Benchmarks and sources
Never invent precision. Here is every number ANIMA HYBRID compares you against, where it came from, and what is estimated.
The rules we hold ourselves to
- No benchmark is shown without the sample size behind it.
- No percentile is shown more precisely than a whole percent, and never better than "top 1%".
- We only say "top X%" while you are in the faster half of your bracket. Past that we say what you actually beat, because "top 77%" would be misleading.
- Anything extrapolated rather than measured is labelled estimated everywhere it appears.
- A bracket with fewer than 25 community results publishes no community number at all.
How distributions are stored
Every public source publishes quantiles — a median, a top 10%, a bottom 25% — and never raw results. So ANIMA HYBRID stores a distribution as seven percentile anchors and interpolates honestly between them, rather than assuming a curve. That means published figures are used exactly as published, and a better dataset can replace them without changing anything else.
How community data takes over
Every race added with sharing on increments an anonymous histogram for that bracket. At 400 community results a bracket is weighted half community and half published; at 1,600 it is 80% community. Delete your data and your contribution is subtracted from those counters. The counters themselves hold no identifier of any kind.
Sources
| Tag | What it gives us | Stated sample |
|---|---|---|
| S1 | Station times by percentile (men and women, all eight) and the mean time for each of the eight kilometres | 395,452 solo results |
| S2 | Finish time by age group, the age decline curve, the women-vs-men gap, and our reference bracket (Open men 30–34: p25 1:12:20, median 1:20:29, p75 1:29:17) | 4,309 men, 943 women |
| S3 | Division means, which set the Pro, Doubles and Pro Doubles factors | 12,479 results |
| S4 | The doubles distribution and the doubles-vs-solo factor | 425,242 team finishes |
| S5 | Relay benchmarks — no sample size is published, so every relay number is estimated | not published |
| S6 | Transition totals by finish bracket (sub-1:10 → 3:41; 1:30+ → 6:58) | 825 athletes, one venue |
| S7 | Which stations spread the field most, which is how the report ranks what is worth training | 700,000+ results |
| S8 | Brandt, Ebel, Lebahn & Schmidt (2025), Frontiers in Physiology — aerobic capacity was the strongest predictor of finish time (ρ = −0.71). Used only to decide what the plan prioritises, never for a number shown to you. | n = 11 |
| S9 | Rappelt et al. (2026), Frontiers in Physiology — seven seasons of Pro and Elite development, used as a sanity check on the fast end of our curve. | 39,696 Pro, 473 Elite |
| H1 | 5 km times by age and ability, with an explicit percentile per level | not published |
| H2 | 1 km row median (men 3:22.7, women 4:02.3) — the spread is fitted from the running percentiles, so it is estimated | not published |
| H3 | Mangine, Cebulla & Feito (2018), Sports Medicine – Open — normative benchmark workout scores (Fran: men 250 ± 106 s, women 331 ± 181 s). Self-reported. | 10,000 profiles |
| H4 | Long-course triathlon finish times by age and sex | 41,000+ finishers |
| H5 | Squat standards by bodyweight, with an explicit percentile per level | 7,039,938 qualifying lifts |
The Anima Performance Rating formulas
The APR does not invent a scale. It puts two published ones — DOTS for strength and VDOT for endurance — onto the same 0–1000 percentile axis.
| Tag | What it gives us | Source |
|---|---|---|
| A1 | DOTS — the open successor to Wilks, adopted by the IPF's national federations. Used exactly as published, clamped to the bodyweight range it is defined over. | OpenPowerlifting |
| A2 | McCulloch / Foster masters age coefficients: exactly 1.000 between 23 and 40, rising either side. Interpolated between the printed anchors. | USAPL / IPF masters |
| A3 | Daniels & Gilbert VDOT — an oxygen-cost curve in velocity and a fractional-utilisation curve in duration. A 20:00 5 km comes out at VDOT 49.8, which is what the published table says. | Oxygen Power (1979) |
| A4 | Cooper 12-minute test: VO₂max = (metres − 504.9) / 44.73 | JAMA 203(3), 1968 |
| A5 | Concept2 pace–power: watts = 2.80 / pace³ | Concept2 |
| A6 | Hagerman rowing oxygen cost: VO₂ (L/min) = 0.01141 × watts + 0.435 | Sports Medicine 1(4), 1984 |
| I1 | Acute:chronic workload ratio. The comfortable band everyone quotes is 0.8–1.3. Contested — Impellizzeri et al. (2020) argue it is statistically fragile — so ANIMA HYBRID treats a flag as a prompt to look at your week, never as a verdict, and prints that caveat next to every one. | Gabbett, BJSM 50, 2016 |
Guidance, not medical advice
The readiness score and the injury-risk flags read patterns in numbers you typed in. ANIMA HYBRID cannot examine you, it never diagnoses anything, and it can be wrong. If something hurts, stop and see a doctor or physiotherapist.
Every Open score is self-reported
Nothing on an Anima Open board or a squad board has been verified. Every entry is typed in by the athlete and badged SELF. The data model already carries a place for camera-counted and sensor-read entries so today's scores stay valid when those arrive — but neither is built, and no row in the app claims either.
What is estimated, and why
- The 5th and 95th percentile of every distribution is fitted from the published quantiles on a log scale. The published values are preserved exactly.
- Station percentiles, run-leg means and transition totals come from different samples, so they do not add up to the finish distribution on their own. All three are scaled by one common factor — typically 3–6% — so the report adds up. A report that does not add up is worse than one that is a few seconds off.
- No public source breaks doubles or relay down station by station, so team station benchmarks are the solo table scaled by the division factor.
- The published women's samples above 50 are 20, 12 and 1 athletes, far too thin to use directly, so the gap above 50 is extrapolated at about 11%.
- Murph is not in the benchmark-workout study, so its anchors are positioned from the Fran and Helen relationship in the same sample.
- Squat standards are graded by bodyweight from published data, but no citable age curve existed, so the age adjustment there is ours.
- The APR's DOTS benchmark curve is the mean of the published strength standards across the whole bodyweight range, not one reference bodyweight. The standards are not bodyweight-neutral even though DOTS is, so picking a single reference would have moved everyone's strength percentile by up to a quarter.
- A VO₂max from a 2 km row chains three published regressions (A5, A6, A3). Each one is published; the chain is not, so the row route is always labelled estimated. A 5 km run is one relationship and is not.
- Enter only one or two of squat, bench and deadlift and the missing ones are read off the published standards at the same ability level as the ones you gave — not a flat ratio, because bench and squat do not scale together. The total is then marked estimated.
- The readiness weights (resting heart rate 30%, sleep 30%, soreness 25%, yesterday's load 15%) are ours. There is no validated consumer readiness formula and we do not pretend otherwise; they follow the direction of the published literature rather than any single paper.
The biggest weakness, stated plainly
Sources S1 to S4, S6 and S7 all come from one analytics site. They publish their sample sizes, they are internally consistent, and they agree with the peer-reviewed elite data in S9 — but they are not primary race data and have not been independently audited. S2 is three European events, so times at large North American or Asian events may differ. That is precisely why ANIMA HYBRID blends in community results: within a season of real use, most brackets should rest on ANIMA HYBRID's own data rather than on an aggregator's.
Found a mistake?
Tell us which bracket and which number: support@animahybrid.app. We would rather fix a number than defend it.