How to auto-detect your pace from a GPX

Load a GPX from a race you have run. Auto-Detect estimates your flat pace, how much you slow down, your strength on climbs and descents, and what night and altitude cost you. It also says how confident it is.

Why fit a profile from a race

A race recording captures you on one course. Its average pace mixes your flat pace with the climbs, the descents, the fatigue, the surface and, on a long race, the dark. Applying that average to another course imports the shape of the old one. Auto-Detect tries to separate the parts, so a different route can be priced from your own response rather than from the old course.

That is also why the file needs variety. A flat run says almost nothing about how you climb or descend; a short run says little about fatigue; a daytime race cannot say what night costs you. The fit reports those gaps instead of inventing values.

What you need

You'll find Auto-Detect in the planner's Profile step, in the Auto-Detect card. The file is parsed in your browser and is not saved. The estimates are shown for review because a fit from one file is an estimate, not a measurement; they are applied only when you choose Apply to Profile. Auto-Detect is part of the free tier.

What it estimates

Estimate What it means
Flat pace Your pace on level ground early in the race, before fatigue builds.
Fatigue How much you slow down per hour, and whether the slowdown flattens at a ceiling.
Climb and descent scalers How your climbing and descending compare with the model. 1.00 matches the model; below 1.00 means you are stronger on that terrain.
Night penalty How much slower you are in the dark, if the recording has night sections.
Altitude penalty How much slower you are above the threshold, if the recording has enough high sections.
Confidence High, medium or low, with an explanation of what limited the fit.

How it works

Auto-Detect samples the recording, removes stops and obvious outliers, and compares your pace on flat, climbing and descending ground. Stops are removed because a long aid-station break is not running pace; obvious outliers are removed because a GPS spike is not a movement signal. It fits the fatigue shape, estimates the penalties and scalers, then calibrates the base pace so the model reproduces the time you actually moved. The result includes a confidence report and notes what the recording did not contain.

An example of the fit

To show what the fit recovers, this example uses a 50 km route with a known profile. A recording was generated from that profile, run through Auto-Detect, and compared with the inputs. The fit recovered the flat pace within about 4% and was reported as high confidence. The other values are estimates, not a recovery guarantee.

Example recording; values are rounded.
Value Input profile Auto-Detect
Flat pace 7:30/km 7:49/km
Fatigue 5.0%/h, 40% ceiling 3.4%/h, no ceiling
Climb scaler 1.00 0.98
Descent scaler 1.00 0.99
Altitude penalty 10% 6%
Night penalty 10% 10% (default; no night sections)
Stopped time 30 min 29 min
Confidence — high

This example shows what the fit recovers from a recording with a known answer. The fit is a statistical estimate; a different race can give a different result.

What it cannot tell you

A short or flat recording cannot estimate fatigue or terrain strength well. A race without night or altitude cannot estimate those penalties; they stay at their defaults. It is not a fitness test or medical advice. The confidence level is about the fit, not about you as a runner. Auto-Detect is a fit to one recording, so use it as a starting point and adjust the values if you know they are off.

Try it with a race you have already run.

Start from a GPX or race

Cite this page: “How to auto-detect your pace from a GPX”, UltraTrailPlan, https://ultratrailplan.com/guides/auto-detect-pace-from-gpx/.