How we test and research

Derived figures: arithmetic, shown

Most numbers on this site are calculated rather than measured, and for these there is nothing to test: there is only whether the formula is right and whether the arithmetic is correct.

We handle that by showing the formula next to the result on every page and inside every calculator. If you disagree with a cm/360 figure, a frame time, a polling interval or a pixels-per-degree value, you can check it in thirty seconds with a calculator of your own, and we would rather you did.

The calculators themselves are pure client-side JavaScript with the formulas visible in the page output. Nothing about them is proprietary.

Verification by hand measurement

Where a derived figure rests on a constant we did not publish, such as a game's yaw value, the check is a physical measurement: mark the pad, rotate exactly 360 degrees, measure the distance, solve for the constant. We describe this procedure in full on the constants page so that any reader can perform the same verification.

This is the method we use ourselves when a constant is disputed, and it is the method we ask for when someone reports that one of ours is wrong.

Manufacturer-stated specifications

Mouse weights, dimensions, sensor ratings and switch specifications are transcribed from manufacturer documentation. We do not weigh mice on our own scales, and we say so rather than implying otherwise.

What that means for you:

The figures are as accurate as the manufacturer's own measurement and quality control.
Weights sometimes exclude the cable, the dongle, or the grip tape in the box. Where a manufacturer's figure is ambiguous, we use the base configuration and note it.
Individual units vary. A mouse specified at 54 g may weigh 53 or 56.
Where our figure disagrees with a measurement you have taken, please tell us, and include the method.

Third-party measurement

Some quantities require equipment and controlled conditions that a written reference cannot substitute for: click latency to sub-millisecond precision, pixel response time across a full transition matrix, wireless link latency under load. For these we rely on published measurement from organisations that do this work properly, and we present the resulting figures as ranges reflecting the spread across current hardware rather than as point values for specific products.

This is why, for example, our latency budget gives "2 to 8 ms" for switch debounce rather than a table of per-product figures. A range we are confident in is more honest than a precise number we cannot stand behind.

Our own frameworks

Several things on this site are categorisations we invented because none existed and the category needed one: the mouse weight classes, the grip-contact ratios in the size finder, the shape archetypes, the upgrade order, the mousepad sizing bands.

These are judgements. We handle them by:

Labelling them as our framework wherever they appear.
Explaining the reasoning for each threshold, so that you can disagree with a specific boundary rather than the whole idea.
Stating the limits. The size finder gives a window and says explicitly that two measurements cannot capture hand shape.

Claims we deliberately do not make

Things you will not find here, and why.
We do not claim	Because
That a specific product is the best	Fit is individual and the differences between good products in the same class are small
Per-product click latency figures	Requires hardware measurement we do not perform, and the differences are below the threshold that affects play
That any sensitivity is optimal	There is no evidence for it, and the professional distribution is very wide
Precise professional player settings	They change constantly and any snapshot is stale within weeks
That a gear change will improve your rank	It might not, and claiming otherwise is how this category lost its credibility

Editorial standards
Corrections log
The reference tables these methods produce