Adjusted Plus-Minus

Many of the projects I've worked on are related to Adjusted Plus-Minus (APM), so I thought I'd make that my first real, actual post. 

APM is a plus-minus-like statistic that improves over plus-minus in many ways.  Over at and/or, I wrote a four-part blog series describing how the statistic is calculated, the improvements that it is has over plus-minus, and some of the results for best offensive, defensive, and overall players at even strength.  Here are links to those posts:

NHL Adjusted Plus-Minus Part 1 (an introduction)
NHL Adjusted Plus-Minus Part 2, Even Strength Offense
NHL Adjusted Plus-Minus Part 3, Even Strength Defense
NHL Adjusted Plus-Minus Part 4, Even Strength Overall

At some point I'll post some special teams results too.

Recently, my updated version of APM was published.  See the version at Journal of Quantitative Analysis in Sports here or if you don't have access to that, see the version here.

This updated version uses a different statistical technique than the previous version (ridge regression instead of ordinary least squares regression).  It also computes APM for goals, and also some statistics other than goals:  Shots, Fenwick (shots + missed shots), and Corsi (shots+ missed shots+ blocked shots).  The results are more stable.  The improvement is greatest for the single-season results and the special teams results, since they use less data.  

In the paper, I also give some updated lists of APM's picks for various awards for 2007-2011: best defensive forward (Selke), best defensemen (Norris), best player overall (Hart).  Spoiler alert: Pavel Datsyuk is the man.  I'll be posting some other results here from time to time, like award winners for 2011-12 and the aforementioned special teams results, but not before posting about some other topics first...

No comments:

Post a Comment