ReactionMetric

Most reaction tests hand you a number and maybe a rank. ReactionMetric takes the measurement more seriously. It times from the moment the browser switches the panel to GO on an animation frame until your click, tap or key press is detected, and it says outright that your display and input hardware add delay on top of that. The site's advice is to compare repeat results on the same setup rather than read a small difference between devices as a change in your reflexes.

Scoring follows the same thinking. You get five valid trials, and the median is the headline figure, with best, worst and consistency beside it. A sample on the page shows why: four trials between 221 and 247 ms and one at 310 ms still produce a median that reflects your usual speed.

The optional community benchmark is built with the same care. After a session you can choose to add your result anonymously. The site then compares you with people on a similar setup, matched by device, input type and refresh-rate bucket, and widens the group only when the closer match doesn't have enough results. It shows a percentile only after the comparison group reaches 100 contributed sessions, so early on you may not see one at all.

Your own history stays in your browser's local storage, in a table of past medians, bests and consistency that you can clear at any time.

Beyond the visual test there's a keyboard version, a sound test using a browser-generated tone, an F1-style five-red-light start with false-start tracking, a clicks-per-second test and a double-click diagnostic for checking a mouse. Guides and a methodology page explain the timing.

There's no account. The privacy policy says trial values are uploaded only when you contribute a session, that the site keeps first-party aggregate counters for reliability, and that Google Analytics is paused and AdSense isn't included for now. The footer is clear that this measures browser-observed performance and isn't a medical test.