[alexandermorgan]

Neat article but I don't get two things: 1. if there are "hard rules" that always override the ranking decisions, why not weight them with a score that is greater than the total possible from the other ranking elements. E.g. if the highest possible score is 100, then a hard rule can have a score of 1,000 so it will override the other decisions no matter what. An even more significant hard rule can be worth 10,000. 2. Since only the top 1 really matters, why not do an nlargest type of thing so that you only do n-1 comparisons instead of (n^2 - n)/2 comparisons?