Recommending users based on ELO Rating

The ELO rating works on the basis of competition between 2 players in chess. There could be 3 outcomes of any chess game : win, lose and draw. New ratings of players are then calculated on the basis of the outcome and the old ratings of the players involved. Extending this model to music, imagine a competition between all the songs to be the favorite song of a user. To this end, let a user randomly shuffle play a set of songs. The current song playing is considered to be competing against all the other songs playing and there could be 3 outcomes to the song : repeat, skip, or let it complete and move to new song. Let repeat be associated with a win, skip with a loss, and allowing to run to completion as a draw. Then ratings of the songs are updated based on the outcome and the existing ratings.

Problems faced by the model

1. ELO rating system was made for 2 player systems. So in a large collection of songs, when a song wins, every other song loses. This would be biased for songs which are way down the queue and havent even been given a chance yet. Using nearest neighbours also doesnt solve the problem as user might not know which song is next. Maybe the next song doesnt deserve to lose and user would prefer to listen to that song than the current one. So, futuristics assumptions like this can deviate the ratings of songs in undesirable ways.
2. Listening pattern of a user is inherently time based. At any point of time, the user might be in any type of mood and may even skip his all time favorite for a newly discovered song that suits his mood. However, if measurements have been done for a long time, the all time favorite song will have such higher rating that it will take a lot of time to sink back down to allow the newly discovered song to be recommended. In fact, this model only ensures that most played songs or least skipped songs get recommended. There is no "hotness" factor. A lot of sites use the newton's cooling formula to estimate the hotness of an item. It gives them the trending the items of the website and is time based. As time passes the item cools down and the items with the highest ratings(non-time based such as reviews etc.) rise up. ELO can only take this into account if the number of plays of every song is roughly the same. Songs do not have a trending factor