Mouse DPI test from one measured drag

On Windows it is almost always the pointer speed slider, and the ratio tells you which notch it is on. That slider multiplies cursor movement by a fixed factor — 1/32, 1/16, 1/8, 1/4, 3/8, 1, 1.5, 2, 2.5, 3 and 3.5 across its eleven positions — and only the sixth is 1:1. So a 1600 DPI mouse reading 2,400 is sitting on position 7 and one reading 600 is down on position 5, and in both cases the fix is to put the slider back to the middle and measure again rather than to distrust the mouse. macOS has no equivalent notch: its tracking slider bends the curve rather than scaling it, which is why a Mac reading is only trustworthy once acceleration is off entirely.

They are the same quantity and CPI is the correct name for it. A mouse sensor photographs the surface thousands of times a second and reports how many counts of movement have happened since the last report, so its resolution is counts per inch. Dots per inch came from printing, where a dot is a spot of ink on paper and a mouse produces none. Vendors print DPI on the box because that is what people search for, and this page follows the box in its title and the sensor in its readout.

Because both of them change what a pixel is, and the page has to convert. Coordinates in a browser arrive in CSS pixels, and devicePixelRatio is how many display pixels one of those covers — it moves with display scaling and with browser zoom alike, so at 150% a drag that reports 1,066 CSS pixels really crossed 1,600 of them. Windows, X11 and ChromeOS move the pointer in display pixels, so that multiplication is the right one there; macOS and iPadOS move it in points, so on those the unconverted CSS figure is the count and multiplying would double it. Each run stores the ratio it was measured at, and a session where the ratio changed part way through is flagged instead of averaged.

None of them — a spread that wide means something is still scaling the pointer, not that one run was unlucky. Acceleration is the usual cause, because it makes the answer depend on how fast each pull was rather than how far it went, and the cursor test will tell you in one sweep whether a curve is in the path. The other cause is the pad measurement moving: if the mouse started a centimeter short of the mark on one run, that is 10% of a 10 cm drag on its own. Tape the marks down, redo three runs at a crawl, and expect the spread to fall under 3%.

That is the case this page exists for. Office mice ship at 800 or 1,000 CPI and almost never say so anywhere on the device or in the manual, and there is no operating system dialog that reports it — Windows and macOS both show you a speed slider, which is a multiplier applied afterwards, not the sensor's resolution. A drag here gives you the number without installing a vendor utility, and without registering anything: the measurement is arithmetic in the tab, so there is no account and no driver involved.

It can, and it shows up as a reading that comes out low rather than as an obvious failure. An optical sensor that briefly loses the surface stops reporting counts, so the pointer covers less ground than your hand did and the division under-reports. Glass, high-gloss desks, mirror finishes and very dark cloth are the usual offenders, along with a lens with a hair across it. If the number rises when you move to a cloth pad, the pad was the problem; if the pointer also drew visible zigzags while your hand went straight, that is the same fault seen from the other side.

Because a fast drag measures your acceleration curve rather than your sensor. Windows' Enhance pointer precision and macOS' default curve both map speed onto extra distance, so the same 10 cm across the pad produces more pixels when you do it quickly, and the faster the pull the more of the answer is curve. Four inches of pad per second is the ceiling here — 2.5 seconds for a 10 cm drag — which is slow enough that every curve worth worrying about is still in its flat region. A slow pull also gives the mouse time to report the whole path instead of a coarse sample of it.