Data

Every published statistic on this site comes from anonymous test results submitted by people who used it — not from an external study, not from an estimate, and not from a number we thought sounded right. This page explains what we collect, why we publish it, and links to the actual distributions.

What we collect

When you finish a test, one anonymous record is sent to us: which test, the test mode, your score, a device category (desktop, mobile, or tablet), a two-letter country code, and the time the record was created. That is exactly six values, and nothing else — no IP address, no user-agent string, no cookies, no account identifier, and no way to link two of your visits together. Full detail is in our methodology and privacy policy.

Why we publish it

Most "average CPS" or "average reaction time" figures circulating online are unsourced — nobody says where the number came from, how many people it is based on, or over what period. We would rather show you our actual distribution, with its actual limits, than repeat a number nobody can trace.

How big is the dataset

Collection began on 29 July 2026, so this is genuinely an early dataset — currently in the tens of records per test, not thousands. We are not going to dress that up: small samples produce noisy percentiles, and a distribution built on a handful of device-category results is not reliable. Where a segment is too thin to say anything meaningful, the page says so instead of showing a number. Every figure we do publish states its exact sample size (N) and collection period next to it, recomputed from the live dataset at most a minute apart — not baked in at a past build.

The two distributions

CPS distribution — clicks per second, broken down by test mode (5s / 10s / 30s) and device category.
Reaction time distribution — reaction time in milliseconds, broken down by device category.

An aim-accuracy distribution page is not published yet — the current aim test sample is too small to say anything honest about it. We would rather publish nothing than publish a percentile built on a handful of points.

How the numbers are produced

The figures on the distribution pages are not written into the page at build time. They are computed from the live dataset each time the page is requested and cached for at most a minute, so what you see is the dataset as it stood moments ago — including outlier handling and the exact rounding rule, both documented on each page. If the underlying data cannot be reached for a moment, the page says so rather than showing a stale or invented number.