We try to keep this content unique and up-to-date. Consider using the social media buttons below, or add the page URL when sharing it anywhere.
Conventional adjusted plus-minus does a poor job of predicting the outcome of future games, particularly when you have less than a season of data. Adding regularization greatly improves accuracy, and some players’ plus-minus ratings change dramatically.
The enhancement with the RAPM is a Bayesian technique in which the data is combined with theoretical beliefs regarding reasonable, large data ranges for the parameters in order to produce more accurate models. That is what ridge regression does.