Warm up → test → repeat → compare
Use repeated runs under similar device conditions. Browser timing includes hardware, display, OS and scheduling effects, so trends are more useful than pretending one run is laboratory latency.
Convert polling rate to report interval
The theoretical report interval is 1000 divided by the polling rate in hertz. For example, 1000 Hz corresponds to 1 ms and 8000 Hz corresponds to 0.125 ms between reports in the simple arithmetic model.
Not total system latency
The calculation does not include sensor processing, USB scheduling details, CPU/game timing, render queues, display scanout or human response. Use it to understand rate/interval relationships, not to predict click-to-photon latency.