All stats for this blog were provided by corsica.hockey
Recently, a thought that has been crossing my mind is the term 'offensive defensemen'. There's no doubt that this should include the d-men who are consistently putting up points, and in no way is this post intended to strip those players of the title they've deserved.
I think we're missing half the story here. What about the d-man who gets the puck up to the forwards and starts the play that leads to the goal? Maybe this player isn't one of the last two players on ice to touch the puck before it goes in, but he was instrumental in taking the play out of his own end, and into the opponents end. I have absolutely no idea how often this happens. There is something to be said about a defenseman who is constantly pushing play out of his own end and into offense. Is this not an offensive defensemen? [I wonder if this is going to be a takeaway from Ryan Stimson's passing project?]
As a proxy, I've been working with the concept of "Impact". This was a concept used by Steven Burtch's dCorsi on the the now defunct War-On-Ice where Burtch would take a players dCorsi for impact and subtract his dCorsi against impact. With this (and all relative per 60 metrics, negative numbers are good for relative shots against), you're essentially adding together a player's impact on both sides of the ice into one nice rounded up number.
We can do this with Corsica.Hockey's relCF60 and relCA60 metrics. [You've probably seen my charts on Twitter recently applying this to scoring chances]. Using relative metrics to avoid punishing players for playing for bad teams.
With this in mind, here are the top-10 from last season:
[5v5 only, minimum 500 minutes on ice]
And the bottom-10
Yes, that is Dan Girardi being 4.44 shots worse than the next worst d-man in the NHL last season.
At face-value, this passes the smell-test. If you're trying to build a proxy for offensive-defensemen and Erik Karlsson is in your top-2, maybe you're on to something. And that's what this is, a proxy. An attempt to see if we can differentiate the players that are driving shots for their team relative to the shots they don't allow in their own end.
Rel Corsi per 60 impact doesn't have any huge relationship to individual points (0.12 adjusted R^2), Goals per 60 impact (0.1 adjusted R^2), or relative goals for % (0.09 adjusted R^2), and if you're trying to build a proxy for offense, you probably want it to relate to goals.
And that's a huge flaw. If you're trying to re-think the concept of the offensive defenseman, and the thought you have doesn't explain much variance in the form of individual points, nor does it have a huge relationship with the impact on goals, then where can we go from here?
There's also a bit of a line to toe here. Doing our best to eliminate team noise by using relative impacts. However, this also punishes players on good teams who happen to have better players in front of them, and then paints bad players in a better light, if they are on a team with other bad (read: worse) players. Thus is the conundrum of the regional manager who wonders why half of his markets are below the regional average. [Yes, I quit that job].
Back to random thoughts
We know that having the puck is important, and we know that shooting the puck is an important factor in having the puck. Where we do see a strong relationship is between Relative Shot Attempts for per 60 and relative Goals For per 60 (0.38 adjusted R^2). However, this relationship is not mirrored between Relative Shot Attempts against per 60 and Relative Goals Against per 60 (0.1 adjusted R^2) [[this data includes all players since 2007-2008 who played at least 300 minutes 5v5]].
Maybe it's best if we just continue to think of offensive defensemen not just as guys who drive the puck into the offensive zone, but as players who can do that, and put up the points to back it up.