IQ Percentile Calculator

The percentile is the honest part

Of everything on this page, the arithmetic is the bit you can trust completely. Converting a score into a percentile is a clean statistical operation with a right answer, and this tool gets it right.

What a percentile tells you is exactly one thing: where a number sits in a distribution. Feed it 130 and it says 97.7, meaning roughly 98 people in 100 scored lower. That is a true statement about the position of a number on a curve, and it is all a percentile ever claims.

Everything else is where it gets complicated. Whether the score you typed in means what you think, which ruler it was measured on, how much of it is measurement noise, and what the label underneath it is implying. So this page is mostly about the assumptions the calculator makes silently, because the arithmetic needs no defending and the assumptions do.

What the calculator is actually doing

IQ scores are built, deliberately, to follow a normal distribution: the bell curve. That is not a discovery about human minds, it is a design decision. Test makers take raw scores from a large standardisation sample and stretch them onto a curve with a chosen mean and spread. The curve is the packaging.

This tool assumes the standard modern packaging:

• Mean of 100. By definition, the average is 100, because that is what the scores were scaled to.
• Standard deviation of 15. So about 68 percent of people fall between 85 and 115, and about 95 percent between 70 and 130.

Then it runs the cumulative distribution function of that normal curve, which is the standard way of asking "what fraction of the area under this bell sits to the left of my score". Under the bonnet it uses a well-known mathematical approximation of the error function to compute it, accurate to about seven decimal places, which is considerably more accuracy than anything else in this process deserves.

The point worth holding onto: the calculator is not measuring anything. It is doing geometry on a curve someone else designed. Its answer is only as meaningful as the score you fed it.

A worked example

An IQ of 130.

Distance from the mean = 130 − 100 = 30 points
In standard deviations = 30 ÷ 15 = 2.0
Area under the normal curve below 2 standard deviations = 97.7%

So a score of 130 sits at about the 97.7th percentile: roughly 2.3 people in 100 score at or above it. That is the number behind most gifted-programme thresholds and behind Mensa's entry requirement, and it is why 130 has the reputation it does. It is not a magic number. It is just two standard deviations, and two standard deviations is a tidy place to draw a line.

The landmarks worth knowing

The whole scale, in the only terms that mean anything:

Notice how the top of the scale behaves. Going from 100 to 115 moves you past a third of the population. Going from 145 to 160, the same 15 points, moves you past about 0.13 percent. The curve is thin out there, which has two consequences: small differences in score translate into dramatic-sounding rarities, and there are so few people in the standardisation sample above 145 that the scores themselves become progressively less reliable. A reported 160 is doing a lot of extrapolating.

Why the number is less precise than it looks

This tool reports your percentile to one decimal place. An IQ of 130 comes back as 97.7. Even that is more precision than the underlying score can support, and here is the arithmetic that shows why.

Every psychological test has a standard error of measurement: the amount your score bounces around purely from the imprecision of the instrument, with no change in you at all. For the WAIS-IV, the most widely used adult test, that figure is about 2.5 IQ points, which gives a 95 percent confidence interval of roughly plus or minus 5 points.

Apply that honestly to our 130:

So a reported IQ of 130 honestly means "somewhere between the 95th and the 99th percentile". That is a band nearly four percentile points wide, on a quantity that is genuinely uncertain in its first two digits. An earlier version of this tool printed five decimal places, which implied precision to one part in ten million and was pure fiction. It now prints one, and even that last digit is decoration.

Which suggests the correct way to read any score: as a range, never as a point. A psychologist reporting properly writes something like "approximately 120, with a 95 percent probability of the true score falling between 115 and 125". "Your IQ is 120" is, strictly speaking, an incorrect statement, and it is the one everybody makes.

Read the tool's output as its first two digits and mentally delete the rest.

Why the tool asks which ruler you used

This is the trap that produces the biggest errors here, and it is the reason the tool asks you a second question rather than one.

A standard deviation of 15 is correct for the Wechsler scales and for the modern Stanford-Binet, which covers most tests you are likely to have taken. But it is not universal, and a score means nothing until you know which ruler produced it:

Those three scores are the same achievement. None is higher than the others. They are the same percentile written on different rulers, which is why Mensa quotes an entry requirement as a percentile, the 98th, and then publishes different score thresholds for different tests.

