has the most cheating?