The 15 millisecond filter, and why it is the useful part
A mechanical mouse switch is a metal contact, and metal contacts bounce. New switches bounce for well under a millisecond and the mouse firmware debounces that away invisibly. As a switch wears, the bounce gets longer and less predictable, and eventually it outlasts the firmware's debounce window. At that point one physical press produces two click events a few milliseconds apart, and the computer has no way to know they were not both real.
A human finger cannot produce two clicks 15 milliseconds apart. Even the fastest specialised techniques sit above 60 milliseconds between clicks. So anything under 15 milliseconds is discarded here and counted separately, and that count is genuinely diagnostic: if a ten-second run discards several clicks, the switch in that mouse has almost certainly started to fail. The same fault shows up elsewhere as files opening when you meant to select them, and drags that turn into double-clicks.
Five seconds and ten seconds measure different things
Rapid clicking fatigues fast. The forearm muscles doing it are working anaerobically and begin slowing within a few seconds, so a five-second run is close to your peak rate while a ten-second run averages that peak with the decline that follows it. Most people are somewhere between half a click and a full click per second faster over five seconds than over ten.
Neither is more correct — they answer different questions. What matters is not mixing them. A five-second score set against someone else's ten-second score is meaningless, and so is comparing your own runs across the two settings.
Technique, and a caution about it
Ordinary clicking with one finger is what the default measures. There are faster methods: tensing the arm so the hand vibrates and the finger clicks on the tremor, or alternating two fingers on the same button. Both push the rate well past what a single relaxed finger achieves, and both are how very high numbers get posted.
They also load the wrist and fingers in a way that is not gentle. This is a novelty measurement with nothing riding on it, and there is no version of a high score here worth a strained tendon. If anything twinges, stop. Discomfort from this kind of repetitive motion tends to arrive after you have finished rather than during, which is exactly what makes it easy to overdo.
Questions people ask
When does the timing actually start?
On your first click on the blue pad, not when you press the button above it. That means you can take as long as you like getting your hand into position, and the first click of the run counts toward the total as well as starting the clock.
Does a touchscreen count fairly?
It counts every tap, but touch panels sample at their own rate and some devices coalesce taps that arrive close together, so a touchscreen result is not comparable with a mouse result. Two thumbs alternating is faster still and is a different activity again. Pick one method and stay with it.
Some of my clicks seem to vanish. Why?
Either they arrived within 15 milliseconds of the previous one and were filtered as switch bounce, in which case the discarded count will show them, or the page fell behind. A heavily loaded machine or a browser with many tabs open can delay click events enough to lose a few at high rates. Check the discarded count first, since a high number there points at the mouse rather than the browser.
Is the score stored?
No. There is no history, no leaderboard and no upload. The count lives in the page while it is open and disappears with the tab.