Repeatability of sv% when adjusting for shot location

Earlier today Paul Riley posted an article with some pretty fascinating stuff. Namely a grid showing the average Premiership sv% by shot location (see also Gabe Desjardins’ plot from a couple of years ago). I’d recommend reading both pieces.

The piece in Paul’s piece that was most interesting to me were the tables detailing the Premiership ‘keepers relative save percentage, as adjusted for the location of the shots a given GK has faced.

I’ve written previously that raw sv% at the team level is a mess – it comes out as ~40% skill and ~60% luck over the course of a Premiership season. If we adjust crudely for shot location, by splitting shots into inside and outside the box, that split drops further – to ~25% skill and 75% luck.

To get an idea as to whether this held true at the GK level I took all of the GK’s from Paul’s table who’d played in both the ’11-12 and ’12-13 seasons and plotted their adjusted save percentages against one another. (I’d add that I got permission to use these numbers – it’s not a lost art)

Screen Shot 2013-12-05 at 7.40.17 AM

So the nub of this plot is that, when shot location is accounted for, sv% at the individual level over the course of the season is ~21% skill and ~79% luck. If we regress the best performance of these years 80% of the way towards the mean we’re basically saying that we’d expect the best shot stopper in the league this year to be less than two goals, or about 1 point better than the league average at shot-stopping next year.

This result also agrees with my crude splits for location for shots from inside/outside the box, and I’d further suggest it marries nicely with mine and Devin Pleuler’s thoughts from a couple of years ago that we shouldn’t really be evaluating ‘keepers based on their shot stopping abilities. It’s not that I don’t think that the best ‘keepers have value, or that the market on ‘keepers has inefficiencies (I honestly don’t know that) – it’s just that I think this adds still further to the pile of evidence suggesting that the true value of a player/team isn’t being found if we evaluate based on sv%.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s