livecamsvr

Methodology

The index, documented.

Everything on livecamsvr comes from one pipeline, and this page documents it so you can judge the numbers yourself.

The live poll

Every 5 minutes, a crawler asks the two native VR platforms for their current VR-roster. Per model we record: live state, viewer count, stream resolution, private and spy per-minute rates (as published by the platform), connected toys, languages and the direct room URL. A model absent from a poll is marked offline.

The 30-day heatmap

Each poll lands in an hourly grid (7 weekdays x 24 UTC hours). For each model and hour we keep how many polls found her live out of how many polls were taken. The "usually live" note on a card and the whole schedule finder come from this grid. The reliability threshold is deliberately strict: a model counts as "reliably live" at an hour only if she was live in at least 40% of that hour's polls over 30 days.

Medians, not averages

Private-rate medians are computed across currently-live models that publish a rate. Medians resist distortion from a handful of extreme rates, which is why we prefer them to averages. The "cheapest live" figure is a floor, not a recommendation.

Freshness honesty

If the crawler is down, the wall shows it: the ticker carries a "refreshed X min ago" line and turns into a warning beyond 25 minutes. We would rather show stale-with-a-label than pretend data is live.

What we do not track

No model-side analytics, no cross-site tracking, no personal data beyond the anonymous daily-rotating visitor key described in the privacy policy.