I feel i'm going to open source this so people can add their own functionality. I will need to refactor first since I just hacked this together rather quickly.
To address the exact situation in your parenthetical, I considered putting Sum(log(post_karma[i]-k)), for a k such that the expected log value is negative unless you get enough upvotes.
Whatever is feasible. For a while I've wanted a list of "blogs/domains that hn likes" that isn't polluted by general-high-traffic domains.