Scroll test

Counts every notch in both directions and catches the backward jump that gives away a worn encoder.

Put the cursor inside the box and spin the wheel up and down. Try tilting it, or holding Shift, and click the wheel too.

Hover here and spin the wheel

Up 0

Down 0

Horizontal 0

Wheel click 0

0% of the test done

Backward jumps: 0
Peak per second: 0
Wheel type: No data yet
Delta per notch: —
Unit your browser reports: —

The notch count is a real measurement. The "delta per notch" figure is whatever your browser reports, in its own units, and is not comparable between browsers: the same mouse gives 100 in one and 3 in another. That is why you will not see "lines per notch" converted into anything here.

Generated right here, in your browser. What you type never leaves your device.

Keep going with Mouse tester Touchpad test Click speed test Touchscreen test

The wheel is the most-used part of a mouse and the one that ages worst. It rarely dies outright — it starts jumping. You are scrolling down a page and it hops up a line, or you turn three notches and the document moves two. Because it happens fast and in the middle of doing something else, most people take months to be sure they are not imagining it.

This test puts those symptoms in writing. It counts notches up and down separately, and flags backward jumps: a notch that opposes the ones immediately before and after it, while you were turning one way. That is the thing your hand cannot do by accident, and it is the encoder’s signature.

It also checks what almost nobody looks at until it fails: horizontal scroll, if your mouse has it, and the wheel click, which is a separate micro-switch and breaks on its own schedule.

Manual

How to use it

1
Put the cursor inside the box

The test only listens in there, and in there it blocks the page from scrolling so it does not run away from you while you turn.
2
Turn slowly downward, about twenty notches

Slowly, and one way only: that is the only way a backward jump means anything. Spin fast and change direction and there is no telling the fault from your own hand.
3
Repeat upward

Both directions use the same encoder but are read differently, and some wheels fail in one direction only.
4
Try horizontal scroll

Tilt the wheel sideways if your mouse allows it, or hold Shift while turning, which is how mice without tilt do the same job.
5
Click the wheel

Press it straight down until it clicks. It is an independent switch and fairly prone to going stiff or stopping altogether.
6
Read the backward jumps

Zero after a few dozen notches is a healthy wheel. One stray one might be your hand; several on every sweep is not.

In depth

Things worth keeping in mind

How to read the numbers

Four figures, each answering a different question. It is worth knowing which looks at what before drawing conclusions.

Notches recorded: the running total. A real measurement, comparable in any browser.
Backward jumps: the one that matters. Zero is healthy; several per sweep is a worn encoder.
Peak per second: notches in the densest second. Depends on how fast you spin, not on mouse quality.
Delta per notch: whatever your browser reports, in its own units. Only good for comparing against yourself.
Wheel type: notched if the delta is constant; free-spin or touch if it arrives as a decaying ramp.

History

The wheel took thirty years to arrive

The mouse dates from 1964 and the wheel did not reach the mainstream until 1996, with the Microsoft IntelliMouse. Thirty-two years in which moving through a long document meant dragging the scrollbar with the cursor, or hitting the page keys over and over.

It was not the first attempt. Mouse Systems had sold the ProAgio with a similar wheel in 1995, and there were earlier prototypes at Xerox. What Microsoft did differently was to build support into Windows and Office at the same time as it sold the hardware, so the wheel did something useful from day one in the programs people already had open.

The IntelliMouse wheel was notched: it turned in steps and it made a sound. That was not styling, it was information. The click and the resistance tell your hand how many steps it has taken without looking at the screen, which is exactly what is lost on the modern free-spin wheels that turn smoothly and keep coasting.

The wheel click came as a by-product, as these things do: if there is already a shaft that can be pushed down, putting a switch under it is nearly free. It became the middle button that opens links in new tabs — a function nobody designed and that is now impossible to give up.

1996: The IntelliMouse popularises the wheel
32: Years between the first mouse and the wheel
2: Contacts deducing direction, in quadrature
3: Different units a browser may report in

How it works

What is inside a mouse wheel

Under the wheel sits a rotary encoder: a notched disc with two contacts brushing against it. As it turns, each contact opens and closes in turn, and the difference between the two gives the direction. Two signals, one trailing the other: that is the whole mechanism, and it is called quadrature.

Direction does not come from one signal but from which of the two changes first. If contact A changes before B you are going down; if B changes before A you are going up. That is why a dirty wheel does not fail gently: when a worn contact bounces, the order inverts for an instant and the computer reads one notch the wrong way. That is precisely the backward jump counted here.

Nearly every mouse encoder is mechanical, with contacts that physically rub, because they are cheap. They pick up finger grease and dust and they wear. Mice that never develop this problem usually carry an optical encoder, which reads the notches with a beam of light and touches nothing.

The browser, for its part, hands you none of that raw. It gives you one event per notch with a delta number, and that number is not normalised: the same mouse reports 100 in one browser, 3 in another and 53 in a third, because each decides for itself whether it is speaking to you in pixels, lines or pages. That is why the notch count is a real measurement and the delta figure cannot be compared across browsers.

By hand

Checking the wheel with no page at all

If you would rather not trust anything running in a browser, the same diagnosis works with a long document.

1

Anything running to many pages. Set it at a size where individual lines are easy to tell apart.
2

One near the top edge of the window. You are going to follow it with your eyes, not watch the scrollbar.
3

One only, then look at where your word landed. Repeat twenty times, counting out loud.
4

If at any point the word moves up instead of down without you changing direction, there is the fault. Once in twenty is already enough to annoy you every day.

