Skip to content
clickidy
🎯 ALL GAMES
REFLEX · SPACEBAR

SPACEBAR
COUNTER

Hammer the spacebar for five seconds. We count every press and work out your presses per second - the biggest key on the board, tested properly.

SPACEBAR COUNTER · 5S RUN · AD-FREE ZONE
00PRESSES
5.0 SECONDS LEFT
PRESS SPACE TO START
3·2·1·GO - THEN HAMMER IT FOR 5 SECONDS (TAP WORKS TOO)

HOW THE COUNT WORKS ◢

Five seconds, one key, no downloads

  1. Press space once to start - you get a 3·2·1·GO countdown.
  2. Hammer the spacebar until the timer hits zero. Only real presses count.
  3. Read your presses per second, then go again - your best is kept on this device.

WHY A SPACEBAR IS NOT A MOUSE BUTTON ◢

A mouse switch has almost no travel - a click is a millimetre and a snap. A spacebar is a full key: it has to move down far enough to reach its actuation point, the depth at which the switch actually registers, and then rebound above that point before it can fire again. On a typical mechanical switch actuation sits around 2mm into a 4mm press. Your thumb is still the binding constraint - the clinical tapping norms put a fast adult at four to five presses a second on a counter with no travel at all - but the key is moving too, and that is one reason spacebar numbers tend to land a little under the same person’s click speed.

WHEN THE KEYBOARD DROPS YOUR PRESSES ◢

If your count feels lower than your thumb, the board may be losing presses. Two usual suspects. Rollover: cheaper membrane keyboards share wiring between keys, and with three keys held on three corners of the wiring grid the board either invents a fourth (ghosting) or suppresses one to avoid it (blocking, which is the one that feels like a dropped press). N-key rollover fixes the wiring, though over USB the standard keyboard report still carries only six ordinary keys at once. Debounce: keyboards ignore a short window around each press so that one physical press isn’t read as several - a few milliseconds, and some firmware waits for the signal to settle rather than reporting immediately. Either way it is far above anything a thumb can reach, so it is not what caps your score.

A STICKING OR CHATTERING SPACEBAR ◢

The spacebar is the most-hit and most-crumbed key on the board, and it’s stabilised at both ends, so it fails in ways other keys don’t. If presses feel mushy or the bar returns unevenly, that’s usually debris or a stabiliser rather than the switch. The opposite fault is chatter - a worn switch registering two presses from one push, which shows up here as a count that’s higher than you actually pressed. Either way, this test tells you which one you have.

HELD KEYS DON’T COUNT ◢

Hold the spacebar down and your operating system starts auto-repeating it, firing key events at a fixed rate for as long as you lean on it. We ignore every one of those, so the only way to move the number is to actually lift and press. The percentile is a model anchored on the clinical tapping norms, not a live scoreboard, and those norms measured a finger rather than a thumb - the workings are in how we count.

THE LONG ANSWER ◢

Four to five times a second, if you are an ordinary adult going flat out, and a shade over six if you are at the top of a published normative sample. That is not a guess or a gaming statistic, but one of the oldest measurements in clinical neuropsychology, taken with a mechanical counter and a stopwatch long before anyone owned a keyboard.

Underneath it sits a second ceiling, and that one belongs to the hardware. Switch contacts settle somewhere between one and five milliseconds, a controller waits a few more to be sure, and the USB bus asks the keyboard for news somewhere between a hundred and a thousand times a second. Stacked together, those still allow well over ten times the rate a thumb can produce.

So the binding constraint is your hand and the key it is moving, not the electronics underneath. What is worth knowing is what has to go wrong for the other ceiling to become the one you hit.

Key Takeaways

  • Published finger-tapping norms put adults aged 18 to 40 at roughly 4.4 to 5.3 taps a second on the preferred hand, and the whole table runs from 4.0 upward. The fastest single person in that sample averaged 6.6.
  • The rate is not fixed even within one attempt. Tapping at maximum speed slows measurably within the first half-minute, and the cause is largely in the brain rather than the muscle.
  • A keyboard's own limits sit far above that: Cherry rates its switches at one to five milliseconds of bounce depending on the part, and the USB specification allows a full-speed device to be polled every millisecond.
  • The ways a keyboard genuinely loses a press are specific and named: a sneak current path across the scan matrix, a debounce window that swallows a transition, and a poll interval that both the press and the release fall inside.
  • The spacebar is struck far from its own plunger, which is why it carries a wire rod that a keytop the width of its own housing does not need.

