Wednesday, August 06, 2008

The +/- Challenge Of Anaheim's Checking Line

In my recent sabermetrics and hockey posts about on/off ice adjusted +/- ratings one group of players that keep coming up with unique stats is the Anaheim Duck checking line of Sami Pahlsson, Rob Niedermayer and Travis Moen. All three of them rank as one in the ten worst quality of teammates rankings and the ten best quality of opposition rankings.

Anaheim is the only team currently in the NHL that gives lots of playing time to a regular checking line that plays together most of the time. The line of Pahlsson, Niedermayer and Moen is very good at preventing goals. All three of them rank very well in a list of low individual player goals against. Among players with more than 50 games played last year, Moen ranks eighth with a 1.35 GAA when he is on the ice in 5 on 5 situations, Niedermayer ranks 12th with 1.42 GAA and Pahlsson has a 1.71 GAA (which is also well above league averages). They are clearly very good at preventing goals from being scored. They do this against very high quality of opposition (in fact Pahlsson has the highest quality of opposition, Niedermayer is 2nd and Moen is 6th). This makes them valuable players to the Anaheim Ducks. They do not score very well in the roles that they play. Last season, Niedermayer has 16 points, Pahlsson had 15 and Moen had 8 points. This gives them relatively poor +/- ratings. When compared to their team (Anaheim was a pretty good team that had the third best team +/- in the league) it gives them on/off adjusted +/- ratings that are well below zero. Pahlsson is -1.23, Moen -1.08 and Niedermayer -0.94. None of these are bad enough to get on the 20 worst list but they are not good. The fact that they continuously play against the top opposition and prevent them from scoring and continuously play with one another instead of Anaheim's higher scoring players makes their ratings worse than they should be. This is identified with the quality of opposition and quality of teammates rankings. It is not clear how to properly adjust for this. Some thought and effort needs to go into that problem.

When making sabermetric ranking systems, one must be aware of the unusual cases that continuously show up in unusual ways in the rankings in order to properly set their rankings. The Anaheim checking line is the most unusual case in the NHL today. Setting up a system that gives them appropriate value is hard. It is a challenge to the existing systems. A good sabermetric theory (beyond that which exists today) will need to be able to handle cases such as these. Sami Pahlsson, Rob Niedermayer and Travis Moen are a good test to any such system. I think the information exists in the analysis of the on/off ice +/- system, but the proper analysis to do so is currently lacking. How best do we account for high quality of opposition and low "quality of teammates" (I put that in quotes because I do not mean to say they are low calibre teammates for one another - but under the system they appear and such and that needs correction). A proper system will show their value without any ad hoc adjustments and without compromising the value of any other player or class of players.

Since these three aren't looking to score goals, would it be a good idea to just take the + part out of the equation, and just look at adjusted goals against/60? Using qualcomp as a modifier, I would think you could get a decent measure of how good defensively these guys are. Whether that makes up for or overcomes their lack of offensive, I've no idea.
Good post, it is a good set of questions raised.

Whether that makes up for or overcomes their lack of offensive, I've no idea.

Well, probably there's a payoff for the rest of the players who do not have to see regular shifts against the Iginlas and the Thorntons. That may be the trick: the benefit of running Pahlsson against the league's best may not be found in their statistics at all, but rather in the results of Getzlaf and Selanne and friends.

You gave an example of an ad hoc system made up especially for those three. It's pretty useless to have a special rating system for a handful of players treated differently than the rest. A good system would treat them fairly without any ad hoc adjustments to help them.
I don't mean for it to be a way to evaluate only those three- I was proposing this system as a way to evaluate a player or line's defensive ability. For this three, it would be a measure of their primary value as players.
This blog is an exact representation of skills. I appreciate the blogger for posting the most excellent thought.


Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?