Diagnosis

If something is not working

Symptom:The wheel jumps backward while I scroll down

Likely cause:: The encoder contacts are dirty or worn, and the order of the two signals inverts for an instant.
What to do:: Blow air into the wheel slot and spin it through several full turns so the contacts wipe themselves. If it comes back within days, the contact is worn past cleaning: the permanent fix is replacing the encoder, a cheap part that needs soldering, or replacing the mouse.

Symptom:The wheel turns but the page does not move

Likely cause:: Almost never the mouse: usually the application, which is not accepting the event where your cursor happens to be.
What to do:: Check here whether the notches are counted. If they are, the mouse is fine and the program is the problem: try clicking inside the region you want to scroll before turning.

Symptom:The wheel click does nothing

Likely cause:: It is a switch independent of the turning, so it can fail on its own. It may also have been remapped by the vendor software.
What to do:: Check the mouse software to see whether the middle button has been given another function — a very common remap. If it is at its default and nothing lights up here, the switch has failed.

Symptom:Horizontal scroll does nothing

Likely cause:: Many mice do not have it, and of those that do, some send it as a keyboard shortcut rather than as scroll.
What to do:: Try holding Shift while turning the wheel: if that registers, your mouse has no tilt but the system supports the gesture. If your mouse does tilt and nothing appears here, check the vendor software for how the tilts are mapped.

Symptom:Each notch moves too much or too little

Likely cause:: Not the wheel: it is how many lines the system decides to move per notch.
What to do:: It lives in the operating system’s mouse settings, usually as "lines per scroll" or similar. Change it there rather than on the mouse. Watch out on free-spin mice: with the brake released the wheel keeps coasting and it will look like it is moving too far.

Symptom:The wheel feels gritty or scrapes as it turns

Likely cause:: Dirt between the wheel and the shell, or a dry shaft.
What to do:: Compressed air into the slot while spinning the wheel. Do not introduce liquid: if it reaches the encoder, the problem goes from gritty to intermittent.

Trivia

Things worth knowing

Direction is not told by a signal, it is told by the order

The wheel has two contacts out of phase. Which of the two changes first is the only thing separating up from down. That is why a dirty contact does not make the wheel slower: it makes it reverse for an instant.

The delta number does not mean the same thing in each browser

The browser event carries a field saying whether it is speaking in pixels, lines or pages. The same mouse on the same machine reports 100 in one and 3 in another. Any page telling you "your mouse does X lines per notch" is half inventing it.

The middle button was invented by the hardware, not the software

Nobody designed the wheel button: it appeared because there was already a shaft that could be pressed and a switch underneath cost almost nothing. Opening links in new tabs, its commonest use today, arrived years after the button did.

Notches are information, not decoration

The click of a detented wheel tells your hand how many steps it took without looking. Free-spin wheels are faster across enormous documents and worse for moving precisely, which is why many expensive mice carry both and a button to switch.

Inverted scrolling is a 2011 argument

Apple flipped the default in Mac OS X Lion so the screen moved like paper under your finger, as on a phone. It called it natural scrolling. Fifteen years on it is still the first checkbox many people change on a new Mac.

Questions

Frequently asked questions

How many backward jumps are normal?

Zero. One stray jump across a long sweep might be your hand reversing for an instant without noticing, so a single one condemns nothing. But if several appear every time you make a twenty or thirty notch sweep in one direction, a hand does not do that: the encoder does.

Can a jumping wheel be fixed?

Often, and it costs nothing to try. The commonest cause is dirt between the contacts, and blowing compressed air into the slot while spinning the wheel through several turns clears many cases. If it returns within a few weeks the contact is worn and no amount of cleaning will save it: the encoder needs replacing, which is cheap as a part but means opening the mouse and soldering.

Why is the delta per notch no use for comparison?

Because it is not normalised. The browser gives you the delta and separately tells you which unit it is speaking in: pixels, lines or pages. Each browser and each system chooses differently, so the same mouse may report 100, 3 or 53. The notch count is comparable, because a notch is a notch anywhere; the delta figure is only good for comparing against yourself in the same browser.

Does it work with a laptop trackpad?

Yes, and you will see the difference immediately: a trackpad sends many small events that taper off as you lift your finger, and the test will classify it as free-spin or touch rather than notched. What does not make sense on a trackpad is the backward-jump count, because the inertia gesture changes value naturally as it decays.

Does it work on a free-spin wheel, the kind with no notches?

Yes, but read the numbers carefully. A free-spin wheel keeps turning after you let go, so the peak per second will be enormous and says nothing about your hand. To hunt for jumps, engage the wheel brake if your mouse has one and turn slowly; without notches it is much harder to be certain.

Can this test measure polling rate?

No. Browsers batch events to the display refresh rate, so there is no way to tell whether your mouse reports at 125, 500 or 1000 Hz. The peak per second here counts how many notches arrived in the densest second, which is a different thing and depends mostly on how fast you spin.

Is anything sent to a server?

No. Everything this page does happens in your browser and disappears when you close it or press Reset. There is no account, no upload, and no record of what you scrolled.

How is this different from the mouse tester?

The mouse tester covers the five buttons, movement and phantom double clicking, and looks at the wheel only in passing: it counts steps and little else. This page is the wheel alone, which is why it reaches things that do not fit there: the middle click, horizontal scroll, telling a notched wheel from a free-spin one, and the unit your browser reports in. If you suspect a button, start with the mouse tester.