What a single finger can actually do, and who measured it

Clinicians have a standard instrument for this, the Finger Tapping Test, built into the Halstead battery in the 1940s and still administered essentially unchanged. One published description of the battery gives the whole protocol in a line: subjects "tap as rapidly as possible using the index finger on a specially adapted tapper for five 10-second trials", and those trials are averaged.

Because that has been stable for eighty years, the normative tables are unusually good. Fromm-Auch and Yeudall published one in 1983 covering 193 volunteers with no psychiatric or neurological history. Their Table 8 is the closest thing there is to an answer for a healthy adult. Every figure below is built from subjects' averages across the five trials, which is what the table reports.

Group (preferred hand) Mean taps per 10-second trial Taps per second Range of individual averages
Males 18 to 23 49.5 4.95 26.6 to 64.6
Males 24 to 32 50.6 5.06 38.2 to 66.2
Males 33 to 40 53.4 5.34 39.0 to 61.0
Females 18 to 23 43.6 4.36 30.6 to 65.6
Females 24 to 32 45.2 4.52 31.0 to 60.0
Females 33 to 40 45.8 4.58 40.6 to 55.6

Three things fall out of that table. Performance peaks in the thirties rather than the twenties, a relationship the authors describe as curvilinear. Males average about five taps more per trial than females. And the individual spread is enormous: the group means sit near five a second, but one subject averaged 66.2 taps a trial and another averaged 26.6. Two age bands are left out of the table above for space, the teenagers and the 41-to-64s, and they pull the bottom of the group means down to 40.4 taps a trial.

Modern work lands in the same neighbourhood by a different route. A 2024 study of 370 people, run on a phone or tablet touchscreen rather than in a laboratory, reported an average of 5.2 Hz for its older group and faster tapping in the young, with the index finger fastest of those tested and the little finger slowest. That last finding matters here, because the norms above are index-finger numbers and a spacebar is usually a thumb. No open normative table for thumb tapping on a full-travel key appears to exist, so treat all of these as the closest available proxy rather than as a thumb measurement.

Why the rate falls away before the clock does

Maximum tapping is not a steady state. It decays, fast enough to see inside a single short trial.

Researchers call that decay motor slowing, and the 2024 study measured it directly. A 2026 experiment in the same journal put the drop from the first ten seconds to the last ten seconds of a thirty-second bout at 17.2 percent in its post-crossover control condition, on a task that alternated the index and middle finger rather than hammering one key. So a five-second window and a thirty-second window are not measuring the same thing.

Why it decays is the surprising part. Fast, low-force tapping barely touches the muscle: the 2024 authors note that contractile properties show no impairment before and after such a task, and that maximal voluntary force is unchanged by it. Slowing is attributed instead to a gradual release of inhibition in the sensorimotor cortex. Your finger is not tired. What reaches it is degrading.

Which is also why a keyboard rate, a click speed test and a reaction time test disagree with each other so often. Three different limits, and being good at one predicts very little about the others.

Ghosting, blocking, and the grid of wires under the spacebar

Nearly every keyboard reads its keys as a grid. Rows and columns of conductor cross without touching, a switch sits at each crossing, and a controller drives one line at a time and reads back which crossing lines went with it. Cheap, and saddled with a well-documented failure.

Maxi-Switch set out the geometry exactly in a patent filed in 1983. With several keys down at once, current can take a path through them that the controller cannot tell apart from a genuine closure: "This condition can occur at any site in the matrix where three keys down are located on three corners of a square or rectangle in the matrix", and "The key at the fourth corner of the square or rectangle is the potential pseudo key." Those three real keys need not be near each other. That phantom fourth key is what enthusiasts call ghosting.

Fixing it takes a diode in series with each switch, stopping current flowing backwards through a closed contact. Cherry sells the same MX switch with and without one, listing a decoupling diode as a per-part-number option in the models table on its own datasheet. Wherever it sits, inside the switch or as a discrete part on the board, it has to exist in hardware; no amount of firmware substitutes for it.

What most people actually experience is not the phantom key but its suppression. A 2007 Microsoft patent application describes the standard behaviour, which is that "current algorithms suppress reporting the current state of key depression until the potential ghost condition disappears". Detecting that it might be about to invent a key, the board reports nothing until the ambiguity clears. That is blocking, and from the player's side it looks exactly like a dropped press.

