Interesting read! I find it funny how we approach this same problem every ~5 years. (https://stackoverflow.com/questions/164831/how-to-rank-a-million-images-wi

Interesting read! I find it funny how we approach this same problem every ~5 years. (https://stackoverflow.com/questions/164831/how-to-rank-a-million-images-with-a-crowdsourced-sort https://stackoverflow.com/questions/164831/how-to-rank-a-mil...) I recently implemented a similar tool to help my team decide which brand and art references to use => http://refsort.com/ http://refsort.com/ One big difference is that I focused on subjective topics, where there is no absolute 'correct' or 'better' option. In that case I think transitivity could actually decrease the quality of the result, so more 'brute force' is required. I could probably implement this algorithm in a branch and see what happens :)