I'd like to propose a modification to the statistical hockey lexicon. We'll see if it catches on. I think the notion of "shot quality" is a bit of a misnomer, and I've been as guilty as anyone of using the term. While we have a database of shot locations, and we can figure out if a shot came off a rebound or a giveaway, we don't really know anything about the "quality" of an opportunity. Screened shot? Odd-man rush? Wide-open net? It would be wonderful to have that data, and we don't. So, going forward, if all I'm talking about is shot location and whether a shot was a rebound, I'm going to call that "shot location."
At any rate, one interesting aspect of shot location is how it varies by shot type. Here we see the seven recorded NHL shot types for initial shots (no rebounds):
|Shot Type||% of Shots||% of Goals||Distance|
Deflections/Tips and Backhands account for a larger proportion of all goals than we'd expect - which isn't surprising given how they occur much closer to the net than other types of shots (except wrap-arounds, but that's another issue.)
We can normalize for shot location and determine how much more (or less) dangerous a given shot type is:
|Shot Type||Expected||Actual||Extra %|
So the shots that are most likely to go in - Deflections and Backhands - are also much less dangerous than other types of shots taken from the same locations. Makes sense to me - if you're on your forehand from 15 feet out int he slot, that's a great opportunity. On your backhand, less so. And if you're trying to deflect a blast from the point, it's difficult to control where it goes (in fact, that's often the intent of deflections.)
We can also see that certain types of shots are more likely to happen on rebounds:
|Shot Type||% Rebounds|
It's not totally clear to me what the difference is between a deflection and a tip-in - the locations at which they occur are indistinguishable - but official scorers seem to be making a judgment call here. Perhaps a deflection is passive, while a tip-in involves motion of the stick? Who knows?
We can see the impact that adding rebound data to the shooting percentage has:
|Shot Type||Non-Rebounds||All Shots|
If we look only at rebound shot location, most shot types actually perform worse than expected:
|Shot Type||Shots||Expect G||Actual||% Extra|
What do I take away from all this? Ryan Smyth should lay off the wraparounds.