Your watch says you ran a 71-second lap. The stopwatch on the infield says 74. Both readings are honest about the time. Only one is honest about the distance.
Every GPS watch cuts the corner on a track curve, swapping the path you ran for a straight line between two points. A straight line is always shorter than the curve it crosses. That’s not a bug. It’s geometry, and it means your reported interval pace lies in one direction, every time you’re on a bend.
Why Your GPS Watch Pace Is Wrong on Every Track Curve
Picture a track curve like a slice of pizza. The true distance around the crust is the arc. Your watch skips the crust and cuts straight across the slice, a shortcut called a chord, always shorter than the arc it crosses. GPS watches build this chord automatically, logging position every second or so and connecting the dots with straight lines. On a straight, that line matches your real path. On a curve, it never does.
Worked out for an 800m pace of 5.33 meters per second, normal 1-second sampling produces a chord-cutting shortfall of roughly 0.2 meters per lap. Barely a rounding error. Drop to a degraded 5-second sampling rate, common with a weak signal or battery-saving mode, and the same lap comes up about 5 meters short.
Sample fast, and corner-cutting is nearly invisible. Sample slow, and it adds up fast. A short lap always reads as a fast lap.
Sampling geometry alone doesn’t explain the errors real watches produce. That’s next.
The Real Curve Radius: A Number Almost Every Track Article Gets Wrong
Nearly every track geometry article quotes the same curve radius: 36.5 meters. That number comes from World Athletics’ facility manual. It’s also not the radius of the line you actually run.
Here’s the arithmetic, and it’s ours, not a quoted figure. Take the published numbers: two straights of 84.39m each, plus a 36.5m radius on both bends. Add it up: 2 × 84.39 + 2 × π × 36.5 = 398.12 meters. That’s 1.88 meters short of the official 400.00m lane-1 distance.
Solve backward for the radius that actually closes the loop at 400.00 meters, and you get 36.80m, not 36.5m. That 30-centimeter gap matches World Athletics’ own rule: lane 1 is measured 30cm out from the inside kerb, not on the kerb itself. The 36.5m figure describes the wall. Runners run the 36.80m line.
We didn’t find this reconciliation stated anywhere. It’s arithmetic, not a citation. But it’s exact, and it quietly explains a common small error in track-geometry content online.
What the Peer-Reviewed Studies Actually Found
Geometry alone predicts a tiny error. Real watches, tested on real tracks, show a much bigger one, because signal noise and reflection, called multipath, matter more than sampling math.
A 2020 study in JMIR mHealth and uHealth tested eight sport watches on a 400m track: six of eight, 75%, came in under 5% distance error. It used one 26-year-old runner across 36 test runs, a small-sample snapshot, and predates today’s dual-frequency (L1+L5) GPS chips, better at rejecting reflected signal.
A 2024 study in Sensors tested hardware closer to today’s watches: 4,000m (10 laps) on a stadium track across 10 smartwatches. Error ranged from 0.8% (Garmin FR955 Solar, best) to 12.1% (Fitbit Versa 4, worst), with seven of ten under 2%.
A 2014 PLoS One study found curved-path running produced a +2.61% distance overestimate, the opposite direction pure geometry predicts. It used decade-old 5Hz hardware, well before dual-frequency chips, so the mechanism can flip on older gear.
The error reaches downstream metrics too. A 2024 European Journal of Sport Science study had 15 trained runners estimate critical speed, the top aerobic speed you can hold before fatigue, via a track time trial. Garmin’s track mode showed 3.4% error on that estimate versus 2.7% for a footpod power meter; D-prime, your finish-kick reserve, showed 20.6% versus 13.8%. Critical speed instead of FTP is least accurate when set from a GPS-only track test.
Official Lane Distances: The Ground Truth Table
Forget the watch for a moment. Here’s what World Athletics certifies for a standard outdoor track, lane by lane.
| Lane | Official Distance |
|---|---|
| 1 | 400.00 m |
| 2 | 407.67 m |
| 3 | 415.33 m |
| 4 | 423.00 m |
| 5 | 430.66 m |
| 6 | 438.33 m |
| 7 | 446.00 m |
| 8 | 453.66 m |
Each lane out adds roughly 7.6 to 7.7 meters, since the curve’s radius grows with every lane. Run lane 3 instead of lane 1, and you’ve added over 15 meters per lap before your watch even enters the picture.
