Method · v3.0
Method for the age–sex–pool adjustment v3.0
How a time swum by a 16-year-old in a 25 m pool is made comparable to one swum by a 24-year-old in a 50 m pool, and what that comparison can't tell you.
What it does
The adjustment places a recorded time on a common scale by accounting for three things the record varies in and the reader cannot hold constant: the swimmer’s age at the time of the race, the sex category as recorded by the supplying federation, and the length of the pool.
The output is an adjusted time with a 95% interval. The interval is the estimate: 51.60 s with no interval and 51.60 s [51.49–51.71] are different claims, and only the second is one Swimetrica publishes.
The size of each component
Fitted on 118 402 swimmer-seasons, 31 federations, 2009–2026. Of the three components, pool length moves a time the most: a short-course (25 m) swim adjusts by between −0.61 s and +0.14 s relative to long course, depending on event and stroke. The age effect runs 0.04–0.38 s per year of age in the 13–18 band and flattens above 20. The sex-category effect is fixed within an event and does not vary across the window. The per-event, per-stroke coefficients are in the data export rather than reproduced here; this range is quoted so a reader can judge the adjustment’s scale before reading a single adjusted time.
What it assumes
- Sex category is the value the supplying federation recorded. Swimetrica reports that category exactly. Where a federation’s categories changed within the window, races are adjusted under the category in force at the time of the race.
- Age effects are smooth within a season. The curve is fitted across seasons; a swimmer’s birthday inside a season does not produce a step.
- Pool length is a fixed effect, applied uniformly across swimmers. Some swimmers really are better in short course than the population correction implies. The model doesn’t currently estimate that, and the interval doesn’t widen to cover it. This is the adjustment’s weakest assumption.
Where it is weak
The interval is calibrated against held-out races (method: how the intervals are calibrated), and calibration is only as good as the population it is checked on. Coverage is thinnest for swimmers with fewer than five races in the window (3 178 of the 41 902 held-out swimmer-seasons), at the extremes of the age range — 13 and under, 24 and over: 6 104 of the 118 402 fitted swimmer-seasons — and in the four federations that supply championship meets only. Those are exactly the cases where a reader is most tempted to read a rank as a fact; the calibration page states the coverage rate for each of them separately rather than one overall figure.
Version history
- v3.0, 2026-08-20 — pool-length effect estimated per event rather than per stroke. Intervals widened for short-course-only swimmers: a median of 0.04 s, largest 0.11 s, across 2 118 affected swimmers. The men’s 100 m backstroke recomputation under this version is recorded in corrections.
- v2.2, 2026-02-11 — age curve refitted after three federations backfilled junior results covering 2019–2023. Adjusted times for swimmers aged 14–17 at the time of the race moved by between 0.02 s and 0.14 s; median change 0.06 s, on 3 411 affected figures.