How accurate is an off-the-shelf GPS for sprints? Tested against a laser and timing gates

The paper: Dennison, L., Duthie, G.M., Ehrmann, F., & Psarakis, M.A. (2025). Concurrent validity and interunit reliability of 25-Hz GNSS units for profiling sprinting performance. Journal of Strength and Conditioning Research, 39(1), 107–114. Read it on the journal site.

The question was simple. Lab-grade tools for measuring a sprint — a laser or timing gates — are accurate but expensive, fiddly, and can only watch one lane at a time. An off-the-shelf 25 Hz GPS unit can sit on every athlete at once. But is it accurate enough to trust?

The study

Thirty athletes — 21 men and 9 women, most of them track and field specialists (100 m bests averaging 11.3 s) — ran 106 maximal 40 m sprints in a normal training session. Every sprint was measured at the same time by three systems: a 25 Hz RaceBox GNSS unit worn on the upper back, a laser (Muscle Lab, 1000 Hz) behind the start line, and dual-beam timing gates (Swift) at 10, 20, 30 and 40 m. The laser and the gates are the two “gold standard” field tools coaches already trust.

To check whether one unit reads the same as another, three RaceBox units were also mounted together on a vehicle for 60 accelerations to 60 m.

The RaceBox is a multiconstellation unit (GPS, GLONASS, Galileo and BeiDou) with a 1000 Hz accelerometer built in — and that accelerometer matters, because it detects the exact instant the athlete first moves.

What they found

The unit matched the gold standards. Split for split, the RaceBox read the same as the laser and the timing gates. Here are the actual numbers from all three, side by side:

Mean ± SD across the 106 sprints of 30 athletes (paper’s Table 1); the ± is the athlete-to-athlete spread, not error. The three systems land on the same split times, to within a hundredth of a second — and the gates can’t measure the start (0–10 m), the full sprint, or top speed, which the GPS and laser can.

And it agreed with the laser more tightly than the laser agreed with the timing gates — the two gold standards disagree with each other more than the GPS disagrees with either.

The unit also turns out a clean velocity profile — here is a single sprint, the raw 25 Hz signal and its smoothed trace:

A single maximal sprint from the 25 Hz GPS — the raw signal (red) and its smoothed velocity trace (dark), the profile you’d actually read. Trace shape is representative, as in the paper’s Figure 1; the accuracy against the laser and gates is in the numbers below, not a curve overlay.

The one interval that reads differently isn’t an error — it’s a different start. The GPS starts its clock at first movement: the accelerometer catches the instant the athlete pushes off. The laser and gates start their clock at the start line. Because they define “go” differently, the 0–10 m interval is the only place the two diverge. From 10 m on they are timing exactly the same thing, and the splits line up.

The GPS starts its clock at first movement — its accelerometer catches the instant of the push. The laser and gates start at the start line. Because they define the start differently, the 0–10 m interval is the only one that differs; from 10 m on they measure the same thing and the splits agree to within 0.7–0.9%. First movement is arguably the more useful start for a coach anyway — it’s when the athlete actually goes, not when they reach a line.

Unit-to-unit, the readings were consistent. Put three RaceBox units on the same runs and they read the same — reliability was excellent from 10 to 60 m and for top speed (ICC 0.91–1.00), and good over the first 10 m (ICC 0.86). That first-interval number came from the vehicle used for the reliability test, which accelerates far more gently than a sprinter and so makes the start harder to detect; off the blocks, with a sharp push, that isn’t the same problem.

Strengths and limitations

Strengths

• A real training session with 106 maximal sprints, three systems recording every run simultaneously — a genuine head-to-head, not a modelled comparison.
• Checked against two independent gold standards (laser and timing gates), which also exposed how much those two disagree with each other.
• The unit detects the start from the accelerometer (first movement) rather than a velocity threshold, which is a cleaner definition of “go” than a modelled start.

Limitations

• There was no 3-D motion-capture “true” reference — the laser and gates are excellent, but the study itself shows they are not perfect.
• The 0–10 m interval can’t be a clean apples-to-apples check, because the GPS and the line-based tools start their clocks on different events — so treat that first split as measuring something slightly different, not as an error bar.
• Interunit reliability was tested on a vehicle, whose gentle acceleration made start detection harder than a sprinter’s would; the 0–10 m reliability figure reflects that, not the unit on an athlete.
• It only tested straight-line running (to 40–60 m). A 100 m is also run straight, so that should carry over — but it doesn’t cover bend running (200 m, 400 m), where the GPS has to track a curved path. That’s a separate question.

Practical takeaways

• An off-the-shelf 25 Hz GPS measures sprint velocity, splits and top speed about as accurately as a laser or timing gates. The lab kit is no longer the only trustworthy option.
• Its real advantage is scale. A laser or gate watches one lane; a unit on every athlete profiles the whole squad in the same run. That is the difference between spot-testing one athlete and monitoring all of them.
• Mind what “start” means. The GPS times from first movement; gates and lasers time from the line — so the first split isn’t wrong, it’s measuring from a different event. Every split after it is directly comparable.
• No measurement is exact — even two gold standards land about 0.03 s apart on a split. So watch trends and changes big enough to be real, not single-run decimals.