The IQ percentile calculator above converts between three representations of the same position on a normal distribution: a score on an IQ-style scale, its z-score, and its percentile. It works in both directions, and it lets you set the mean and standard deviation, because different test scales use different values and the same number means different things under each. The rarity figure — one person in how many scores at or above this point — is included because it is far easier to read than a percentile with several decimal places.
Arb Digital publishes this as a statistics tool, and it is worth being blunt about what that means. This page does arithmetic on a normal distribution. It is not an IQ test, it does not measure anything, and no number it prints tells you anything about any person's intelligence. What it does is answer a narrow, well-defined question: given a normal distribution with a stated mean and spread, what proportion of it lies below a given point? That is a legitimate calculation with a definite answer, and it is the only claim made here.
What This IQ Percentile Calculator Does
It standardises a score into a z-score with z = (score − mean) ÷ standard deviation, then evaluates the standard normal cumulative distribution function at that z to get the percentile. Reverse the direction and it inverts the same function: you give a percentile, it returns the score at that position. Alongside the percentile it reports the proportion below, the proportion at or above, and the reciprocal of that upper proportion expressed as a one-in-N rarity.
Our general percentile calculator does a related but genuinely different job: it ranks a value inside a data set you supply, making no assumption about the distribution's shape. This page assumes a normal distribution with parameters you specify and never looks at raw data at all. If you have actual observations, the data-driven tool is the better instrument. If you have a scaled score and want its theoretical position, this one is. The z-score calculator handles the first half of the conversion on its own, and the inverse normal distribution calculator handles the reverse direction for any quantity, not just this scale.
How to Use It
- Choose the direction. Score to percentile is the common case. The reverse is useful for working out what score corresponds to a stated cut-off.
- Check the standard deviation before anything else. A score report should state the scale it uses. If it does not, the percentile you calculate is not reliable.
- Leave the mean at 100 unless your scale says otherwise. IQ scales are constructed so the norming sample averages 100; that is a definition, not a finding.
- Read the rarity figure. "One in 44" communicates the same information as "97.7th percentile" and is much harder to misread.
- Treat extreme values with suspicion. Percentiles far out in the tail depend entirely on the normal assumption holding there, and it generally does not.
The Formula and How It's Calculated
Two steps. First standardise: z = (score − μ) ÷ σ. Then apply the standard normal CDF, percentile = Φ(z) × 100. Φ has no elementary closed form, so it is computed here through a high-accuracy complementary error function, and the reverse direction uses a rational approximation to the quantile function refined by Newton iteration. The NIST/SEMATECH e-Handbook page on the normal distribution states the same point: the percent point function of the normal has no simple closed form and is computed numerically.
Take the default. A score of 130 on a scale with mean 100 and standard deviation 15 gives z = (130 − 100) ÷ 15 = 2.0 exactly. Φ(2.0) = 0.97725, so the percentile is 97.725. The proportion at or above is 1 − 0.97725 = 0.02275, and one divided by that is 43.96 — roughly one person in 44 of the reference population.
Now change only the standard deviation to 16, as some scales use. The same score of 130 becomes z = 30 ÷ 16 = 1.875, the percentile drops to 96.96, and the rarity falls from one in 44 to about one in 33. Nothing about the person changed. Nothing about the test changed. A single parameter in the conversion changed, and the apparent rarity moved by a third. This is the most important thing on the page: a raw IQ number quoted without its standard deviation is not a complete piece of information.
Why the Standard Deviation Has to Be Stated
An IQ score is not a measurement in natural units the way height or mass is. It is a standardised position, defined by where a raw performance falls relative to a reference sample. To turn it into a number, the scale designer picks a mean and a standard deviation and stretches the distribution to fit. Wechsler-type scales conventionally use 15. Some other instruments have used 16. The older Cattell scale used 24, which is why scores from it look dramatically inflated when read against a 15-point assumption.
The consequences grow the further from the mean you go. At a score of 115 the percentile is 84.13 under σ = 15 and 82.58 under σ = 16 — a small difference. At 145 it is 99.865 versus 99.754, which sounds small until you convert it to rarity: one in 741 against one in 407. The same score describes a group over eighty percent larger under the second assumption. Any comparison of scores from different instruments without converting to a common scale first is comparing two different things.
Norming: What the Percentile Is Actually Relative To
A percentile from this calculation is a position within a modelled distribution, and that distribution stands in for a particular norming sample — a group of people, of a particular age range, in a particular country, tested in a particular year. Change any of those and the same raw performance maps to a different score. This is why scores are always tied to a specific test and a specific norming edition, and why a score is not portable between instruments.
