How much of your reaction time is your computer?

If you take a reaction test on a phone and then on a desktop, you will probably see a gap of 50 milliseconds or more. Almost none of that is you.

Where the milliseconds actually go

A click test measures a round trip: something changes on screen, light reaches your eye, you decide, your hand moves, a switch closes, the operating system notices, the browser fires an event. Your nervous system is responsible for the middle. The machine owns both ends.

The two large, controllable ones are the display and the input device.

Why refresh rate matters more than it sounds

A 60Hz screen redraws every 16.7 milliseconds. When the page decides to turn green, that change waits for the next frame — on average about 8 milliseconds, and up to 16.

That is a floor you cannot get under, and it applies before anything else. A 144Hz display cuts the wait to under 7 milliseconds on average. Between two otherwise identical people, the one on the faster panel scores better, and the difference is roughly the distance between “average” and “good”.

Display lag is separate, and larger

Refresh rate is how often the panel updates. Display lag is how long the panel takes to process a signal before showing it at all, and it varies far more between models than refresh rate does.

Televisions are the worst offenders — many add 20 to 40 milliseconds in their default picture mode, because the image processing that makes films look good takes time. Most have a “game mode” that bypasses it. If you have ever felt that a game was sluggish on a TV and fine on a monitor, that is what you were feeling.

Touch is a different measurement entirely

A touchscreen is not a slower mouse; it is a different instrument.

The screen samples touches at its own rate, usually slower than a mouse polls. It has to distinguish a tap from the start of a scroll, which takes time. And your finger covers the target it is reaching for, so the visual confirmation arrives differently.

The practical consequence is simple: do not compare a phone score against a desktop one. Our aim trainer records which input you used alongside the result for exactly this reason, so the two do not end up averaged into a number that describes neither.

Wired against wireless

Modern wireless mice are much better than their reputation, and a good 2.4GHz dongle is close to indistinguishable from a cable. Bluetooth is measurably slower, often by a few milliseconds, and it is more variable — which matters more than the average, because variability widens the spread of your results rather than shifting them.

If your history looks noisier than it used to and nothing else changed, a nearly flat mouse battery is a genuine suspect.

What you can reasonably conclude from a browser test

Given everything above, it is fair to ask what the number is good for at all.

It supports within-person, same-equipment comparisons. That covers most of what anyone actually wants to know: whether you are slower today, whether sleep helped, whether the new mouse made a difference.

It does not support between-person comparisons, because the hardware term is large, uncontrolled and unknown. A leaderboard across mixed equipment ranks equipment.

It does not support absolute claims about you. “My reaction time is 240ms” is really “my reaction time plus this laptop and this mouse is 240ms”, and the second half of that sentence is worth 30 to 40 of the milliseconds.

Every design decision on the reaction test follows from those three lines: a median over five rounds rather than a single number, your own history in front of you, and no global rank to compare against.

Why browser timing is less precise than it looks

Even with perfect hardware, the timestamps a web page can read are not exact.

Browsers deliberately reduce the precision of their high-resolution timers and add small random jitter, as a defence against timing-based attacks that could otherwise read data across site boundaries. The exact treatment varies by browser and by whether the page is cross-origin isolated.

For a reaction test the effect is small — a millisecond or two — and it is noise rather than bias, so it widens the spread slightly without shifting the median. It is worth knowing because it is another reason a single round means little and a median of five means more.

Background tabs are the bigger risk. Browsers throttle timers in tabs that are not visible, sometimes to once a second. Any test measuring elapsed time in a hidden tab is unreliable, which is why anything timing-critical here runs only while the page is in front of you.

How to make a fair test

You cannot remove the hardware from the measurement. You can hold it still.

1. Same device, every time. This is the one that matters. Changing laptop resets your baseline entirely and any comparison across the change is meaningless.
2. Same input device. Not the trackpad one day and the mouse the next.
3. Same time of day, roughly. Most people are slower in the early afternoon.
4. Close what is running. A busy machine adds scheduling delay, and the effect is irregular rather than constant.
5. Ignore your first few sessions. Early gains are you learning the task.

Do that and the remaining variation is mostly you, which is the only thing worth tracking.

What this means for comparing against other people

It means you largely cannot, and any site presenting a global average as though it were a level playing field is overstating what it knows. A table of percentiles built from mixed hardware is measuring the distribution of equipment at least as much as the distribution of people.

That is why the reaction test leads with your own median and your own history rather than a rank, and why the daily round scores every test against a fixed range rather than against other players. Neither choice would matter if the measurement were clean. It is not, and pretending otherwise would make the numbers less useful rather than more.

Two things worth changing, if you care

Most of the hardware contribution is not worth spending money on. Two parts are.

A wired or 2.4GHz mouse rather than Bluetooth. Not because the average is much better, but because it is much more consistent. Variability widens the spread of your results, and a wide spread is what makes a trend hard to see. A nearly flat battery does the same thing and is a genuine suspect when your history suddenly gets noisier.

Game mode on a television. If you are using a TV as a monitor, the default picture processing can add 20 to 40 milliseconds, and the setting that turns it off is usually one menu deep. This is the single largest free improvement available to anyone it applies to.

Beyond those, a faster panel helps and costs money, and everything else is small enough to be lost in your own night-to-night variation.

The one comparison that survives

You against yourself, on unchanged equipment, over a fortnight.

That is enough to see genuinely useful things — what a short night costs you, whether your afternoon slump is real, how much a week of proper sleep gives back. Those effects are large enough to clear the noise your hardware introduces, which is precisely why they are the things worth looking for.