Why does GPS pace jump around during interval training?

The short answer

Pace is distance divided by time, and an interval set deliberately makes both of those numbers small — a sprint might last twenty or thirty seconds and cover a couple of hundred metres. Over a distance that short, the ordinary few metres of GPS position noise stops being a rounding error and starts being a meaningful fraction of the total distance the calculation is working with. The same noise is there during a slow, steady walk too; it just gets buried under a much larger distance and a much longer time, so it never shows up as anything worth noticing.

Why a sprint is not just a faster version of a normal pace reading

A GPS receiver reports where it thinks you are, not how fast you are moving — speed and pace are both worked out afterwards, by comparing two positions and the time between them. During a long, steady stretch of walking, small errors in individual points barely matter, because the real distance covered between readings is large compared with the noise. An interval flips that ratio. A sprint covers ground fast, but it also ends fast, which means the total window the pace calculation has to work with — however many GPS points landed inside those twenty or thirty seconds — is short on both axes at once. A wobble of a metre or two in any one of those points now sits next to a real distance that is not all that much bigger than the wobble itself.

This is a different situation from a live pace reading flickering when you stop walking, even though the underlying cause — GPS position noise divided by a small distance — is the same mechanism. Stopping drives the distance toward zero while you are not moving at all. An interval keeps you moving the entire time; it is short, not stationary. The number it produces is not meaningless in the way a pace reading is meaningless while you stand still, but it carries proportionally more noise than a pace figure built from several minutes of walking would.

Rest periods have the opposite problem for a related reason

The rest between sprints is usually spent standing still or walking slowly, which is closer to the stopped case than the sprint is. Distance covered during a rest period is small on purpose, so any GPS drift during that stretch makes up a larger share of it — the same arithmetic that makes a live pace reading unstable the moment you stop walking. A rest interval is not trying to report a meaningful pace in the first place, so this rarely causes confusion in practice, but it is worth knowing that both halves of an interval set — the sprint and the rest — sit on the noisy end of what GPS pace can measure well, for slightly different reasons.

Why longer intervals settle down faster than you might expect

The noise does not scale with the length of the interval — a fixed few metres of position wobble stays roughly the same size whether the interval is fifteen seconds or two minutes. What changes is the real distance it is being compared against, which grows the longer the interval runs. A two-minute effort covers enough ground that the same wobble which dominates a fifteen-second sprint shrinks back down to a small share of the total, the same way it does over a full hike. Interval training does not have to be long to be trustworthy, but very short, very fast bursts are the specific case where a small amount of GPS noise gets to matter the most.

Which number to actually trust

The pace for a single short sprint, read the instant it ends, is the least reliable number an interval set produces, for the same reason a live pace reading is the least reliable figure on an ordinary walk. The total distance and total time for the whole set of sprints and rests — added up across the entire session rather than read off one interval at a time — behaves much better, because it is built from a much larger amount of real movement relative to the same fixed GPS noise. If a single interval's pace looks implausible, that is normal and not a sign that anything was tracked incorrectly; the pace for the session as a whole is the steadier figure to look at afterwards.

Things that trip people up

• Comparing the pace of one short sprint against another and reading the difference as real effort — a lot of that difference can be GPS noise on a small distance, not a change in how fast you actually moved.
• Assuming a faster GPS device or app would fix this — the underlying limit is how much real distance a short interval covers, not the precision of any one receiver.
• Trusting a rest-period pace number at all — a rest is meant to be slow or stationary, which is exactly the situation GPS pace is worst at describing.
• Judging an entire interval session by its single fastest or slowest split — the total distance and time for the session is a steadier measurement than any one short interval inside it.

Where TrailCam fits in

TrailCam can run interval sets with spoken cues — set a sprint and a rest, and it calls out each one aloud so the phone can stay in your pocket, with distance and pace announcements spoken as you go. The GPS track underneath keeps recording the same way through every sprint and every rest, and the distance, moving time, pace and elevation gain reported for the finished session are built from the whole recorded route rather than from any single interval read in isolation. TrailCam is free to download.