Urban Rating

The Elo algorithm is a ranking system that was originally developed for chess tournaments. It establishes the average performance level of each player which remains fairly constant through time. In a chess game, a strong player may perform poorly and it will not have a short-term effect on their average performance score but in the long run, it may decrease their average performance score if they continue to play inadequately. We are applying the same method to our project.

As users undergo our pairwise comparison, each image will develop a performance score that will be held onto and used in the algorithm once we finalize the results. As each image is either selected or not selected, it will move up or down in the ranking system according to its outcome. The algorithm results in small changes of comparison scores if the outcome is similar to the predicted outcome. Although, if the outcome is not to be predicted then there is a possibility of a larger jump in the rankings for that player. The number of times an image is chosen is the only way to change the rating score for that image. This method is therefore highly dependent on the outcomes.