Hogging the Puck: A Follow-Up (All That You Wanted & More)

Ben Wendorf
November 15 2012 07:45AM


After initially publishing the piece on players who "hog" the puck (players who take a high percentage of the Fenwick attempts-for {shots + missed shots} when they're on the ice), I received a lot of helpful feedback and queries about the metric, which I called "percentage of attempted shots," or %AttSh. Some of the questions revolved around, "What if the player is playing with someone with a high %AttSh, like Rick Nash or Jeff Carter?" I had another question wondering if the %AttSh had a normal distribution like all of us stats folks love. And our own Eric T. wondered aloud what a chart of the player's shooting percentage minus his linemates' shooting percentage (x) would look compared to the player's %AttSh (y). Some of these questions I'd been wondering about myself, but some were angles I hadn't considered, so I figured I ought to put together a follow-up post to tie up some of those loose ends. Enter if you dare...

Dicing the data

So, initially I think the distribution question should be addressed first. Those of you who follow my Twitting will have already seen this, but here is the distribution of forward %AttSh performances, with one percentage-point intervals.

Embedded image permalink

Juicy.

The next question, Eric T.'s, is more about level of effect. That being, there is a minimal amount of effect of teammates on a person's %AttSh, but there is an effect, so what does it look like? Taking the player's shooting percentage minus their linemates' shooting percentage on the x axis, and the player's %AttSh on the y, our chart looked like this:

Embedded image permalink

As you can see, it doesn't seem to follow that a better shooting percentage than your teammates means much at all in regards to how much you shoot; if anything, it means you shoot less. I'm not willing to admit that latter point, though...I think the balance is tipped a little due to shot volume, that the highest-volume shooters are not necessarily the best shooters.

When puck hogs collide

The final question gained some steam off this interesting post from Canucks Army's Thomas Drance, and was further suggested by Cam Charron, that teammate %AttSh could be having an effect on the player's %AttSh. This was something that I had though about before, and wanted to explore, but I also knew that putting together the data-set would be kind of a pain in the ass. Thankfully, Gabe Desjardins' Behind the Net data once again made everything easier (he's made this whole thing easier, people; seriously, if you are interested in this stuff, Gabe's Behind the Net site should be your #1 bookmark with a bullet). I was able to compile a list of forwards over the past five years who had a drastic change in linemates - basically, any forward who went from spending 50%+ of their ice time with a primary forward and 30%+ with a secondary forward, then spent 50%+ and 30%+ of their ice time the following year with different forwards, would make my list. I could then compare the change in the %AttSh of linemates to the change in %AttSh of the players in question.

I ended up with a data-set of 293 forwards. Among that data-set, the correlation between the change in linemates' %AttSh from one year to the next to the player's %AttSh one year to the next was 0.267, suggestive of a slight positive (though very weak) correlation. Since the differential took the previous year's %AttSh's and subtracted the current year's %AttSh, this mean's that when your linemates' %AttSh went up, yours has a slight possibility of going down. In a chart, it looks like this:

Embedded image permalink

To build this chart, I ordered the "Tm %AttSh Diff" from lowest to highest, then diced them into consecutive buckets of 10 (the last two buckets were 12 and 11) and took the average "Plyr %AttSh Diff" related to those buckets. This sharpens our focus on the trend a bit better than the scatterplot, by bring our variance into line a bit. As you can see, the change in linemate %AttSh doesn't seem to swing player %AttSh much further than 2.5%, which doesn't move beyond the normal variance of a player's %AttSh from one year to the next. So it's not an expected effect, though I speculated to Cam that at the extreme ends of the spectrum (the Rick Nashs or, on the flip side, Kyle Wellwoods) you'll probably see some effect due to the fact that we're comparing two things (Tm %AttSh and Plyr %AttSh) that are closely related.

The last thing I want to say relates to some of the things Eric T. mentioned in the first post. If this is behavior, how might it be changed? What is the ideal combination of %AttSh linemates? Does simply taking a low %AttSh with a high shooting percentage and getting them to increase their %AttSh equal success?

417ea714751a4580f06253a09a79bab0
Benjamin Wendorf was co-manager of the SB Nation Winnipeg Jets blog Arctic Ice Hockey (formerly Behind the Net); he is currently in the PhD program in Africology at the University of Wisconsin at Milwaukee. You can follow his graph work on Tumblr or his tweets @BenjaminWendorf. He can also be reached at bwendorf84 AT gmail DOT com.
Avatar
#1 Eric T.
November 15 2012, 10:27AM
Trash it!
0
trashes
+1
0
props

It looks like the slope on your last plot is something like 0.25-0.3. That might look pretty flat, but assuming the x-axis is the difference in combined (rather than average) %AttSh of the two new linemates, I think it means that puck-hoggery is fairly elastic.