Six keys at once, and why that number is written into a standard

A second, entirely separate rollover limit lives in the USB specification rather than in the wiring.

Version 1.11 of the Device Class Definition for Human Interface Devices defines the keyboard report as an array of key codes, and its Appendix C is blunt about what happens when that array runs out: "The keyboard must report a phantom state indexing Usage(ErrorRollOver) in all array fields whenever the number of keys pressed exceeds the Report Count. The limit is six non-modifier keys when using the keyboard descriptor in Appendix B." Modifiers such as shift and control sit in a separate byte and do not count against it. Meanwhile the boot protocol that your firmware talks to before an operating system loads makes the same six-key cap mandatory, and demands the same rollover error whenever a key combination cannot be accurately determined, naming ghost keys explicitly.

Keep that separate from the matrix question, because the two get conflated. A keyboard advertising N-key rollover has solved the wiring problem with diodes, but it has not repealed the six-key array; it works around it by describing a different report format to the operating system, and still falls back to six in the boot protocol. None of this binds a spacebar test, where you press one key. It binds the moment a game asks you to hold three.

Contact bounce, and the window that both hides and doubles a press

A metal contact does not close cleanly. It makes, breaks and remakes for a short interval before settling, and a controller that believed every transition would report one press as several.

Manufacturers specify that interval directly, and Cherry states it in more than one way across its own range. Some sheets give a bounce time of typically under 1 ms. Others give under 5 ms with a test condition attached, during actuation at 0.4 m/s, among them the Nature White and the MX Speed Silver, both revised in 2017.

Newer parts are not simply better here, and the clearest evidence sits inside a single PDF. Cherry's MX Black document carries two datasheets. The first, dated 17 December 2015, covers the SMD and RGB variant and gives typically under 1 ms. The second, dated 16 February 2016, covers the original gold-crosspoint key module and gives under 5 ms during actuation at 0.4 m/s. Both are a black switch at 60 cN. Cherry's current datasheet for the MX2A Red, a later switch generation entirely, states the bounce time flatly as 5 ms with no condition and no qualifier.

So the honest range is roughly one to five milliseconds, and how it is written down depends on the document as much as on the part.

Travel belongs in the same discussion, because a key cannot fire twice without resetting. Cherry gives that 2015 MX Black an actuator travel of 4.0 mm with a pretravel of 2 mm, so the switch registers halfway down and the cap has to climb back to at or above that depth before the next press counts. No return-time figure appears on any Cherry datasheet read for this article, so the ceiling that geometry imposes cannot be calculated from published data at all. Bounding it from the other side is all that is available, and the bound is loose: at the five-a-second the norms describe, the cap has 200 milliseconds to make a round trip of a few millimetres, which is ample.

Firmware then adds its own margin, and the shape of that margin is where both failure modes come from. QMK, the dominant open-source keyboard firmware, states that "Default debounce time is 5 milliseconds" and offers two families of algorithm. Deferred debouncing will "wait for no changes for DEBOUNCE ms before reporting change". Eager debouncing is the mirror image: "any key change is reported immediately. All further inputs for DEBOUNCE ms are ignored".

They fail in opposite directions. Deferring delays every press and, if the contact never stays still for a full interval, reports nothing at all. Eagerness is documented as not noise-resistant, so a worn switch whose chatter falls outside the window produces a second press that your thumb never made. QMK adds that "The time it takes for the switch to settle might vary with switch type, age, and even pressing technique", which is why a keyboard that was fine for five years can start double-counting.

Either way the arithmetic is generous. Apply a 5 ms window to each transition and a full press and release still fits into 10 ms, which is 100 a second, about fifteen times the fastest human in the norm table.

What a polling interval costs, in milliseconds

Last comes the bus, which is the stage gamers argue about. A USB keyboard does not interrupt the computer when you press a key. Instead the host asks, on a schedule the device requests when it is plugged in.

Section 5.7.4 of the USB 2.0 specification sets the bounds: "A full-speed endpoint can specify a desired period from 1 ms to 255 ms. Low-speed endpoints are limited to specifying only 10 ms to 255 ms." High-speed endpoints ask instead for a power-of-two multiple of the 125 microsecond microframe. In the HID specification's own example keyboard endpoint descriptor, the requested interval is 0x0A, expressed in milliseconds, which is ten. Whatever the number, the guarantee runs one way only: the same section promises merely that "the host will ensure that the time duration between two transaction attempts with the endpoint will be no longer than the desired period".