Same lap. Different lane. Different distance, every time.
Track Mode: What It Fixes and Where It Still Breaks
Most GPS watches include a dedicated Track Run mode. Turn it on, and the watch stops relying on raw GPS points, snapping instead to a modeled oval built from two straights and two same-radius curves.
On a genuine single-radius track, that fixes most of the corner-cutting. On a double-radius track, common at many public and school facilities, the single-radius model breaks. Garmin’s own support documentation and forum reports confirm it: the watch still cuts the corner on part of the bend and overshoots the other.
Coros and Polar publish similar track-mode claims, but we couldn’t find independent testing of their curve-specific accuracy, only vendor statements. Treat those as a starting point, not a guarantee. Our 2026 watch buying guide covers which models handle it best.
Turn on track mode. Just don’t assume it’s flawless on a track built with more than one curve radius.
Footpods, Lap Buttons, and Instant Pace: Ranking What to Trust
If GPS on a curve can’t be fully trusted, what should you actually watch mid-rep?
Your watch’s instant pace is the least reliable number on the screen, even on a straightaway. It’s a rolling average of noisy fixes, lagging true speed by 5 to 7 seconds and swinging 15 to 20 seconds per mile at constant effort. Coming off a curve, it’s still catching up from the fast reading the bend just gave it.
Lap pace, your average for the rep, is more honest, but it still inherits whatever corner-cutting happened during that lap.
A calibrated footpod is the strongest wearable alternative: uncalibrated, about 90% accurate; calibrated, 98 to 99%, ahead of GPS’s typical 97 to 98%, and immune to curve geometry since it measures stride, not satellite position. Accuracy drops if your pace strays far from calibration pace. Stryd, NURVV, and Garmin’s running dynamics differ noticeably in accuracy.
| Method | Typical Accuracy | Main Error Source | Best Use |
|---|---|---|---|
| Standard GPS (1Hz) | 0.8%-12.1% MAPE, device-dependent | Chord-cutting plus multipath | Roads, easy runs |
| GPS Track Mode, single-radius | Fixes most corner-cutting | Fails on double-radius tracks | Standard single-radius tracks |
| Uncalibrated footpod | ~90%+ | Stride-model mismatch | Backup only |
| Calibrated footpod | 98-99% | Pace far from calibration pace | Track intervals, treadmill |
| Lap button at a painted mark | Limited by reaction time | Human timing, ~0.2-0.5s | Track intervals, time trials |
The cheapest fix on a track costs nothing extra. It’s your thumb, not your wrist.
Take Dana, 29, a 5,000m specialist. Her watch showed 71-second 400m reps, right on target. Her coach’s stopwatch clocked every rep at 73 to 74 seconds instead. A field test, not peer-reviewed, puts the pure corner-cutting shortfall over a 5,000m race at roughly 32 meters, about a tenth of a lap, enough to make a GPS 5k time read several seconds per mile fast. Dana switched to lap-button splits. The watch and the track finally agreed.
How a Short Lap Can Quietly Skew Your Training Load
This part is a reasoned inference, not a lab finding. No published study has measured it directly, so hold it loosely.
Most training-load formulas score a workout by pace relative to threshold, squaring that ratio to weight hard efforts more. A shortened GPS distance on a curve inflates pace there, and squaring roughly doubles that segment’s stress-score inflation.
On one rep, that’s noise. Stack six weeks of twice-weekly track sessions, and a fast-looking GPS trace could nudge logged training load upward for reasons that have nothing to do with effort.
A short lap doesn’t just lie about your split. It might be quietly inflating how hard your season looks on paper.
AthleteOS sidesteps this by scoring track work off lap-button times or a calibrated footpod, not raw GPS distance. Because fitness score, fatigue score, and form score run on formula-based math, not raw GPS pace, a track repeat reflects what you actually ran. Worth checking before you set a threshold pace target off a track time trial.
Try AthleteOS free and see your next track session scored off the times you actually ran, not the corners your watch cut.