>>38509800>I think YouTube verification is to blame for this.It is. Actually for a period you could predict even regular streams (after the waiting room cull and rebound) based on a simple linear model.
The reason for that is simple, see this stream right here
>Anya debuthttps://holo.poi.cat/stream/BET9wXW37mwIt was right after the 2020 algo change, and it was Youtube's first version of the "waiting room cull".
Back then they would drop the CCV to **zero** and take whatever time they needed to verify the waiting room before disclosing the number.
In Anya debut's case, eight minutes on ZERO CCV.
Second interation they started doing it by "batches", meaning
>validate 1k>release 1k>validate another 1k>release another 1kand that created a continuous flow of numbers which had no relation to the actual number of people watching but with the speed their validation servers could validate a batch.
That "validation speed" is the slope of that curve, if they can validate 2k people per minute (let's say) and there are 50k people from the getgo watching a stream, it will take 25 minutes to reach 50k.
Only after the whole initial audience is validated, and after the audience that onboarded during said validation is also validated, is when you get the real CCV number on screen and the variations within the stream
