Most disputes about a tracking platform come down to a misunderstanding of what it measured rather than to bad faith. Elevation is estimated, position is computed, power is often inferred, and segment placement follows from all three. Each has known failure modes worth recognising. Recognising which one is weakest on a given ride explains most of the arguments that follow.
The six subjects
They are ordered roughly by how often they cause an argument. Flagging and elevation account for the bulk of them; the rest matter once a specific result is in question and somebody wants to know what can actually be established from the file. The last two matter mainly when a result is being defended or challenged in public.
- 01
Flagged activities
Why a ride is marked, who can mark it, what a flag actually changes on a leaderboard, and how an appeal is decided.
The most common dispute
- 02
Elevation that looks wrong
Barometric against satellite altitude, why gain only ever inflates, and why two riders on one road record totals a third apart.
The noisiest channel
- 03
KOM verification
What runs automatically, where the automated checks stop, and what it takes to settle a contested crown between two riders.
Leaderboards
- 04
GPS accuracy
Multipath, drift and signal loss, which conditions produce which error, and which device settings genuinely improve a trace.
Position
- 05
Estimated against measured power
Where a calculated wattage diverges from a power meter, by how much, and which terrain makes the estimate trustworthy.
Power
- 06
Getting your data out
Bulk archive against single-activity files, what each format preserves, and what is lost in the conversion between them.
Exports
A note on what a platform can and cannot know
A tracking service receives a file and evaluates the file. It can check that a recording is internally consistent, that speeds match the declared activity, and that a track follows the segment it claims. It cannot check the world the recording describes, because it was not there. That single boundary explains almost everything else about how leaderboards behave.
Almost every limitation described across these pages follows from that single boundary. Recognising it makes the behaviour of leaderboards, flags and elevation totals considerably less mysterious, and makes it obvious why community reporting still does so much of the work. People supply the local knowledge no automated check has access to.
The three estimates every activity rests on
Position is computed from satellite timing and is accurate to a few metres in the open, far worse among buildings. Altitude comes from air pressure or from the same satellites, and is the weakest channel in either case. Power, unless a meter is fitted, is inferred from the first two.
Each estimate inherits the errors of the one beneath it. A displaced position shortens or lengthens a segment; a drifting altimeter inflates climbing; inflated climbing raises an inferred power figure. By the time a number reaches a leaderboard it may be three estimates deep. Each layer is reasonable on its own and the stack is still wrong.
None of this makes the figures useless. It makes them figures with error bars, which is a different thing, and the practical skill is knowing which channel is weakest in the conditions a particular ride was recorded under. A city-centre sprint and an open moorland climb fail in entirely different ways.
| Channel | Typical source | Weakest when |
|---|---|---|
| Position | Satellite timing | Cities, valleys, dense tree cover |
| Altitude | Barometric or satellite | Weather changes, tunnels, buildings |
| Speed | Derived from position | Any moment position is wrong |
| Power | Meter, or inferred | Flat ground without a meter |
Questions about platform data
Why do two devices report different distances for one ride?
Because each computes distance by summing straight lines between recorded points, and the points differ. A device recording every second follows corners more closely than one dropping points during steady riding, and positional noise adds length that was never ridden. Neither figure is wrong exactly; they are answers to slightly different questions.
Is the platform or the device responsible for a wrong figure?
Usually the device, occasionally the platform's correction of it. Recording faults originate in the hardware and the conditions. Platforms then apply smoothing and terrain lookups that fix some errors and introduce others, which is why the same file can produce two different totals. Uploading one ride to two services is the quickest way to see it happen.
Can a leaderboard position be trusted at all?
On a long climb with many participants, largely yes, because errors are small relative to the effort and many people would notice an anomaly. On a short sprint segment ridden by a handful of riders, considerably less, since a momentary position error is a large share of the time. Participation count is as good a reliability signal as any.
What is the single most common cause of an argument about data?
Elevation gain, by a wide margin. It is the channel most sensitive to sensor noise, the only one that sums in a single direction, and the one where two riders on the same road most reliably record totals that differ by a third. It is also the figure people quote most confidently in conversation.
To put a number through the model yourself, see the instruments.