How to tell if your sprint speed is improving
A hand-held stopwatch has roughly 0.2 seconds of error. Six months of serious speed training might buy you 0.15 seconds over 40 yards. Your measurement tool is less precise than the thing you're trying to measure.
That single fact explains why most athletes have no idea whether they're getting faster. It's fixable with a phone and a consistent protocol.
Why hand timing can't detect real improvement
When you time someone by thumb, you're adding your own reaction time twice — once at the start, once at the finish. Human reaction to a visual cue is around 0.2s, and it isn't consistent. Worse, it's biased: timers anticipate the finish, so hand times run systematically fast, and hand times taken by an excited timer run faster still.
Put in perspective:
| Source | Typical error |
|---|---|
| Hand-held stopwatch | ±0.15–0.25s |
| Phone video at 60fps | ±0.017s (one frame) |
| Phone slow-mo at 240fps | ±0.004s |
| Real 8-week improvement, 40yd | 0.05–0.15s |
The gap is the whole story. Hand timing cannot see an 0.08s gain. A 60fps phone can see it about ten times over.
The fix: count frames
The protocol
- Mark a fixed distance. 20m or 40yd — whatever you'll repeat forever. Measure it once with a tape and mark both ends permanently if you can.
- Phone on a tripod or propped object, perpendicular to the track, wide enough to see both marks. Same position every time — cones or a chalk mark for the phone too.
- Record at 60fps minimum, 240fps slow-mo if your phone has it.
- Use a flying start or a fixed static start — and never mix them. A 3-point start and a rolling start are different tests by a large margin. Pick one.
- Full recovery: 3+ minutes between reps. Speed work with short rest becomes conditioning, and conditioning times are slower for reasons that have nothing to do with your speed.
- Record every rep, and note which rep it was. Rep 1 and rep 6 are not comparable — see below.
- Note the conditions: surface (track, turf, grass), shoes, indoor or outdoor, wind. These matter more than most training variables.
To get the time: scrub the video to the frame where you cross the first mark, note the frame number, do the same at the second mark, then divide.
The rep-order trap
Sprint times get slower within a session — almost always. Rep 1 is your fastest, and by rep 5 or 6 you can be several hundredths down purely from fatigue, even with full rest.
So if you compare last month's rep 2 to this month's rep 5, you'll conclude you got slower when you may well have gotten faster. Compare rep 1 to rep 1, or take the average of your first two reps every session and always compare that. Consistency of what you compare matters as much as the measuring.
What counts as real
The rule
With frame counting, your measurement error is about 0.02s, so almost any real change is detectable — the limiting factor becomes your own day-to-day biological variation, which is roughly 1–2% (about 0.03–0.06s on a 3-second sprint).
So: 0.05s+ over 20m, holding across two or three sessions, is real. Single-session PRs are usually a good day, not a new level — which is also why chasing a PR every session is a bad way to train.
Expect roughly 1–3% improvement over a well-run 8–12 week block if you're not already highly trained. On a 3.00s 20m that's 0.03–0.09s. Small numbers — which is exactly why the measurement method has to be better than a thumb.
What good progress looks like on paper
| Week | Rep 1 | Rep 2 | Avg of first two | Read |
|---|---|---|---|---|
| 1 | 3.04 | 3.08 | 3.06 | Baseline (turf, trainers) |
| 4 | 3.03 | 3.05 | 3.04 | −0.02 — inside daily variation |
| 8 | 2.99 | 3.01 | 3.00 | −0.06 and holding. Real. |
| 9 | 2.91 | 2.96 | 2.94 | Ran on track, not turf. Not comparable. |
That last row is the one to internalise. A surface change bought a bigger "improvement" in one week than eight weeks of training did. If you don't log conditions, you will eventually record a shoe change as a training breakthrough — and then wonder why it never happened again.
Mistakes that make the data lie
- Timing yourself. Starting a watch while sprinting is unmeasurable. Video or nothing.
- Mixing surfaces and shoes. Track versus turf can be worth more than a training block. Log it, and only compare like to like.
- Mixing start types. Blocks, 3-point, standing and flying starts all produce different times for identical speed.
- Only keeping PRs. Your best-ever rep tells you about one good day. The average of first-two-reps tells you about your ability.
- Testing tired. A speed test at the end of a hard session is a conditioning test.
- Ignoring wind. Outdoors, a tailwind is free time. Note it or test indoors.
How I do this now
Sprint work was one of the unorthodox, mostly unweighted things I trained as a hockey player, and it's a perfect example of why I built Athlete Craft: the useful inputs here are reps + time, and there is no mainstream workout app that will let you track a set in anything other than reps and weight.
In Athlete Craft you build a preset with the inputs you actually need, type your own exercise name (mine is just "20m sprint"), and log every rep. A View can show the average or lowest time per interval across 30/90/180/360 days, so the trend appears without you maintaining a spreadsheet. Different surfaces are just different exercise names, which keeps the comparisons honest.