Short answer: no height prediction is exact, and anyone who tells you theirs is 98% accurate is not telling you what "accurate" means. The best methods available today land within about 4 to 5 cm of your real adult height 9 times out of 10. Simpler methods are wider than that. That is the honest state of the science, and it is still useful, as long as you know what the number can and cannot tell you.
Key Takeaways
- Mid-parental height (the formula doctors use first) is right within about ±8.5 cm for 95% of people. Wide, but a solid starting point.
- Khamis-Roche (parents' heights + your current height and weight + age) tightens that to roughly ±5.3 cm for boys and ±4.3 cm for girls, 90% of the time.
- Bone-age X-rays (Bayley-Pinneau, Tanner-Whitehouse) are the clinical gold standard, but studies still show a typical error of 4 to 6 cm, and they overestimate in late bloomers.
- A prediction is a range, not a point. A good app shows you the range and tightens it as you log real measurements.
- "Accuracy" percentages with no error margin attached are marketing, not measurement.
What "Accurate" Actually Means for a Height Prediction
When a study tests a height predictor, it does not ask "was it right?" It asks: how far off was it, on average, and how far off was it in the worst 10% of cases? Those two numbers are the 50% and 90% error bounds.
So a method with a 90% error bound of 5 cm means: for 9 out of 10 teens, the real adult height ended up within 5 cm of the prediction. For the tenth teen, it was further off than that.
Notice there is no percentage-style "accuracy" in there. Height is a continuous number. A prediction of 180 cm for someone who ends up 179 cm is not "99.4% accurate"; it is 1 cm off. Any app quoting accuracy as a single percentage is either dividing two heights (meaningless) or has never validated its model against real adult heights at all.
Method 1: Mid-Parental Height (Tanner Method)
This is the formula pediatricians reach for first, and it is the baseline TallerTeen uses too (here is our full model).
- Boys: (mother's height + father's height + 13 cm) ÷ 2
- Girls: (mother's height + father's height − 13 cm) ÷ 2
Accuracy: Tanner's original work put the expected span at ±8.5 cm, covering roughly 95% of children. That is a 17 cm window. It sounds huge, and it is, because the formula only knows two facts about you: your parents' heights.
Where it goes wrong:
1. Parents guess their own height. In one study of mid-parental target height, both mothers and fathers over-reported their height, which shifts the whole prediction upward. Measure your parents; do not ask them.
2. Genetics is not an average. You inherit a random shuffle of both parents' height genes, not the midpoint. Siblings from the same parents routinely differ by 10 cm.
3. It ignores you. Your current height, your age, whether you have hit puberty: none of it goes in.
Verdict: the right starting point, the wrong finishing point.
Method 2: Khamis-Roche (Parents + Your Current Numbers)
Published in Pediatrics in 1994, Khamis-Roche is the most-validated method that does not need an X-ray. It takes your age, sex, current height, current weight, and mid-parental height, then applies age-specific coefficients built from the Fels Longitudinal Study, which tracked children into adulthood for decades.
Accuracy (from the original study, ages 4 to 17.5):
| Boys | Girls | |
|---|---|---|
| 50% of predictions within | ±2.2 cm | ±1.7 cm |
| 90% of predictions within | ±5.3 cm | ±4.3 cm |
That is a big improvement over mid-parental height alone, because adding your own height and weight tells the model where you sit on the growth curve right now.
Where it goes wrong:
- It was built on a mostly white, American, mid-20th-century cohort. It is less validated for other populations.
- It assumes an average puberty timeline. Early and late bloomers get pulled toward the middle, which means late bloomers are usually under-predicted by this method.
- Accuracy is lowest at the youngest ages and during the peak of the growth spurt, when height changes fastest.
Verdict: the best no-X-ray method available, and the accuracy most online calculators are quietly borrowing when they say "±5 cm".
Method 3: Bone-Age X-ray (Bayley-Pinneau, Tanner-Whitehouse, RWT)
A doctor X-rays your left hand and wrist, reads how mature the growth plates look, and assigns a bone age. Bone age tells you how much growing is left far better than birthday age does. This is the method used in endocrinology clinics.
Accuracy: better in principle, messier in practice. Across studies, the Bayley-Pinneau method shows a standard deviation of roughly 4 to 6 cm against real adult height. In comparisons of boys followed through puberty, the Tanner-Whitehouse (TW2) method was the most accurate, followed by RWT and then Bayley-Pinneau.
