Mouse wheel test
A page that jumps back up while you are scrolling down is not the page and not the browser. It is one notch of the wheel reporting the wrong direction, and it is countable: scroll steadily one way, and every event that arrives with the opposite sign is the encoder getting it wrong. A healthy wheel produces none at all.
Scroll inside the box, steadily and in one direction. Forty notches each way is enough.
| Direction | Notches | Went the other way | Step | Between notches |
|---|
- Verdict—
This looks like a trackpad rather than a wheel: the events arrive as a continuous stream of small steps instead of one burst per notch. A trackpad has no encoder to wear out, and its momentum scrolling reverses on purpose at the end of a flick, so the count above does not mean anything here.
Did any notch go the wrong way?
Running it
Scroll inside the box, steadily, in one direction. Forty notches is enough — the counter shows how many have arrived. Then do the same upwards.
The strip along the bottom draws one bar per notch. Bars going with the run are dark; any notch that reported the opposite direction is red. That single red bar in a row of dark ones is the fault, and seeing it is more convincing than being told a number.
Scroll at the speed you normally do. A wheel that is beginning to fail is often clean when moved slowly and wrong when moved quickly, so a careful, unnaturally slow scroll can pass a mouse that fails in use.
Reading the result
No notches the wrong way is the only clean result. This is not a measurement with a tolerance: a working encoder does not report backwards, ever, and there is no acceptable rate.
One stray notch is worth repeating rather than acting on. A hand can change direction without meaning to, particularly at the end of a run, and one event is not enough to separate that from the hardware.
Three or more is the encoder. It will not recover.
The step and between notches figures are context rather than verdict. They tell you how far one notch moves the page and how fast you were scrolling, which is what makes the reversal count meaningful — forty notches delivered over four seconds is a real test, and forty delivered in half a second is a flick.
Windows 11
- Settings → Bluetooth & devices → Mouse sets how many lines a notch scrolls. It changes the step figure here and nothing else; it cannot cause or cure a reversal.
- Some manufacturer utilities offer a free-spin or ratchet mode. In free-spin the wheel disengages from the notches entirely, and the readings here become meaningless — switch to ratchet mode before testing.
- If scrolling is erratic in one application only, it is that application. The fault this page looks for is in the hardware and appears everywhere, including here.
macOS
- macOS applies momentum to a wheel as well as a trackpad on some configurations, which adds events after you stop and can look like a stream rather than notches. If the test says it thinks you are on a trackpad but you are not, that is why — scroll in short deliberate bursts and read the bars rather than the counts.
- Natural scrolling inverts the direction. It does not affect this test, which works out which way each run is going rather than assuming.
When it is not the wheel
A page that jumps as you scroll is not always the mouse. Content loading in above where you are reading pushes everything down; smooth scrolling can overshoot and settle back; some extensions replace the wheel handler with their own. All three feel like a wheel fault and none of them are.
That is precisely what makes this test worth running before buying anything: it separates the mouse from everything else standing between it and the page.
Common questions
What is actually broken when the wheel jumps back?
The encoder. Inside the wheel is a slotted disc turning past a sensor — optical on better mice, a set of mechanical contacts on cheaper ones — and it reports each notch along with which way the disc moved. Dust in the slots, or worn contacts, make it misread the direction of a single notch. The computer receives a perfectly valid scroll event pointing the wrong way, which is why no driver setting and no browser fixes it.
Why does it happen more when I scroll fast?
Because the encoder has less time to resolve each notch, and a marginal reading becomes a wrong one. This is also why the fault appears in long documents and long web pages first: that is where people scroll fastest. If a slow, deliberate scroll is clean and a fast one is not, the encoder is on its way out rather than already gone.
I use a trackpad. Does this test mean anything?
No, and the test says so when it detects one. A trackpad has no wheel and no encoder, so there is nothing here to wear out. It also reverses direction deliberately: momentum scrolling bounces at the end of a flick, which this would count as the fault it is looking for. The test is for a wheel.
What is the step number?
How far one notch scrolls, in whatever unit your browser reports. Chrome and Safari usually give pixels, Firefox usually gives lines, so the numbers are not comparable between browsers — but they should be consistent within one run. Wildly varying steps in the same direction point at the same worn encoder, arriving as missed and doubled notches rather than as reversals.
Can I fix it without opening the mouse?
Sometimes. Compressed air into the gap around the wheel, from both sides, clears the dust that causes a large share of this. Do it with the mouse upside down so what you loosen falls out rather than further in. It is worth trying before anything else because it takes two minutes and costs nothing.
And if that does not work?
The encoder is a small soldered component, widely available, and replacing it is a well-documented repair on any mouse with a following. Whether it is worth doing is a question about the mouse rather than the fault: on something with a good sensor and a comfortable shell, yes. On a mouse that came free with a keyboard, no.