The clock
All timing uses performance.now(), the browser's monotonic high-resolution timer. Monotonic matters: unlike Date.now(), it cannot jump backwards when the system clock is corrected by NTP or by a daylight-saving change mid-test. A test that used the wall clock could quietly report a negative elapsed time, and some do.
Its resolution is not unlimited and we are not going to pretend it is. Browsers deliberately coarsen this timer as a defence against timing side-channel attacks, typically to about 0.1 milliseconds and sometimes worse depending on how the page is isolated. For a click test that is irrelevant, because your fingers are four orders of magnitude slower. For the reaction time test it is the smallest term in a much larger error budget, which is the next section.
The countdown is driven by requestAnimationFrame, so it advances once per rendered frame rather than on a fixed interval. One consequence is worth stating plainly: if you switch to another tab mid-test, the browser stops delivering frames and the countdown stops advancing. The test does not silently keep running in the background and it does not credit you with the missing time. Come back and it resumes from where the last frame left it.
Click speed: what is counted
The click area counts a mouse button press, not a completed click. The distinction is real: a completed click event only fires when press and release land on the same element, which at high click rates throws away presses that a player genuinely made. Press events are what the hand actually did.
The first press starts the clock and also counts as click one. That convention is what every site in this niche uses, so scores stay comparable, but it has an arithmetic consequence that nobody states and we will: it credits you with one click for zero elapsed time. On a ten second run that inflates your score by 0.1 CPS. On a one second run it inflates it by a full 1.0 CPS, which on a typical result is more than ten percent. If you want the strict interval rate, subtract one from your click count and divide again. Short burst scores are the least comparable numbers on this site and that is the reason why.
Your score is rounded to two decimals for display, and it is rounded in the ordinary way, not upward. There is no bonus for a final click, no padding, and no adjustment to make a number look better than it is. A personal best is only overwritten when the new score is genuinely higher.
Synthetic clicks are rejected. Any event whose isTrusted flag is false, which is what a script dispatching new MouseEvent(...) produces, is discarded rather than counted. Real input from a mouse, a touchscreen or the keyboard is always trusted, so nothing legitimate is thrown away.
What that check does not catch, and we are not going to pretend otherwise:a hardware auto clicker presses a real switch, and a browser extension driving the debugger protocol produces genuinely trusted events. Neither is distinguishable from a person by any web page, on this site or any other. Claims elsewhere that a browser test "detects auto clickers" are, at best, describing this same flag and overselling it. In practice it does not matter here: your scores live only in your own browser, there is no leaderboard and nobody to fool but yourself. It would matter the moment a public ranking existed, which is one of the reasons there is not one.
Reaction time: why your result is not purely your reflex
A typical adult reaction time on a simple visual test is somewhere around 200 to 300 milliseconds.
This is the range quoted almost everywhere online. ClickStorm has not measured it and has not verified a primary source for it, so it is presented as an estimate rather than a fact. What the site can state precisely is how much of your measured number is the browser and the hardware rather than you: see the measurement chain below.
A browser reaction test measures the round trip from "the colour changed in a frame we submitted" to "a click event reached our handler". Several fixed delays sit inside that round trip and none of them are your nervous system:
| Stage | Typical cost | Why it is there |
|---|---|---|
| Display refresh | up to 16.7 ms at 60 Hz, up to 6.9 ms at 144 Hz | The colour change is drawn into a frame. If the frame has just been sent, the pixels do not change until the next one. That wait is added to your score and it is not your reflexes. |
| Panel response | roughly 1 to 20 ms depending on the panel | The time the pixels themselves take to actually change colour after the frame arrives. An office IPS panel sits at the slow end of that; a gaming panel at the fast end. |
| Mouse polling | up to 8 ms at 125 Hz, up to 1 ms at 1000 Hz | Your click is not reported the instant you press. It waits for the mouse's next report to the PC. Run the polling rate test to see which of these applies to you. |
| Event dispatch and timer resolution | sub-millisecond, plus browser clock coarsening | Browsers deliberately blunt performance.now() as a defence against timing attacks, typically to 0.1 ms or coarser. That is small next to the display, but it means the last decimal of any browser timing is decoration. |
Where this comes from: arithmetic and published hardware specifications, not an opinion. A 60 Hz display sends a frame every 16.7 milliseconds by definition, and a 125 Hz mouse reports every 8 milliseconds by definition. You can verify your own polling figure with the polling rate test.
