Penalty Killers who Take Penalties
by Hawerchuk on Oct 5, 2009 3:43 AM EDT
Email
Print
The Falconer posed an interesting question last week: "[what is the] effect of a PK going to the box? I did a post on "costly penalties" a couple of years ago. J.P. read my post and noted that it seemed like whenever a member of the Thrashers PK unit took a penalty the opposition was more likely to score. Sure enough that is the case, the opposition PK% rose to something like 25% when an Atlanta PKer went to the box. I'd love to see a breakdown of this for the entire NHL..."
I'm going to approach this from a slightly different perspective - if a regular penalty-killer goes to the box, how does that change the rate of shots allowed by the penalty-killing team? Because there are so many more shots than goals, shot rate gives you a better sense of the opponent's offense.
So far, I ran this analysis for 2008-09 - I picked out every forward who played more than two minutes at 4-on-5 per game, played more than 70 games, and didn't change teams. I then looked at the following quantities:
1) Shot rate against at 4-on-5 when this group of players was in on the ice.
2) Shot rate against at 4-on-5 when this group of players was not on the ice - but not while they were in the box and only in games that they actually played in.
3) Shot rate against at 4-on-5 when this group of players was not on the ice - but only in the 120 seconds following one of these players taking a penalty.
And the results:
| On-Ice ShA/60 | Off-Ice ShA/60 | TOI On/G | TOI Off/G | |
| In Box | 78.7 | 0:18 | ||
| Not In Box | 72.5 | 70.8 | 2:37 | 3:46 |
We see several effects here - the first is that the opponent's shot rate is substantially higher when one of these penalty-killing forwards is in the box. So The Falconer's insight is correct - when a first line PP unit goes up against a PK unit that's missing one of its top players, we should expect more shots on goal.
The second might be surprising: the shot rate against the our top penalty-killers is worse than against their lesser teammates. But this is a reflection of the drop-off in talent on the opponent's 2nd-line PP relative to the 1st line, as opposed to the PK unit's performance.
***Update***
I mentioned below in the comments that PK save percentage is mostly luck. To elaborate on that - if you take the individual team save percentages at 5v4 over the last two seasons (just G and SOG, but excluding missed shots), the average shooting percentage is 12.67%, with a team standard deviation of 0.01892. If I run a simulation of random shooting over 10000 seasons - where I assume every shot has a 12.67% chance of going in - the standard deviation of shooting percentage is 0.01603. So 72% of the variance of shooting percentage is due to luck, and 28% to other factors.
You can read much, much more about this at Objective NHL.
19 comments
|
0 recs |
Do you like this story?
Comments
Good stuff, but I wonder whether it still makes sense to look at goal rates — shot counts, especially on the power play, can be deceptive — and I might like to see the goals and shots side-by-side, if it’s easy enough.
But yeah, it’s always a bit distortive — the Pahlssons of the world will see the world’s best PP units, and that will make them look sour, but coaches hopefully know what they’re doing to some degree.
http://www.battleofcali.com/
I don’t think it makes sense to look at goal rates. Shooting percentage is highly-influenced by luck, while shot volume is much more of an inherent quality of a team.
Luck? I suppose, but it seems that on the power play, there’s opportunities to improve probable shot accuracy — I don’t know what harm there is to see both stats — you can call it luck, I might read some skill into it.
http://www.battleofcali.com/
I’m not sure I follow the explanation. The fact that power play scoring is not well related to the raw number of shots taken has to be written off to luck? I’m not the math-speaky guy, but the fact that all shots are not equally valuable is not surprising to me (maybe that’s not what you’re saying, but then I’m confused) — that’s kind of why I’d like to see goal totals next to shot totals to help me get a sense of shot selection.
http://www.battleofcali.com/
If 30 teams took 500 shots a year on the PP and they were all equally likely to go in (12.7%), we’d see a variance in shooting percentage that’s 70% of the variance that we actually see. The remaining 30% is due to a lot of other things, but mostly goaltender skill. There isn’t a lot of proof out there for deterministic shot selection at the team level (I’ve looked at a few cases where I believe it exists).
