How WIMSTA calculates
Every body-composition formula was built on a different population with a different reference method, so they disagree by several percentage points on the same person. WIMSTA runs several published formulas at once and shows the range they agree on, along with each formula’s own published error — instead of picking one and presenting it as fact.
Same measurements, several published methods. 18.4% is the lowest of them — and it’s the one you’re usually shown.
Every formula we run
Sources, samples and published error for each formula in the product. Verification status is shown honestly: ✅ checked against the primary source, 🟡 checked against secondary sources.
| Formula | Source | Sample | Validated against | Published error | Limitation | Status |
|---|---|---|---|---|---|---|
| Body fat | ||||||
| Deurenberg 1991 | Br J Nutr 65(2):105–114 Deurenberg, Weststrate & Seidell | n = 1,229, Netherlands | densitometry (underwater weighing) | SEE 4.1 pp · R² 0.79 | Overestimates in obesity (per the authors) and not validated in Asian populations. BMI-based, so it cannot tell muscle from fat. | 🟡 |
| CUN-BAE 2012 | Diabetes Care 35(2):383–388 Gómez-Ambrosi et al. | n = 6,510, Spain (white adults 18–80) | air-displacement plethysmography | SE ≈ 4.7% · r 0.89 | BMI-based, so it shares the muscle blind spot; validated mainly in white adults. Not for under-18s or pregnancy. | ✅ |
| US Navy (Hodgdon–Beckett) 1984 | U.S. Navy NHRC Reports 84-11 / 84-29 Hodgdon & Beckett | U.S. military personnel | hydrostatic weighing | ±3–4 pp · r ≈ 0.90 | Developed on a military population; loses accuracy at extreme builds. Not valid in pregnancy. | 🟡 |
| RFM (Relative Fat Mass) 2018 | Sci Rep 8:10980 Woolcott & Bergman | NHANES, n ≈ 12,000 (US, multi-ethnic) | DXA | more accurate than BMI vs DXA | Needs only waist and height — a strong all-rounder, but still an estimate, not a measurement. | ✅ |
| Covert Bailey 1999 | “Ultimate Fit or Fat” (book) Covert Bailey | — | — | not formally published | From a popular book, not a peer-reviewed paper; the primary validation is hard to source. Needs extra circumferences (wrist/forearm or thigh/calf). | 🟡 |
| Healthy-range zones | ||||||
| Gallagher healthy ranges 2000 | Am J Clin Nutr 72(3):694–701 Gallagher et al. | n ≈ 1,626, multi-ethnic adults | four-compartment model | age- and sex-specific ranges | Source of the healthy-range ZONES, not a body-fat estimate. Ranges depend on age and sex; Asian ranges differ. | 🟡 |
| BMI | ||||||
| WHO BMI cut-points 2000 | WHO Technical Report Series 894 World Health Organization | international expert consultation | population mortality/morbidity risk | bands 18.5 / 25 / 30 / 35 / 40 | Population cut-points, not a personal verdict. Class I/II/III come from TRS 894, not the current fact sheet. Asian risk thresholds are lower. | ✅ |
| Trefethen “new BMI” 2013 | The Economist (letter) / Oxford page Nick Trefethen (Oxford) | — | — | 1.3 · w / h^2.5 | A proposal, not an official standard — gentler on very tall and very short people. Uses the same category cut-points as classic BMI. | ✅ |
| Waist-to-height | ||||||
| Ashwell boundary 2016 | BMJ Open 6:e010159 Ashwell & Gibson | systematic review / meta-analysis | cardiometabolic risk outcomes | boundary 0.5 (“keep waist under half your height”) | A single screening boundary across sex and age; the full scale also flags very low ratios (<0.4). A screening signal, not a diagnosis. | ✅ |
| Max heart rate | ||||||
| Tanaka 2001 | J Am Coll Cardiol 37(1):153–156 Tanaka, Monahan & Seals | meta-analysis, n ≈ 18,712 | measured max heart rate | 208 − 0.7·age · SD ±10 bpm | Population-recommended over 220−age, but any prediction is ±10 bpm for an individual. | 🟡 |
| Gulati 2010 | Circulation 122(2):130–137 Gulati et al. | women's cohort, n ≈ 5,437 | symptom-limited stress test | 206 − 0.88·age | Derived specifically in women, where 220−age fits worst. | 🟡 |
| Fox (220 − age) 1971 | no strict primary derivation Fox, Naughton & Haskell | observational, small | — | 220 − age · SD ±10–12 bpm | The famous one, and the least accurate for an individual — shown with its spread, plus six more formulas in the heart-rate calculator. | 🟡 |
What we don’t use — and why
- The YMCA formula. Its primary source can’t be found: calculator sites cite each other in a circle, and several conflicting variants exist. On lean women it returns physiologically impossible values (below essential fat), and it was the worst performer by a wide margin on NHANES data in our testing. Removed.
- “Metabolic age.” No agreed scientific definition, so we don’t report it.
- Silhouettes / body-type images. Body-fat percentage doesn’t map to appearance: two people at the same percentage look different depending on fat distribution, muscle and frame. Drawing “what 25% looks like” would invent a precision that isn’t there.
- “Ideal weight” as a prescription. Ideal-weight formulas were created for drug dosing, not as health goals. We show them as a historical range, not something to aim for.
How to read the error
Each formula’s halo is its standard error of estimate (SEE) — how far a single reading typically falls from a lab measurement for that method alone. The range between methods is a different thing: it’s how much the formulas disagree with each other. Both are real, and both are usually wider than the gap between “categories.” That’s why the honest read is the direction over weeks — measured the same way each time — not the first decimal.
Limitations
- Every formula is a population estimate, not an individual measurement.
- None is validated across all ethnic groups.
- Not applicable in pregnancy, or for children and teenagers (this is an 18+ site).
- Even the reference methods (DXA, hydrostatic weighing, plethysmography) carry 1–2 points of error themselves.
- An estimate is not a diagnosis.
Updates
Last reviewed August 2026.
- August 2026 — YMCA formula removed; clinically-meaningful weight-change threshold corrected from 5% to 3–5% per the 2013 AHA/ACC/TOS guideline; CUN-BAE added and verified against the primary source.