There is also a well-documented drift over time in measured performance on these tests across the twentieth century, which is why test publishers re-norm periodically. Re-norming resets the mean back to 100 for the new sample, so a person whose performance is unchanged can be assigned a lower score against newer norms than against older ones. None of that is visible in a bare number. A score without the test name and the norming year attached is missing most of the information needed to interpret it at all.
Where the Normal Model Breaks Down
The normal distribution is imposed on these scales by construction, not discovered in the data, and it fits reasonably through the middle and poorly at the extremes. Real score distributions on cognitive tests tend to show more mass in the tails than a normal curve predicts, so the one-in-N rarity figures for very high and very low scores are model outputs rather than population counts.
The arithmetic makes this concrete. A score of 160 on a σ = 15 scale is z = 4, which the normal model puts at roughly one in 31,600. A score of 175 is z = 5, or about one in 3.5 million. Those numbers are produced by extrapolating a smooth curve into a region where almost no norming data exists — a norming sample of two thousand people contains no information whatsoever about a one-in-a-million position. Most tests also have a ceiling above which they simply do not discriminate. Treat any extreme figure this page produces as a property of the normal curve, not as a fact about people. Penn State's open STAT 500 applied statistics notes cover the normality assumptions that this kind of extrapolation quietly depends on.
Reading Percentiles Without Fooling Yourself
Percentiles are compressed in the middle and stretched at the edges, which distorts intuition in a specific direction. Between scores of 90 and 110 — a twenty-point span — sits about 49% of the modelled population. Between 130 and 150, also twenty points, sits about 2.2%. Equal score differences are not equal population differences, so the gap between two scores near the mean represents far more people than the same gap out in the tail.
The reverse trap matters too. Small measurement differences near the extremes swing the percentile a lot, and every real test has a standard error of measurement, typically several points. A score reported as 132 might reasonably be 127 or 137 on a retest, and the rarity figure moves substantially across that band. Anyone quoting a percentile to two decimal places from a single administration is reporting precision the underlying measurement does not have. Our standard deviation calculator and normal distribution calculator cover the general machinery behind both of those points.
Arb Digital writes data-heavy pages that state their assumptions instead of hiding them, which is also what makes them defensible.
Explore Content Marketing Talk To Our TeamCommon Mistakes to Avoid
- Quoting a score without its standard deviation — the same number gives materially different percentiles under a 15-point and a 16-point scale, and wildly different ones under a 24-point scale.
- Comparing scores from different tests — different instruments, different norming samples and different years are not on a common footing, whatever the numbers look like.
- Treating a tail rarity as a population count — one in three million is an extrapolation of a smooth curve, not a headcount, and no norming sample contains that information.
- Ignoring measurement error — a single test administration carries a standard error of several points, which moves the percentile more than most people expect.
- Reading this page as a test — it converts a number you already have into another representation of the same number. It measures nothing.
Related Free Tools From Arb Digital
Standardise any value with the z-score calculator, work forwards from value to probability with the normal distribution calculator, go from probability back to a value with the inverse normal distribution calculator, rank a value inside real data with the percentile calculator, or measure spread with the standard deviation calculator. Everything we publish is listed on the free online tools hub.
Frequently Asked Questions
Subtract the distribution mean from the score and divide by the standard deviation to get a z-score, then evaluate the standard normal cumulative distribution function at that z. Multiply by 100 to express it as a percentile.
No. It is a statistical conversion between a score, a z-score and a percentile on a normal distribution. It administers no test, measures nothing, and tells you nothing about any individual's abilities.
Because it sets how far a given score sits from the mean in standard units. A score of 130 is two standard deviations above the mean on a 15-point scale but only 1.875 on a 16-point scale, which changes the percentile and roughly a third of the rarity figure.
Whichever one your score report states. Wechsler-type scales conventionally use 15, some other instruments have used 16, and the older Cattell scale used 24. Guessing produces a percentile that does not correspond to your score.
They are exact outputs of the normal model and not reliable descriptions of a population. Real score distributions have heavier tails than a normal curve, and no norming sample is large enough to contain information about one-in-a-million positions.
Not directly. Scores are relative to a specific instrument, a specific norming sample and a specific norming year, so two numbers from different tests are not measured against the same reference even if both are labelled IQ.
Because measured performance drifts over time, so the mean of a new sample would otherwise no longer sit at 100. Re-norming resets the scale, which means unchanged performance can map to a different score against newer norms.
This page performs a statistical conversion on a normal distribution you specify. It is not an IQ test, it is not a psychological assessment, and it provides no medical, educational or diagnostic advice. Any question about a real assessment belongs with the qualified professional who administered it and holds the official score report.
