Set up a reproducible rest session
Connect the controller using the method you want to investigate and keep the browser page in the foreground. Press a controller button so the browser can expose the device, then scan and choose its numbered entry. Select the hands-off rest protocol before starting. Place the controller on a stable surface and release both sticks.
The browser returns normalized axis values, commonly between minus one and one, through its input mapping. The lab records those values without adding a deadzone or centering correction. This is useful because a corrective filter could hide the very center behavior you wanted to inspect. It also means that the results reflect the browser and driver representation rather than an electrical reading from the sensor.
Read more than one center statistic
The mean vector describes the average center offset of an axis pair. RMS radius includes variation around that mean, while maximum radius shows the largest sampled displacement. A few brief excursions and a persistent offset can therefore produce different summaries. Inspect the trace and the individual axis statistics alongside the headline numbers.
The review radius is a user-chosen reference circle. Counting snapshots outside it helps compare repeated runs, but the threshold is not a manufacturer specification. Moving a stick during a rest run invalidates the hands-off interpretation. If you touch the controller or change a setting, stop the session and start a clearly separate measurement.
Compare conditions before comparing results
Keep the connection, browser, selected controller and protocol the same when using a baseline. Let the controller settle after movement and repeat the same approximate session length. A difference between two short runs may reflect normal variation or a changed input mapping rather than a hardware change.
Export the JSON if you need the underlying samples, not just the summary. Reports omit the raw controller identifier but retain protocol, review radius, mapping and relative times. Browser sampling can miss short transitions and does not locate the cause of a symptom. Treat the result as reproducible input evidence for a wider troubleshooting process, not as an automatic instruction to replace a device.