Imagine right now, you're on a line with two average guys who each take 25% of the shots, and you're an average guy who takes 25% of the shots (leaving 25% for the defensemen, presumably).

Your team makes a trade for a 35% puck-hog (something that NHL teams have done fairly often recently), and drops him on your line. So the combined %AttSh of your linemates goes up by 0.1, which means we should expect your shots to drop by 0.1 * 0.3ish = 0.03ish, to 22ish%. And since you don't spend 100% of your time with that linemate, that 3% drop in your overall numbers might come from a 5% drop when you're with him. Add in the same factor for the third linemate, and it seems like your puck-hog linemate's extra 10% shots have come almost entirely at the expense of the other forwards.

This leaves me with two questions:

1) Is a defenseman's %AttSh affected in any significant way by which forwards he's playing with? That is, are the forwards just competing with each other for the 75% of the pie that forwards normally get, or do defensemen see a lower %AttSh when they're on the ice with the Carter-Nash line?

2) Is the puck-hog linemate really costing you shots? In one scenario, he takes shots at times when an average forward would have created a shot for you -- this decreases your shots/60 (the numerator) without affecting the team's shots/60 (the denominator). In the other scenario, he takes shots at times when an average forward would have been unable to create a shot -- this has no effect on your shots/60 but increases the team's shots/60. I guess what you'd see then is that a puck-hog would tend to be a high-event player: his tendency to shoot wouldn't affect his team's shots-against-per-60, but would increase his team's shots-for-per-60, resulting in more total shots (regardless of whether his various possession skills lead to a positive or negative Fenwick).

Avatar
#3 David Johnson
November 16 2012, 06:42AM
Trash it!
0
trashes
+1
0
props

The way I view a typical quality line (at least among the top 2 lines) is you have a playmaker, a shooter/goal scorer, and a grinder/secondary goal scorer/stand in front of the net collect rebounds guy. Think H. Sedin, D. Sedin, Burrows. Datsyuk, Zetterberg, Holmstrom of a few years ago. Backstrom, Ovechkin, Knuble in 2010-11. Giroux, Jagr, Hartnell. Second tier lines often don't have as well defined roles but they still loosely fit that model.

My guess is, if a player switches lines/teams but remains in a similar role his %AttSh would not change as much. i.e. Knuble on the Flyers to Knuble on the Capitals might have similar %AttSh. I suspect AttSh% is far more likely to change significantly when players change roles. i.e. a center who may have been more of a playmaker shifts to wing and becomes more of a shooter or a player gets moved from primary goal scorer on a second line or a first line on a weak team to a secondary goal scorer on a stronger first line.

Eric, as per your question relating to defensemen %AttSh varying based on the set of forwards they play with. I suspect you will find this and zone starts may play a factor. Defensemen generally don't get shots on the rush but rather get shots during sustained play in the offensive zone. Forwards that get a lot of neutral and defensive zone starts will probably see a lower %AttSh from defensemen. Forwards that see a high OZone% will probably see a higher %AttSh from defensemen. We know that shots taken immediately after a face off are low quality shots (significantly lower shooting percentage on those shots) and I suspect a big reason for this is they are low quality point shots by a defenseman after a face off win. A way to test this would be to look at Canuck defensemen with the Sedin line and with the Malhotra line.

Avatar
#5 David Johnson
November 16 2012, 08:55AM
Trash it!
0
trashes
+1
0
props

@Ben Wendorf

I am surprised zone starts didn't show much of an effect. What if you took two groups of forwards, those with OZone%>55% and those with OZone%55% group should have a lower %AttSh mean than the

Avatar
#6 David Johnson
November 16 2012, 08:58AM
Trash it!
0
trashes
+1
0
props

@Ben Wendorf

Ok, that didn't post my comment correctly. Seems like it cut stuff off between my greater than and less than signs. Probably assumed it was html and stripped it out. Let's try again.

I am surprised zone starts didn't show much of an effect. What if you took two groups of forwards, those with OZone% greater than 55% and those with OZone% less than 45% (i.e. the extremes and probably best to use only players with 10m/60). What is the mean %AttSh of each group? If you can calculate weighted mean weighted by total shots WOI it would probably be better but maybe not necessary. If my hypothesis is correct, the greater than 55% group should have a lower %AttSh mean than the less than 45% group but it is certainly possible that the small effect that might occur during that small segment of ice time after a zone face off is just overwhelmed by all the other factors that influence %AttSh.

Avatar
#8 David Johnson
November 16 2012, 09:11AM
Trash it!
0
trashes
+1
0
props

@Ben Wendorf

Ok, so zone starts have no effect. Thanks for checking it out.

Comments are closed for this article.