A rank is only
as honest as its ruler.
Every test in this room hands you a number and a place in a line. This page is about where that line comes from, what it is measured against, why the questions are built the way they are, and the several things a score like this cannot tell you about you. Read it once and the rest of the room reads differently.
A percentile here is not a percentile out there
When a test says you placed above 72%, that is 72% of the people who finished the same test on this site. It is not a national sample, not a standardised population, and not a comparison against anyone who started and walked away. Three things follow from that, and they matter more than the number itself.
- The pool decides what the number is worth. A percentile drawn from forty finishers and one drawn from forty thousand are different animals wearing the same clothes. Every result screen prints the pool size next to the rank for exactly this reason. If the pool is small, read the rank as a rough bucket — upper third, middle, lower third — and ignore the digits.
- The people here are not a random sample of anyone. Nobody arrives at a reaction-time test by accident. The room self-selects for people who like being measured, which tends to pull the whole distribution one way and makes the middle of this line sit somewhere other than the middle of the population. Being average here is not being average.
- Unfinished runs are dropped, and that flatters everyone. A rank is computed only over completed attempts. People who abandon a test halfway are disproportionately people who were doing badly, so every distribution on this site is quietly missing its own left tail. There is no honest fix for this that does not involve counting quitters as zero, which would be its own kind of lie.
The fast answer is wrong on purpose
Questions here are not sampled from the world. They are selected, deliberately, from the narrow band where the answer that arrives first is the answer that is wrong. That is the house style, and it has two honest consequences you should know before you read your own score.
The first is that your hit rate will read low — lower than it would on a general knowledge set, and lower than you probably deserve. A bare “you got four out of ten” would be a misleading headline, which is why no test in this room uses one. The second is that a slow answer beats a fast one more often here than anywhere else. On a normal quiz, hesitation is a sign of not knowing. Here it is frequently the sign of having noticed the trap.
So the number a test reports is never simply “how much do you know”. Each one is built to isolate something narrower, and the table is the short version of what each one is actually holding still.
| Test | What it isolates | What moves the number most |
|---|---|---|
| The ten-question round | Whether you can stop the first answer from leaving your mouth | How long you are willing to sit with a question that looks easy |
| How sure were you | Whether your stated confidence means what it says | Your habit at the top end — what you call 90% is the whole test |
| How fast are you, really | Simple visual reaction time, minus your hardware | Your display and input lag, often by more than your age does |
None of the three is a measure of intelligence, and the room does not report one.
Four things a score like this cannot tell you
- One sitting is one mood. Reaction time swings by a tenth of a second across a single day in the same person — more than the gap between two neighbouring decades of age. Calibration moves with how much you care about the subject. A single run captures a person on one afternoon, and the honest way to read it is as one reading, not as a trait.
- Ten items is far too few to measure anybody. This is not modesty, it is arithmetic. Split ten answers across three confidence levels and some of those groups contain two items; one lucky guess then moves the whole picture. The research these tests point at — that most people’s 90% behaves like 70% — comes from studies running hundreds of items per person. We are illustrating that finding, not reproducing it.
- Your equipment is in every number. This is sharpest on reaction time, where the gap between a 60 Hz laptop panel and a 144 Hz screen is worth more milliseconds than most of the biological variation the test is nominally about. The reaction test estimates and subtracts what it can, and says so on the page; no estimate of that kind is exact.
- Nothing here is diagnostic. Not for attention, not for memory, not for cognitive decline, not for anything with a clinical name. These are browser toys built on real scoring rules, which is a different thing from an instrument. If a result worries you, that is a conversation for a doctor, and a slow score on a web page is not evidence of anything you should act on.
What gets recorded while you play
Enough to compute a rank and to see where people stop reading, and nothing that identifies you. There is no account, no sign-in, and no name or email attached to a run. What is stored is which test you opened, how far down the page you got, your answers for the run in progress, the country your connection appears to come from, and a random identifier held in your own browser so a repeat visit is not counted as a stranger. Clearing your browser data ends that identifier and there is no copy of it anywhere else. The full version is in the privacy policy, which is written to be read rather than agreed to.