jack schitt
New member
I'm building a website which will host fanpics (including rule34 type stuff.) One of the features I'm interested in implementing is an image ranking system where the users will be able to rate images. I don't want to do this the way hotornot.com does where you select a number between 1 and 10 for any given image. I want the user to be presented with two images and have them select which is "better".
A proof of concept of this is available as a feature of OKCupid called MyBestFace. I want to duplicate some of it's functionality. The difference is that I want to derive a score for any image on the site so I can have a search page that lists the top 50 (or so) images on the site and also give a "this picture is ranked at $score out of the $total images on this site." when viewing a particular image.
My problem? I have no background in statistics whatsoever. If anybody is familiar with comparative ranking (it might be known under a different name), can you tell me step-by-step how to implements something like this? I can create new tables in the database and new fields in an existing table if needed. My intention is to require a user comparison before a search is executed which should give a large sample base.
It looks like what I'm trying to do might be known as Ranked pairs or the Tideman method.
A proof of concept of this is available as a feature of OKCupid called MyBestFace. I want to duplicate some of it's functionality. The difference is that I want to derive a score for any image on the site so I can have a search page that lists the top 50 (or so) images on the site and also give a "this picture is ranked at $score out of the $total images on this site." when viewing a particular image.
My problem? I have no background in statistics whatsoever. If anybody is familiar with comparative ranking (it might be known under a different name), can you tell me step-by-step how to implements something like this? I can create new tables in the database and new fields in an existing table if needed. My intention is to require a user comparison before a search is executed which should give a large sample base.
It looks like what I'm trying to do might be known as Ranked pairs or the Tideman method.