Adding the chain up, a browser reaction test on typical hardware reports roughly 10 to 45 ms more than your actual reflex, and almost all of that is the screen. That overhead is why a result here is only comparable with another result measured on the same setup.
The practical rule that follows: a reaction time from this site is a fair comparison against your own earlier results on the same screen and the same mouse, and a poor comparison against a number someone else posted on different hardware. The test averages five attempts for the same reason, since a single trial is dominated by whether you happened to be mid-blink. A click before the colour changes is a false start and is discarded rather than counted as a very fast time.
Polling rate: a floor, not a ceiling
A browser can only count the mouse events it is actually handed, and it is handed them per animation frame. Your measured Hz is therefore a floor, not a ceiling: it lands at or below your true polling rate, and it only approaches the true figure while you are moving the mouse continuously and fast. A reading well under a standard value usually means you did not move enough, not that your mouse is slower than advertised.
| Rate | One report every | What it is good for |
|---|---|---|
| 125 Hz | one report every 8 ms | The USB default. Completely fine for browsing, office work and most single-player games. If your mouse has no configuration software, this is almost certainly what you have. |
| 500 Hz | one report every 2 ms | The point where the improvement over 125 Hz stops being subtle for fast mouse movement. Most gaming mice ship at or above this. |
| 1000 Hz | one report every 1 ms | The long-standing gaming standard and the setting most people should pick. Beyond here the returns get very small very quickly. |
| 2000 to 8000 Hz | one report every 0.5 ms down to 0.125 ms | Available on recent high-end wired and wireless mice. Measurably lower latency, but it raises CPU load and some systems handle it badly. Worth testing rather than assuming. |
The result is rounded to the nearest standard rate before it is shown as a "likely polling rate", because a raw reading of 948 Hz almost certainly means a 1000 Hz mouse and a couple of dropped frames. The unrounded maximum and average are shown next to it so you can see how far the rounding travelled. If those two numbers are far apart, move the mouse faster and run it again.
Your device changes the measurement
None of these tests measure a person in isolation. They measure a person operating a specific piece of hardware, and swapping the hardware moves the number more than most people expect.
| Device | Compared with a mouse | What to do about it |
|---|---|---|
| Mouse | the reference device | A physical switch with a short travel and a defined actuation point. Every benchmark on this site assumes a mouse. |
| Trackpad | typically well below a mouse | A trackpad click has longer travel and a firmer press, and most trackpads reject presses that arrive too close together as accidental. A trackpad score is not comparable with a mouse score. |
| Touchscreen | a different measurement entirely | Taps are sampled at the screen's touch rate and multi-finger tapping changes the ceiling completely. Use the tap speed test instead of reading a CPS score as if it were a mouse result. |
On a touchscreen, run the tap speed test instead. It is built for taps and it will not invite you to compare a two-thumb result against a mouse benchmark, which is a comparison with no meaning.
Where the benchmark numbers come from
Most people land between 4 and 6 clicks per second on a 5 or 10 second test using ordinary clicking.
There is no published dataset on human click speed, so this range is a community convention rather than a measured average. ClickStorm states it once, here, so every page on the site agrees.
Every benchmark figure on this site is defined exactly once, in a single file, and every page reads it from there. That is a deliberate fix: before August 2026 this site gave four different answers to its own core question, ranging from 3 to 6 up to 6 to 8 depending on which page you landed on. It now gives one answer, and it labels that answer an estimate, because an estimate you can see is an estimate is worth more than a confident number you cannot check.
ClickStorm publishes no measured click-speed dataset, and will not claim one until it has collected the data honestly and can state a sample size and a measurement period alongside it. Nothing is uploaded from your browser today, so there is nothing to publish today.
Privacy, unchanged
All of the above happens in your tab. Scores and personal bests are written to localStorage on your own device and no test result is ever sent anywhere. There is no account, no upload and no server-side scoring, which also means there is nothing for ClickStorm to leak, sell or lose.