[crazygringo]

Totally agreed -- there are so many different factors involved in each comparison, and I feel like I'm easily paying attention to different things on each comparison.

Unfortunately, I don't believe there's any established algorithm for how to repeatedly sample pairwise preferences to convert them into a strict ranking, which would ideally be with an active learning component to really drill down into the comparisons that are the closest. Would be a fascinating thing to try to develop, though.