Jitter or butterfly: which is fastest?
Jitter clicking tenses the arm so the finger vibrates on the button; butterfly alternates two fingers on one button. Normal clicking is one finger, one press at a time.
We work the technique out from the run itself — the rhythm between clicks and how far they scatter across the pad — rather than asking, because an answer nobody can check is not worth publishing. Runs we cannot read confidently are counted in none of these.
Click speed records by test length
CPS is a rate, so a one-second burst and a ten-second run are not the same test. Almost everyone is fastest over one second and slows down as the clock runs — which is why a burst record and an endurance record belong in separate columns, and why one leaderboard ranks everybody on rate rather than splitting them by clock.
Each figure is the fastest run we have recorded at that duration, and the average of every player's best run at it. Multiply by the seconds for a click total: 10 CPS over 10 seconds is 100 clicks.
What is a good CPS?
Anything above the ~6–7 CPS average counts as good. The top 10% of players reach 13.9 CPS or higher, and the top 1% reach 13.9 CPS. The current record is 13.9 CPS.
How these numbers are measured
Every figure on this page is computed from runs taken on this site. Your score is the clicks you landed divided by the seconds you had — no smoothing, no rounding in your favour, and no adjustment for the device you used. The timer starts on your first click rather than when the page loads, so reaction time is never counted against anyone.
Ranked runs are validated server-side before they reach the board, and a rate no hand produces is rejected. That check is the only reason these averages mean anything: an average that includes automated runs measures software, not people.
Why these differ from the numbers you see elsewhere
Most click-speed pages quote one global figure with no clock attached, no sample size and no date. That is not a measurement, it is folklore that gets copied between sites until it sounds official. The most-repeated example is the Guinness “most mouse clicks in one minute” record of 760 clicks, about 12.7 CPS, which is a sixty-second sustained average — quoted next to one-second burst scores as though the two were comparable.
The figures here are split by the things that actually change the answer: how long the run was, how it was clicked, and whether it was a touchscreen or a mouse. Each carries the number of runs behind it, and any bucket below its sample floor says “not enough runs yet” instead of showing a number computed from a handful of attempts.
How to read the percentiles
The median (8.7 CPS) is the middle of the board — half the players measured here are faster and half are slower, which makes it a better description of a typical player than the average, since a handful of very fast runs pull an average upward. The 90th percentile (13.9 CPS) is the line the top tenth clear, and the 99th (13.9 CPS) the top hundredth.
So a score between the median and the 90th percentile is genuinely above average rather than merely above the mean, and the gap between the 90th and the 99th is a useful measure of how steep the top of the range is.
What this data does not tell you
It describes the people who came here and clicked, not the general population. Anyone taking a click-speed test has selected themselves into a group that cares about the result, so these figures skew faster than a random sample of people using a mouse would.
- Anonymous practice runs are not ranked and are not counted here — only verified runs reach the board these numbers are computed from.
- Input type and technique were not recorded from the beginning, so those breakdowns cover fewer runs than the overall totals. Each one states its own count rather than implying it covers everything.
- Technique is inferred from the rhythm of a run, not asked. A run the classifier cannot read confidently belongs to no technique rather than being guessed into one.
- A browser measures every press your mouse sends. A game does not — it applies its own tick and cooldown — so a score here is an upper bound on what a game will register, never a prediction of it.
These limits are published because a dataset is only usable as a source if you know what it excludes. Numbers presented without them are the ones worth trusting least.
Want these numbers in your own page or paper? They are available as a free JSON API — every figure with its sample size and the date it last changed, free to use with attribution.