Pace analysis bug

Dafftt · February 7, 2026, 1:58pm

Hello,

I noticed during my runs that the pace recorded by Athletica was incorrect, and as a result, the coach’s summary was wrong.

For example, today I did a 20-minute tempo run, and I strictly maintained the pace for the entire 20 minutes (around 6:30/km). Garmin shows that I never exceeded 6:00/km (which is also how I felt) :

However, Athletica tells me that I spent less time in Z3 than expected :

and here is the analysis of pace:

According to the graph, I spent 4 minutes running at a pace faster than 6:00/km, which is impossible.

Another example is an endurance run where I was almost never faster than 7:00/km:

and the analysis tells me that I ran 3 minutes faster than 6:30/km, which is impossible :

Do you know where this bug might be coming from?
Thank you.

Dafftt · February 7, 2026, 4:08pm

I notice that this problem occurs during all my running sessions, even those I did months ago.

mateotito · February 7, 2026, 4:55pm

The same thing happens to me. In a series of 5,000 meters, my fastest pace was 4:16 and my slowest was 4:44 on Garmin, but the Athletica IA platform shows me completely unrealistic times of 3:55 and 5:14.

Dafftt · February 10, 2026, 2:03pm

During the uphill strength session, I pick up speed a little on the downhill, I’d say up to ~5:30/km. Athletica tells me that I spent 2 MINUTES between 5:00/km and 4:15/km. Even on the Athletica speed graph, you can see that I never exceed 5:10/km. It’s totally crazy.

And the worst part is that it completely skews my Running Pace Profile :

so all my athletica is skewed

Marjaana · February 10, 2026, 3:08pm

Hello @dafft @mateotito This isn’t a “bug” in the sense of wrong data being created, but it is a known issue with how we currently parse .fit files. Garmin (and some other platforms) applies smoothing/filters so brief speed spikes don’t show up as sharply in their graphs. Those spikes can still exist in the raw data- you just don’t see them as clearly because they’ve been processed out.

Right now, we’re displaying the raw signal more directly, which can make these short spikes look exaggerated (like in your examples). We agree it’s not ideal for readability, and we’re actively working on improving the parsing + smoothing so the graphs are more consistent and intuitive.

One helpful lens in the meantime: if it’s ~4 minutes inside a 45–58 minute session, that’s roughly 6–8% of the workout. Still annoying (totally fair), but it doesn’t mean the whole session is “totally crazy” - it’s a brief raw-data artifact showing up more loudly than it should. When it comes to your profile data, the short duration skew doesn’t affect your overall load in any meaningful way.

Appreciate you calling it out: it’s on our fix list.

Dafftt · February 10, 2026, 4:02pm

Thank you for your reply. If I understand correctly, it was my watch that recorded these outliers?

Marjaana · February 10, 2026, 4:13pm

Smart watches use GPS, gyroscopes and accelerometers to detect motion. That signal is never perfectly smooth.
Tiny pace spikes happen because motion sensors are affected by:

Arm swing and cadence changes
Trees, buildings, or turns
Small moment-to-moment position errors

So the raw pace signal always contains brief jumps , even when your effort is steady. Watch brands then apply strong smoothing using those motion sensors to hide these spikes in the display. It is the way different companies handle the signal. Makes sense?