Live streaming · measured, not estimated

Who is being watched,
right now.

The top five hundred live channels on Twitch, ordered by the people actually watching them this minute — with the categories they are in, the languages they are in, and what has moved since the last reading. Under that, the two questions a leaderboard cannot answer: who gained the most, and who is growing fastest from a small base.

Reading the wire…

The platforms

What each one is carrying.

Every figure on this row is computed from the channels this desk can actually see: the live five hundred on Twitch, counted at the moment you loaded the page. Where a platform publishes no live ranking, the card says so instead of guessing.

The watchlist

On, or not on.

A named set, drawn whether or not they are broadcasting. Twitch answers for the ones who are; the “last seen” on the ones who are not is this desk’s own memory of its snapshots, because no platform publishes it.

The live hundred

Ranked by who is watching.

# Channel Category Lang Watching Change

The movers

Three ways to be the biggest.

A mover is the difference between two readings this desk took itself, and both readings are still on disk. Pick the window; the panel says how far apart the two it actually used were, never the distance you asked for.

Momentum

The small ones going up.

The largest gains as a SHARE of where a channel started, which is the only way a room of four hundred shows up beside a room of forty thousand. There is a floor under it, and the floor is stated: a channel going from eight viewers to eighty is a thousand per cent and means nothing.

Method

What is measured, and what is not.

Twitch is live and it is theirs

Get Streams returns every channel broadcasting right now with the platform’s own concurrent viewer count. Five pages of a hundred is the top five hundred. Nothing here is modelled, smoothed or estimated: it is the number Twitch publishes, read through a relay that holds the credential so your browser never does.

YouTube is scale, not a live ranking

The Data API publishes per-channel statistics — subscribers, lifetime views, uploads — and a most-popular chart. It does not publish “the channels with the most people watching right now”, so this desk does not draw one. What YouTube contributes is size, and it is labelled as size.

Kick is absent on purpose

Its v2 endpoints are reverse-engineered rather than published. A property built on an undocumented internal endpoint is a property that starts lying the day it changes, quietly. If the official API earns a place it will arrive the way these did: with a credential and a name on it.

The history is this desk’s own

A snapshot of the top two hundred is filed every time the relay is asked for one, and kept for fourteen days. Every change, every mover and every momentum figure is the difference between two of those files. That is why this page is thin the day it is installed and better a week later — it is measuring, not remembering.