Further to that, if I'm reading how they determine shot quality correctly, it's about distance to the goal more than anything else. You don't get things like how much traffic is in front of the net or anything of that matter.
For instance, a shot from the point that hits my defenceman and goes in from 3 feet awya is a shot from the point. A shot from the point that hits the opposing player and deflects in from 3 feet a way is a deflection from 3 feet. Likewise a shot from the bottom of the faceoff circle is approximately 19.4 feet away from the goal (depending on your rink... I'm using minor hockey because that's the one I have handy). A shot from the top of the face-off circle is about 44.3 feet away from the goal (again, depending on the rink), but the angle is a superior shot and a much better area to score from.
I've been using the NHL gamecenter data to try and map out the best scoring areas in the NHL. The problem initially is that the sample is only for one season so it's tough to draw too many conclusions from it, but it does seem pretty straight forward.
I brought up 2 problems initially and this addresses one of them. That is, where the registered shot came from on the ice. What that doesn't account for is where the initial shot came from (i.e. when the offensive team tips a shot) or what it hit on the way to the net (i.e. when the defensive team tips a shot).
There's simply no way to account for the latter based on the information we have. You'd have to scour the game tapes and that's not something I'm willing to do without getting a decent salary from an NHL team.
That being said, we can gain an idea of where goals at the NHL level are scored from and attempt to identify the Prime Scoring Area (PSA). The PSA is the area on the ice where you have the greatest chance to score from. Often times that is not simply the distance to the net, but that in concert with the angle to the net. If you look above, it's pretty easy to see that the 19 foot shot is much closer but at a significantly worse angle. Unless it was a one-timer (or the goaltender was out of position because of a rebound or weird bounce), this shot should be stopped by an NHL goaltender virtually all the time.
Conversely, the shot from 44 feet away is farther away, but the angle is significantly better. I'd still expect an NHL goaltender to stop this shot the majority of the time, (95% of the time), but it's a significant improvement on the first shot.
The other aspect to consider is how missed shots come into play (in other words shots that miss the net). I'm going to assume that a shot from the point is more likely to miss the net than a shot from the hashmarks. Just seems like common sense to me. When Identifying the prime scoring area it should not just be shooting % (goals/shots on goal) that is considered, but instead goals/shots at net. I have a feeling that they correlate well anyways, but I can think of examples where it might not, especially around the edges of the PSA.
The reason for that is due to the angles goaltenders play and the quality of the shooters. If you've got a blocking style goaltender who only gives you a couple of inches up high and you find that hole, it's in the net. If the only shot you have is up high, there are more ways to miss the net (wide and high) than shooting low (wide). So even if I'm scoring (shooting% wise) from that area, if I am missing the net twice as much it skews the data.
So despite these flaws (and they aren't insigificant) it's still not a pointless task. We may not have the most accurate shooting charts, but we can still get an excellent indicator of where the most dangerous shots come from and who takes them/gives them up.
It's a bit of a process but I hope that there will be enough concrete data across all NHL games between last season and this season to get a somewhat reliable chart. Once we have that, we can start to see how various goaltenders do with their respective clubs based on that chart.
I have a feeling though that we'll run into the same conclusions Vic did... which is it doesn't make much of a difference from season to season on the whole, but it might help in identifying why one guy saw a huge spike in sv% relative to his past performance.
3 comments:
I just realized that you've got a bunch of new work up. Thanks for putting it out! With respect to the project you've outlined here, you may be interested in this piece by Desjardins who has done some work in this area, especially as it pertains to some of the concerns you've raised here:
http://www.behindthenet.ca/blog/2009/07/shooting-percentage-estimation-errors.html
If he can somehow get the x and y (which he seems to have, I can't find it though on ESPN), that makes a huge difference and makes it much easier to plot this out.
One thing I was always curious about... is the distance tracked from the middle of the net or the closest part of the goal (that the puck could go in)? Not that it makes much of a difference, but that can change things by a foot or two.
You know, I'm really not sure. I don't think they give actual coordinates to be honest, just a picture that Gabe has then created coordinates for. But that might be wrong. Someone tried to ask him in the comments to the first post I threw out there and he didn't give a clear answer (at least, it wasn't clear to me). In an earlier post he links to one of the ESPN pages and I can't find the coordinates but they do show up on the picture. Here's the link to his earlier work:
http://www.behindthenet.ca/blog/2007/12/shooting-percentage-by-zone.html
Post a Comment