It's a manual process of discovery. As in, I actually go out and hunt down groups talking about stocks and if it looks interesting, I'll add it into the algorithm. I'll then analyse the data over a period of a week or two and determine whether it's worth continuing. If it is, I move it to production and it'll perform as the rest do.
If a community is not publicly visible (discord, telegram, etc) I'll ask an admin if they're okay with what I'm trying to do. If it's a no, it's a no.
There are those groups that do try and force a pump and dump but I'm against using them. That's ultimately not what I'm after as I prefer organic conversation with genuine reasons to buy. Maybe the reason doesn't turn out to be true (BBBY) but that's the case with a lot of stock analysis anyway. Sometimes the reason a stock makes it to me is due to a pump and dump but there are measures taken to make sure that we're not acting on artificial metrics.
There is also a period of premarket validation which tries to measure the stocks that are deemed interesting each day, so people are (hopefully) not getting sent duds. It can happen, yesterday ADTX only managed 0.58% but moments like that help us fine tune the algo and those happen fewer and fewer.
