The Glicko system

Arguably one of the greatest fascinations of tournament chess players and competitors of other games is the measurement of playing strength. The Elo rating system, developed by Arpad Elo in the early 1960’s, was the first chess rating system that had probabilistic underpinnings, and was then adopted by many chess federations, and eventually by organizations for other games (e.g., Scrabble, table tennis, etc.). While Elo’s system is a great improvement over earlier systems, it too has its problems. In 1995, I created the Glicko rating system in response to a particular deficiency in the Elo system which I describe below. My system was derived by considering a statistical model for chess game outcomes, and then making mathematical approximations that would enable simple computation. The Elo system, coincidentally, turns out to be a special case of my system. The mathematical details of the derivation can be found in a technical paper called “Parameter estimation in large dynamic paired comparison experiments” which is published in the refereed statistics journal Applied Statistics (48, pp. 377–394), but can also be downloaded from http://www.glicko.net/research.html. The Glicko system is currently implemented on the free internet chess server (FICS), and variations of the Glicko system have been adapted for several commercial internet gaming organizations such as ChronX, Case’s Ladder, and the Gothic Chess Association.