The Case for THoR (December 2013)
This past February we introduced the Total Hockey Ratings or THoR at the MIT Sloan Sports Analytics Conference where it won 2nd place in the research paper competition. THoR rates players by taking the probabilities of a subsequent goal for every event for which they are on the ice and accounting for who they played with, who they played against, where their shift started, the score of the game as well as whether or not they were at home. In doing so we attempt to isolate the impact of individual players. The original paper is found at this link. Recently, we have put together an evaluation of THoR that makes the case in detail for the use of THoR. THoR has roughly the same validity as Corsi and Fenwick but it is considerably more reliable. That is, THoR’s performance is highly correlated from year to year and, hence, more useful for predicting future performance. This version of THoR is based solely upon even strength plays. Below are links to the three part to the case for THoR.
Part I of that case shows that there is high year to year correlation between THoR Values meaning that THoR values are more likely to predict future THoR values. THoR has a year to year correlation of about 0.65 which is considerably higher than CorsiRel.
Part II furthers the exploration of THoR by looking at how well THoR is associated with winning a game in the NHL. The response that THoR uses is in the neighborhood of Corsi and Fenwick in this regard though it is slightly less than both.
Part III looks at the yearly THoR values for some specific players to give a sense for how THoR performs. There are certainly some surprises and they have received a good deal of attention but in reality the top THoR players ( Bergeron, Zetterberg, Toews, Kopitar, etc) are generally not surprises.