People are discussing a lot about the ratings and the ratingsystem here:
Why not evaluate it?
First and simplest idea:
Calculate the probability of going wrong/right (acktesting)
improvements: take the ratingdifference (of the two players) in account!
these were my first ideas: what about YOU?
(it would be nice to know: he probability that you beat/draw/loose a player with this ranking is 34% / 5% / 61%... and RHP informs you: you'll then have a rating of 1314 / 1288 / 1264. So you can expect a rating of 1282,2 which is a plus of 4 points to your actual rating 1278 - his rating is 1305)