Height predictor methods explained: Khamis-Roche vs mid-parental vs bone age
Compare the three main ways pediatricians and AI tools predict adult height. Accuracy ranges, limitations, and which one to trust.
Every height predictor app is built on one (or more) of three methods. Knowing which one you're looking at — and its real accuracy — is the difference between an estimate you can trust and a number-out-of-a-hat.
1. Mid-parental height (MPH)
The simplest formula. Add both parents' heights, adjust ±6.5 cm for sex.
- Boys: (dad + mom + 13 cm) / 2
- Girls: (dad + mom − 13 cm) / 2
Accuracy: Typically ±8-10 cm. It captures genetics but ignores your current growth status. A 13-year-old who's 6 inches taller than peers will likely overshoot MPH; one who's behind will undershoot.
Use it for: A quick sanity-check estimate of your genetic ceiling.
2. Khamis-Roche method (KR)
Published by Khamis & Roche in Pediatrics (1994). The formula:
Predicted adult height = β₀ + β₁ × current_height + β₂ × current_weight + β₃ × mid_parental_height
The β coefficients are sex- and age-specific (in 6-month increments from age 4 to 17.5), derived from regression on the Fels Longitudinal Study data.
Accuracy: ±5 cm (mean absolute error) when applied within its trained age range. Better than MPH because it incorporates your current trajectory.
Use it for: A clinic-grade estimate. It's the same method many pediatricians use.
Want to try it without doing the math? Our free predictor runs the full Khamis-Roche blend and adds a visual share card.
3. Bone age / Greulich-Pyle method
Requires an X-ray of the left hand and wrist. A radiologist compares it to a reference atlas to determine "skeletal age" — how much growth potential remains in the growth plates.
Accuracy: ±2-4 cm. The gold standard.
Use it for: Clinical decisions (e.g., investigating short stature or precocious puberty). Not realistic for casual use — requires a hospital visit, radiation exposure, and a trained radiologist.
Which should YOU use?
| Method | Accuracy | Effort | Cost | Best for |
|---|---|---|---|---|
| Mid-parental | ±8-10 cm | 10 seconds | Free | Quick sanity check |
| Khamis-Roche | ±5 cm | 30 seconds | Free | Realistic prediction |
| Bone age | ±2-4 cm | Hospital visit | $200-500 | Medical evaluation |
For 95% of people asking "how tall will I be?" — Khamis-Roche is the right tool. It's accurate enough to be useful, fast enough to actually do, and free.
Common pitfalls
1. Trusting predictors that don't ask your weight. If a tool only asks for parent heights, it's just MPH dressed up. Real KR needs your current weight (it correlates with growth velocity).
2. Treating the number as exact. Every method has a standard error. A prediction of "178 cm" really means "178 ± 5 cm, with about 80% confidence." Outliers exist.
3. Predicting too young. Below age 4, accuracy drops sharply. Below age 2, any height predictor is essentially guessing.
4. Ignoring puberty timing. A late bloomer might look "short for age" but end up taller because they grow for longer. KR partially accounts for this; MPH doesn't.
What AI adds
Modern AI predictors (like Tallevate's) use Khamis-Roche as the math engine but layer on:
- Vision-based onboarding: Skip typing your age — the AI estimates it from a selfie.
- Confidence visualization: Show the prediction range, not just one number.
- Personality matching: Match your predicted height to celebrities (Pedri, LeBron, Zendaya) for intuitive context.
- Future-self generation: DALL·E paints what you'll look like at full adult height.
The math doesn't change. The UX does — and that matters for actually getting people to engage with the data instead of bouncing.
Bottom line
If you want one number you can trust: use Khamis-Roche. If you want medical certainty: get a bone-age X-ray. If you want a number plus a shareable visual of your future self: try our AI predictor — it's free and takes 30 seconds.