Abstract:We present an Elo-based rating method for ranked multi-player games. We justify a definition of performance using the logarithm of a player's rank. We customize the method for rating TopCoder SRM. We choose parameters which maximize the rating's prediction accuracy when applied to past SRM, and preserve similarities with existing SRM ratings. We evaluate the accuracy of the proposed system on the available data. Results show that the proposed system has a higher predictive accuracy than the existing SRM rating system, and hence could be a good alternative.