So the familiar 125 Hz and 1000 Hz figures are two legal points on that range, an eight-millisecond period and a one-millisecond one. Going faster buys you up to about eight milliseconds on an unlucky press and roughly three and a half on average. Against a press every 200 milliseconds, that is a few percent, and it is a delay rather than a loss.

What can actually lose a press is subtler, and it sits in the HID keyboard requirements: "Reports must contain a list of keys currently pressed and not make/break codes". A USB keyboard sends state, not events. Should a press and its release both fall between two consecutive polls, the two reports are identical and nothing happened as far as the host is concerned. At an eight-millisecond period that needs a press shorter than 8 ms, far outside human range, and the same appendix at least permits a keyboard to buffer events rather than fold them into a single report. So the hole is real. It is not the one costing you presses.

The spacebar is wider than its own plunger, so it needs a rod

Every limit so far applies to any key on the board. The last one is not the same for every key.

Oak Industries stated the problem in one sentence, in a patent filed in 1982: "Because the space bar will be actuated at locations remote from the plunger, means must be provided to prevent the plunger from binding in the housing." A normal keytop, the same patent notes, "is usually about the same width as the housing", so you always strike it directly above the plunger. A space bar spans many key positions, at a standard lead spacing Cherry gives as 19.05 mm, and you hit it wherever your thumb lands.

Its solution is the one still in use. "The torsion rod is an elongated wire which is rotatably mounted in two or more of the hooks", the patent says, and "The torsion rod has a crank arm at each end which is connected to the space bar." Bend the ends through ninety degrees, hook them into the keys of the adjacent row, and the bar stays level whatever part of it you hit.

So the spacebar carries a second mechanical system, as every key wider than its own housing must. Enter, Backspace and the shift keys are stabilised the same way, and the patent's own scope is "support structures for keys having an elongated keytop, such as a space bar on a standard typewriter keyboard". What singles out the spacebar is length. The plunger needs clearance to move at all, and the patent explains what that costs: "While this play is negligible with a short keytop, it is multiplied by the length of a space bar and can become unacceptable." The fast tests hub collects the others, if you want the same hand measured against a different limit.

Frequently asked questions

How many times can a person press the spacebar in one second?

Around four to five times for a typical adult tapping flat out, based on clinical finger-tapping norms that put the preferred hand between 4.0 and 5.3 taps a second across every age band, and between 4.4 and 5.3 for ages 18 to 40. The fastest single person in one published sample of 193 volunteers averaged 66.2 taps per 10-second trial, or 6.6 a second. Those figures are for the index finger on a mechanical counter, so a thumb on a full-travel key is not directly covered by them.

Is my keyboard limiting my spacebar score?

Almost certainly not. A mechanical switch settles in roughly one to five milliseconds, typical firmware adds a five-millisecond debounce window, and USB polls a full-speed device as often as every millisecond. Those combine to allow around 100 press-and-release cycles a second, which is about fifteen times the fastest rate in the published human norms.

Does a 1000 Hz polling rate help on a spacebar test?

Only marginally, and not in the way it is usually described. Moving from an eight-millisecond to a one-millisecond polling interval saves up to roughly eight milliseconds on an unlucky press and about three and a half on average, against a gap of roughly 200 milliseconds between presses at human speed. It shifts when the press is reported, not whether it is counted.

Why does my rate drop off before the timer ends?

Because maximum-speed tapping decays within seconds, an effect researchers call motor slowing. One 2026 experiment measured a 17.2 percent fall from the first ten seconds to the last ten seconds of a thirty-second bout in its post-crossover control condition. The cause appears to be central rather than muscular: contractile properties and maximal voluntary force are unchanged by the task, and the slowing is attributed to a gradual release of inhibition in the sensorimotor cortex.

What is the difference between ghosting and blocking?

Ghosting is a key the keyboard reports that you never pressed, caused by current finding a path across three keys that are down on three corners of a rectangle in the scan matrix. Blocking is the keyboard's defence against it: rather than invent the fourth key, the controller suppresses reporting until the ambiguous condition clears, which reads as dropped presses instead of phantom ones.

Sources