Research protocol · Version 1.0 · 11 Aug 2026
A sports-streaming benchmark another tester can reproduce.
The method controls device, software, network, content state, repetitions, and reporting so a ranking cannot be reverse-engineered from a marketing conclusion.
Scope
The protocol measures authorized live or replay playback on named devices and platforms. It does not test unlicensed sources or generalize one event to an entire service.
Required test record
- UTC date and region
- ISP and advertised tier
- Router, connection type and measured network conditions
- Exact device model, operating system and app version
- Platform, content type, resolution, HDR, codec and frame rate
- Startup time, rebuffer count/duration and dropped frames
- Caption availability and audio languages
- At least three repetitions per condition
- Known incidents, maintenance and limitations
Network controls
Run wired and wireless conditions separately. Record download speed, latency, packet loss, Wi-Fi band and signal context. Do not change the platform, device and network simultaneously when diagnosing a difference.
Startup time
Measure from confirmed play selection to the first continuously moving decoded frame with audio. Discard accidental navigation delays and report median plus range across repetitions.
Rebuffering and frame behavior
Record every playback stop attributable to buffer depletion, total stalled duration, and dropped-frame count where the player exposes it. Distinguish an intentional ad break, stream restart, or source cut from rebuffering.
Latency
Use a stated reference and synchronized clock. Report the method and uncertainty. Notifications, crowd noise and social posts are not precise reference sources.
Accessibility
Test caption presence, language, synchronization, remote/keyboard access, focus, seek controls, and whether replay behavior differs from live behavior.
Publication gate
Results require raw data, sample size, exact dates, conflicts or sponsorship disclosure, limitations, and a correction path. A score without the underlying observations is not published.
Download the open template
Use the CSV as-is or publish your changes so results remain comparable.
Download CSV