>>102995443
This is actually kinda interesting, despite some caveats. He's
a) comparing users to viewers (more or less totally valid) and
b) running some analysis on chat and using botted chats as a proxy for viewbotting
Apparently, a lot of the very active chats on Twitch have what appear to be bots in them to simulate fast chat speed, spamming identical generic messages across multiple channels. They also tend to be pretty obvious, with automatically-generated usernames like "aaaa1" or "horatio8gonzales". So, essentially, having a fast chat isn't by any means a guarantee of legitimate users, you have to actually look at who's chatting and what they're saying.

That said, I'm not sure it's going to give people what they want. Finding a large number of these botted chatters would indicate botting, but if you don't have them you're left with the question of whether someone is botted or just has a slow chat. You also want to look at things like unique chatters, and see how many of them are legitimate. Also, his methodology is just bizarre here, he's taking screenshots of chat and running it through OCR rather than just directly pulling the raw text, which is something you can easily do, it comes off as very amateurish, at a lower level of coding skill than you see on many free third-party sites. That raises questions about whether he actually has the sophistication to make strong and specific claims about this.