Now watch what happens if you get it wrong. A Cattell score of 148 sits at about the 97.7th percentile on its own scale. Score it as though it were SD 15 and it comes back as the 99.93rd percentile, telling someone in the top 2 percent they are in the top 0.07 percent. That is a thirtyfold error in rarity, and nothing on screen warns you.

So before you trust any answer here: check what test your score came from and what its standard deviation is. If it is 15, you are fine. If it is anything else, this tool is the wrong instrument.

Where the score came from matters more than the score

Worth being direct about, because it is the most common reason people arrive at a page like this.

If your number came from a free online test, the percentile this tool gives you is arithmetic performed on a number of unknown provenance. Not necessarily wrong. Just unverifiable, and unverifiable in ways that all point the same direction.

The differences between a clinical instrument and a web quiz are not small:

• Standardisation. A test like the WAIS-IV was normed on a large, carefully constructed sample designed to represent the population. Most online tests are normed on whoever happened to take them, which is a self-selected group of people who wanted to take an IQ test.
• Administration. Clinical tests are given one-to-one by a trained psychologist over a couple of hours, under controlled conditions. A web test is you, on a sofa, possibly distracted, possibly having a go for the second time.
• Error. A properly administered test has a confidence interval of about plus or minus 5 points. A short online test may have a band of plus or minus 10 to 15, which is wide enough to swallow most of the distinctions people care about.
• Scoring. Some online tests are calibrated so that most takers get a flattering result, for reasons that are commercial rather than psychometric.

None of which makes them worthless as entertainment. It does mean that running a web-test number through a precise percentile calculator produces false precision squared: an exact percentile for an inexact number from an unknown distribution.

Why there is a band and not a label

Alongside the percentile, the tool prints a descriptive band: "average range", "above average range", and so on. It deliberately does not print a classification like "Superior" or "Borderline impaired or delayed". Two reasons, and they are the reason those labels were taken out.

The words are conventions, not findings. The numeric structure of IQ, mean 100 and a fixed spread, has been stable since the 1930s. The words attached to the bands have been rewritten repeatedly, because they are cultural judgements about how to describe people, and each generation finds the previous generation's vocabulary uncomfortable. Older schemes used terms that would be unpublishable now. Current Wechsler practice has moved towards neutral, descriptive language, preferring formulations like "extremely high" and "extremely low" over the older evaluative labels. Labels of the "impaired or delayed" kind belong to an earlier convention, which is why they are no longer printed here.

More substantively, the lower labels describe something the field no longer diagnoses this way. This matters more than the vocabulary. A low IQ score, on its own, is not intellectual disability and has not been for some time. Modern diagnostic practice under DSM-5 requires three things together: significant limitations in intellectual functioning, significant deficits in adaptive functioning across everyday domains, and onset during the developmental period. All three. And crucially, DSM-5 bases the severity of intellectual disability on adaptive functioning rather than on the IQ number, precisely because the number turned out to be a poor guide to how much support a person actually needs.

So a calculator that reads a score and prints "moderately impaired" is performing a classification that no clinician would make from that information. If you or someone you know has scored low on any cognitive assessment and it matters, the answer is a full evaluation by a qualified psychologist, not a label from a web page. And if the score came from an online test, it does not bear on this question at all.

The Flynn effect, or why 100 keeps moving

Here is the fact that makes the whole scale stranger than it looks.

Average raw performance on IQ tests rose substantially across the twentieth century, in many countries, at something like three points a decade. But the average score stayed at 100 the whole time, because 100 is not a fixed quantity of anything. It is a percentile in disguise. Every time a test is renormed against a fresh sample, the average is reset to 100 by definition.

Which means a score of 100 in 1960 and a score of 100 today are not the same performance. They are the same position against different populations. Your grandfather's average is not your average, and comparing scores across eras without adjusting is a category error.

This is not a curiosity. It is consequential enough that the Flynn effect is formally built into diagnostic practice around intellectual disability, and it has been argued over in capital cases, where a few points of adjustment for the age of a test's norms can change an outcome entirely.

Interestingly, the effect may be fading. When the newest Wechsler adult scale was validated against its predecessor, the expected three-point gain came out at about 1.2 points, which is well short of the historical pattern and part of a wider debate about whether the rise has slowed or stopped in some countries.

