I've been pondering how to solve this problem for years, and I'm no closer to solving the problem than anyone else.
Your approach involves freezing the user base so that nobody new can join. Granted, it's an innovative take on it I've never seen before, but that's still what it is. The problem with all such approaches is that a certain percentage of the user base wanders off over time, so your frozen userbase constantly shrinks. Eventually it gets below the point you care. (Or, in most cases, below the critical mass of the community.) It only works correctly in the static case where nobody ever leaves.
In your case you could raise the threshold, but all that does is expose you to the very things you were trying to get away from. If there was a correlation between "the people who stuck around" and "quality posters" that might work, but I doubt such correlations exist. (Then again, I don't know, as my words imply.)
About two years ago, I scraped a single article with over a thousand comments on Slashdot and looked at all the UIDs. I theorized that all the old folk would be gone. Instead what I found is that the UIDs were fairly uniformly distributed across the entire base, which, if you think about the temporal distribution of people joining, is actually a moderately interesting result. I just eyeballed the list of unique UIDs, I didn't actually plot it out, but it wasn't what I was expecting. Some small core stick around and keep posting, but over time the vast bulk of the early group leave.
Over time the vast bulk of any group leave, not just the early adopters.
They'd have to, there are just too many reasons to leave and only a very limited few to stick around indefinitely.
Eventually everybody dies, gets a life or lands in some sort of crisis in their lives that changes it to the point where spending time on social sites is just too expensive.
Your approach involves freezing the user base so that nobody new can join. Granted, it's an innovative take on it I've never seen before, but that's still what it is. The problem with all such approaches is that a certain percentage of the user base wanders off over time, so your frozen userbase constantly shrinks. Eventually it gets below the point you care. (Or, in most cases, below the critical mass of the community.) It only works correctly in the static case where nobody ever leaves.
In your case you could raise the threshold, but all that does is expose you to the very things you were trying to get away from. If there was a correlation between "the people who stuck around" and "quality posters" that might work, but I doubt such correlations exist. (Then again, I don't know, as my words imply.)
About two years ago, I scraped a single article with over a thousand comments on Slashdot and looked at all the UIDs. I theorized that all the old folk would be gone. Instead what I found is that the UIDs were fairly uniformly distributed across the entire base, which, if you think about the temporal distribution of people joining, is actually a moderately interesting result. I just eyeballed the list of unique UIDs, I didn't actually plot it out, but it wasn't what I was expecting. Some small core stick around and keep posting, but over time the vast bulk of the early group leave.