For example, on Reddit, the score that you see is extremely simple/primitive: upvotes - downvotes. The entire premise behind Caucus is that content can and should be scored using much more sophisticated algorithms. Some of the many factors that we use to generate scores: poster's reputation, voters' reputations, posters/voters' activity levels, community size, number of times the post has been seen by voters, etc.
If this all sounds very abstract and you're wondering what a sophisticated scoring algorithm looks like, here's a famous one that we drew inspiration from: https://en.wikipedia.org/wiki/PageRank.
Anyway, long story short: the scores that you see on Caucus aren't an indication of the number of people who voted on it. They are the output of our reputation-system-based scoring algorithm. Hope that clears things up!
You can have a magic algorithm, that's fine. You can have it rank submissions behind-the-scenes, that's fine. But you can't copy Reddit's UI/UX and claim it doesn't behave the same.
(Also, PageRank is based on relationship networks, which makes no sense for a link aggregation where each link functions independently from each other. Saying an algorithm is sophisticated requires substantial proof which you do not provide.)
You are incorrect about the timing of the 2000 point threshold on Reddit though. Those have been regularly hit since the Digg migration in 2010.
In that case, why are the vote scores on Caucus submissions blatantly inflated? There is zero probability scores are in the hundreds for completely random content. (glancing at the number of unique accounts who have made submissions, I question if 100+ active accounts even exist.)