The takeaway for this page: your percentile is relative to a specific standardisation sample from a specific time. It is a statement about where you stand among those people, not a measurement of a quantity you carry around.

What the number does and does not predict

Worth being even-handed, since this topic attracts both dismissal and reverence.

On the side of the sceptics being wrong: IQ scores are not meaningless. They are among the better-studied measures in psychology, they are reasonably stable in adults, and they correlate with real outcomes, particularly educational attainment and, more modestly, job performance, with the relationship strongest in complex work. Tests measure something, and that something is not nothing.

On the side of the enthusiasts being wrong: the correlations are moderate, not deterministic. They describe tendencies across populations, and a tendency across a population tells you remarkably little about a specific person, which is the mistake almost everyone makes with these numbers. The variance left unexplained is much larger than the variance explained, and it is filled by everything else: what you do with your time, conscientiousness, health, opportunity, teaching, money, and the ordinary accidents of a life. Plenty of high scorers do very little with it. Plenty of average scorers build extraordinary things.

There is also the standing question of what the tests measure. They are good at capturing certain kinds of reasoning, working memory and processing speed. They do not attempt to measure judgement, creativity, persistence, or the ability to work with other people, and nobody serious claims otherwise. A test that has not tried to measure something cannot tell you that you lack it.

So the honest summary of the number this calculator gives you: it is a real position in a real distribution, measured with meaningful error, on a scale that has moved over time, using a test whose quality you should check. That is genuinely interesting. It is not a verdict, and it is a poor thing to build an identity on in either direction.

Questions people ask

What percentile is an IQ of 130?

About the 97.7th, meaning roughly 2.3 people in 100 score at or above it. That is two standard deviations above the mean and is the usual threshold for gifted programmes and for Mensa. Read honestly with the measurement error included, a reported 130 means somewhere between the 95th and 99th percentile.

What IQ do I need for Mensa?

Mensa's criterion is a percentile, not a score: the top 2 percent, meaning the 98th percentile or above. That works out at roughly 131 on a scale with a standard deviation of 15, about 133 on the old Stanford-Binet, and about 149 on Cattell. Those are the same achievement expressed on different rulers, which is exactly why Mensa quotes the percentile and publishes separate thresholds per test.

How precise is the percentile?

Less precise than it looks. The clinical standard error of measurement is around 2.5 points, so a reported score of 130 covers roughly the 95th to 99th percentile. The tool prints one decimal place; read the whole number and treat even that as the middle of a range.

I got my score from a free online test. Is the percentile valid?

The arithmetic is valid; the input probably is not. Online tests are typically normed on whoever chose to take them rather than a representative sample, are taken in uncontrolled conditions, may carry error bands of 10 to 15 points, and are sometimes tuned to flatter. Running that number through a precise calculator gives you an exact percentile for an inexact score from an unknown distribution.

Does it matter which test my score came from?

Enormously, and this is the tool's biggest blind spot. It assumes a standard deviation of 15, which is right for Wechsler and modern Stanford-Binet tests and wrong for others. A Cattell score of 148 is about the 97.7th percentile on its own scale, but this tool will report it as the 99.93rd, telling someone in the top 2 percent that they are in the top 0.07. Check your test's standard deviation before trusting any answer here.

References

Where the figures come from. The scale this tool assumes, a mean of 100 with a standard deviation of 15, along with the standard error of measurement of roughly 2.5 points for the Full Scale IQ and the descriptive classification scheme, are properties of the Wechsler Adult Intelligence Scale as documented in its technical manual. The diagnostic framework described above, in which intellectual disability requires deficits in intellectual functioning, deficits in adaptive functioning and onset in the developmental period, and in which severity is determined by adaptive functioning rather than by IQ score, is set out in DSM-5. The Flynn effect, its incorporation into diagnostic criteria by the AAIDD and DSM-5-TR and into capital cases, and the finding that the expected three-point generational gain came out at about 1.2 points in the WAIS-5 validity studies, are discussed in the analysis linked below.

1. Wechsler D. Wechsler Adult Intelligence Scale, Fourth Edition (WAIS-IV): Technical and Interpretive Manual. Pearson; 2008.
2. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5). Arlington, VA: American Psychiatric Publishing; 2013.
3. Wait, Where's the Flynn Effect on the WAIS-5? https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11595985/