How percentiles are calculated
Last updated 14 November 2026
What a percentile means here
A result such as "Top 8%" means your score is ahead of 92% of the people it is ranked against. Below the middle we say "Ahead of X%" instead.
Each result leads with one place: your city, your country or the world. Another place is shown beside it only when it says something different and has at least 100 real results behind it. All three are always one tap away.
How a test is ranked
Every test starts with a written-down starting curve: how scores are expected to spread, worked out from how the test is built and how hard each level is. Until 1,000 people have a saved result on a test, its percentiles come from that curve alone, so they are the same for everyone, everywhere. From 1,000 the test is live: you are ranked against the people who took it, in your city, your country and the world, and the curve counts for less with every result.
Only results saved after sign-in enter a ranking. You get one free counted try at each test every 24 hours, and your ranking is your best counted try of the last 30 days; other goes are practice and are not ranked. Playing on a run of days adds free extra counted tries; a missed day in any seven is let go. Answers that come in faster than a person could read them are not counted.
Speed tests that depend on the pointer, such as aim, are ranked separately for a mouse and a touch screen.
On a sprint you answer as many questions as you can on one clock. A right answer is +1, a wrong answer is -1 and a skipped question is nothing; the score never goes below 0.
Tests that rank nobody
Self-ratings and tick lists, such as How Desi Are You?, show your own answers as a percentage: 73% means your answers sit 73% of the way from one end of the scale to the other. They are not compared with anyone. Mental Age and Blue or Green are the exceptions: they say where you sit among everyone who took them, with no better or worse.
Type tests (Aura, Spirit Animal and the like) give the type most of your answers point to, and games such as the name game give their own result. None of these is ranked or kept in a pool, and no result is saved. On self-ratings, totals of how people answered each question are kept, counted once for each signed-in account, to show how others answered.
Kid mode
In kid mode a test gives a score and one to five stars, and no percentile. The stars come from where the score sits on the test's starting curve. That curve is drawn for adults, so the bands are set low: finishing a test earns one star, and five stars do not need an adult's best score.
Where the comparison data comes from
For numbers you type in, a public source is the comparison where one exists: a government survey, an open dataset or a published study. Each one is listed with its name, year and sample size.
Some sources cover only one country or region. Where your country has no data, the closest similar country is used.
Apart from each test's starting curve, which is written down and gives way to real results once 1,000 people have taken the test, no comparison figure on this site is generated by AI or estimated without a named source.
Exam papers
An exam paper on this site is written by us to the pattern of a public exam: the same sections, timing and marking. No question comes from a real paper, and the site has no link to the body that runs the exam.
Its result is an estimate of where the same marks would have stood among the people who sat the real exam that year, and is marked "Estimated". The body that runs the exam publishes no table of marks against percentile, so the estimate comes from publicly available figures: scores that coaching firms and candidates published for stated percentiles after the result. For each percentile we take the middle of the sources that state it, so no single source decides the curve. Published scorecards of individual candidates are used to check it.
The real exam reports a scaled score, adjusted between its sittings. Our paper is one sitting with no such adjustment, and it is not the same paper, so the same marks on it may be a little easier or harder to reach.
Results are whole percentiles below 99 and carry one decimal from 99 to 99.9. Beyond that the result reads "Above 99.9 percentile": the public figures are too few there to say more. Below the lowest published figure the result reads "or below".
A paper is sat once. Its clock runs from the moment you start, whether or not the page is open. The free short version of a paper is an ordinary test, ranked the way every test here is; it says nothing about the real exam.
From a published table to your result
Sources usually publish a handful of percentile points, for example the 5th, 10th, 25th, 50th, 75th, 90th and 95th. Your value is placed between the two nearest published points by straight-line interpolation. For long-tailed measures such as money, the interpolation is done on a logarithmic scale.
We do not guess beyond the published points. If your value is past the highest published point, the result reads "Top 5% or better" (or whatever that point is) rather than a made-up figure.
How results are shown
Percentages are whole numbers. A result is never shown as "top 0%"; the limit is "top 0.1%".
Among the real results on a test, your rank is the share of results below yours plus those equal to yours: a tie counts as a place gained.