Copyright © 1999-2005
Wayne T. SmithThis page was most recently revised on Sunday, October 19, 2005
Your comments and suggestions are most welcome!
A major concern with HEAL and RPI is that the majority of games will be conference games, and that the different conferences will have significantly different overall strength. When conferences have different levels of strength, RPI and HEAL undervalue the teams of a stronger conference and overvalue the teams of a weaker conference.
HEAL can adapt to this in an adhoc manner by adjusting the value for a win (or tie) according to conference. The Maine Principals Association does this by assigning varying number of points for a win according to the size of a school (40 for Class A, 35 for Class B, etc.). See the Heal Point System page at the MPA web site for more information.
RHEAL takes the HEAL method a step further. Once initial HEAL ratings are calculated for each team, they are then used as the number of points for a win of that team. The process simply iterates until a steady state is found. In practice, this steady state is reached relatively quickly ... perhaps 5-20 iterations. Although it is not known if this method always reaches a steady state or even if there is just one steady state, some experimentation has shown this method to be consistant, with predictions (rankings) closer to what is probably considered the best ranking method: KRACH (uses the Bradley-Terry method). Please see our KRACH page, too!
Game scores for this ranking have been provided by US College Hockey Online, sometimes supplemented by other sources. USCHO's composite of all Division I men's ice hockey games is available at
in the section marked Men's Division I schedules.
Copyright © 1999-2005
Wayne T. Smith
This page was most recently revised on Sunday, October 19, 2005
Your comments and suggestions are most welcome!