I don't have a score in my head for each politician: I just know which are better, and which are worse. Consequently, > Most obviously, ranking is effectively

I don't have a score in my head for each politician: I just know which are better, and which are worse. Consequently, > Most obviously, ranking is effectively a bubble sort, О(n^2), whereas rating is just two passes: 1. Get the max and min. 2. Normalize, thus O(n). is meaningless to me: rating is impossible, because it's asking me for numbers that I just don't have.