Where it goes wrong:
- Late bloomers. Bayley-Pinneau markedly overestimates adult height in about half of boys whose bone age is delayed by more than 2 years. Ironically, the group most desperate for a prediction gets the least reliable one.
- Reader variation. Two radiologists can read the same X-ray a year apart in bone age.
- Access. It needs a clinic, a referral, and radiation exposure (small, but not zero). Nobody is doing this monthly.
Verdict: the most information-rich method, and still not a crystal ball. If a doctor recommends one, it is worth doing. If you just want to know your range, you do not need one.
Method 4: Apps and Online Calculators
Every calculator on the internet is running one of the three methods above, usually mid-parental or Khamis-Roche. There is no secret fourth formula. What separates a good app from a bad one is not the formula, it is:
1. Does it show a range? A single number with no ± is hiding the uncertainty from you.
2. Does it update? A prediction made once at signup is frozen. One that re-runs every time you log a measurement gets tighter the longer you use it.
3. Does it account for what you can change? Sleep, nutrition, and activity do not change your genetic ceiling, but they decide whether you reach it. A model that ignores them cannot tell the difference between a teen sleeping 6 hours and one sleeping 9.
4. Is it honest about the error? If the marketing says "98.5% accurate," ask: 98.5% of what? Measured against whom? If there is no study, there is no number.
TallerTeen's estimate is built on mid-parental height, adjusted to CDC growth curves for your age and sex, with a lifestyle modifier from the habits you log. We publish the whole model, including its limits. Read it here: How TallerTeen predicts your height potential.
The Comparison Table
| Method | What it needs | Typical accuracy | Best for |
|---|---|---|---|
| Mid-parental (Tanner) | Both parents' heights | ±8.5 cm (95%) | A quick genetic baseline |
| Khamis-Roche | Parents + your age, height, weight | ±5.3 cm boys / ±4.3 cm girls (90%) | Teens without an X-ray |
| Bone age (BP / TW / RWT) | Hand X-ray read by a doctor | SD ≈ 4 to 6 cm; worse in late bloomers | Clinical decisions |
| Doubling height at age 2 | Height at age 2 | Very rough; no published bounds | Toddlers, not teens |
Why No Method Gets Below ±4 cm
Three sources of noise that no formula can remove:
- Puberty timing. A boy who starts puberty at 11 and one who starts at 15 can end up the same height by very different paths. Until it happens, the timing is unknown. This is the single biggest reason predictions miss. See when growth plates close.
- Measurement error. You are about 1 to 2 cm shorter at night than in the morning because your spinal discs compress. Measure inconsistently and the model is learning from noise. Here is how to measure correctly.
- Environment. Nutrition, sleep, illness, and stress during the growth years move the outcome by a few centimeters in either direction, and none of it is knowable in advance.
That last point is the useful one. The 5 cm of uncertainty is not just noise. Part of it is the space where your habits decide the result.
So Should You Trust a Height Prediction?
Yes, as a range. A Khamis-Roche style estimate that says "about 180 cm" really means "somewhere from roughly 175 to 185, most likely around 180." That is genuinely useful: it tells you whether your goal is realistic, and it gives you a baseline to track against.
What it is not is a promise. Treat any single number as the center of a window, then spend your energy on the part you control: sleep, food, posture, and consistency while your plates are still open.
FAQ
Which height predictor is the most accurate?
Without an X-ray, Khamis-Roche (90% of predictions within about 5 cm). With one, a bone-age method read by an experienced clinician, ideally Tanner-Whitehouse.
Can a height prediction be 98% accurate?
Not in any meaningful sense. Height prediction accuracy is measured in centimeters of error, not percentages. No peer-reviewed method claims a percentage-style accuracy figure.
Why did my prediction change?
Because you logged new data. A prediction that updates is a feature, not a bug. Why your predicted height keeps changing explains what each kind of change means.
Are predictions less accurate for late bloomers?
Yes, across every method. Khamis-Roche tends to under-predict late bloomers; Bayley-Pinneau tends to over-predict them. If you are 15 or older with no growth spurt yet, treat your range as wider than usual and check the signs your growth plates are still open.
Sources: Khamis & Roche, Pediatrics 1994; Pediatric Research on target height from parental heights; NIH/PMC study on self-reported parental height; PubMed comparisons of Bayley-Pinneau, RWT and Tanner-Whitehouse methods; CDC Clinical Growth Charts.