What there is is a lot of proof that teams have a fair amount of control over the number of shots they take (which is a proxy for puck possession.) And good teams tend to take a lot more shots as opposed to taking them from closer to the net. This is not true of individual players, obviously.
If 30 teams took 500 shots a year on the PP and they were all equally likely to go in (12.7%), we’d see a variance in shooting percentage that’s 70% of the variance that we actually see.
I don’t expect every shot to go in with the same probability, though, so I guess I’m interested in which teams benefit from these discrepancies. It may suggest nothing but luck, or it may suggest some notion of “taking better shots” or “setting better screens” or whatever — while it may resemble noise, I’d still be curious who it favors.
http://www.battleofcali.com/
Even if you create a distribution of shooting percentages from 0 to 100% that maps to the same likelihoods as in the NHL, you get the same answer: 70% of the variance in shooting percentage is luck. Shots are much more reflective of past and predictive of future offensive performance than goals – that’s why NHL teams count shots and/or chances.
Sure, but there’s also a reason I don’t ask for shot totals when I want to find out who won a game. I understand why data density matters, and why shots are more stable, but flat-out-ignoring goals seems premature — we don’t have a strong enough handle on shot quality to say that we can tell who won a game or what the score was based on shots taken.
So yeah, I don’t mind one bit that we look at shots — that does tell us a lot, but I think goals still tell us something additive to that story.
http://www.battleofcali.com/
When you’re counting shots for purposes of this discussion, are you only counting SOG or are you including missed and blocked shots?
Never lie, steal, cheat, or drink. But, if you must lie, lie in the arms of the one you love. If you must steal, steal away from bad company. If you must cheat, cheat death. If you must drink, drink in the moments that take your breath away.
I would like to see some statistics that deal with the shots and goals scored during shorthanded situations, but in terms of minutes shorthanded, not times shorthanded. Giving the coaches the benefit of the doubt that they put the best person for the job out there when possible, it might be helpful to compare the number of powerplays to powerplay time, to see if they are shortened by offsetting penalties or by goals.
Do you have a source for this data or do you have to scour oodles of boxscores? If the answer is boxscores, where can you find a better analysis for depth than just the boxscores in the newspaper (or yahoo…)?
Jon Casey fan since '84
here are rate stats at every strength for every team:
http://www.behindthenet.ca/2008/team_data3.php
(Change the 2008 to 2007 or 2009 to get other seasons of data)
This is a really good analysis. I’d be interested to look at each team individually as well. If the expected decline in shot prevention does not occur with a regular penalty killer in the box, it might indicate that the coach is making a poor choice with respect to his PK personnel. We could also see how much variance there is from team to team since what we have here is the averaged rate. One would think that this effect would be more devastating for some teams than others. For example, teams that use four forwards on the PK would be hurt more than those that use six.
34 of the 58 players (59%) had a lower Shots Against rate when they weren’t in the box. The average PKer in the sample had 205 minutes on vs 23 minutes off.
That’s not as much unanimity among the results as I was expecting. Do you think the other 40% were lucky (23 minutes really isn’t much)? Do they have better teammates? Was it typical for teams to have one player on each side or was it more likely for all of them to have a lower (or higher) Shots Against rate? If 41% of players don’t have this effect at all, is the effect much larger than 8 shots/60 for the other 59%? Is it fair to make this generalization if only 59% of the players are effected? Questions! (sorry, I’m just fascinated by the subject).
by Scott Reynolds on Oct 6, 2009 12:52 PM EDT up reply actions
I think 59% is pretty reasonable given the luck involved in the 23 minute (off-ice) sample.
Here’s a piece from Tyler Dellow on goaltender save percentage: http://www.mc79hockey.com/?p=3181#more-3181, where he finds that a goaltender who had a 0.5%-1% higher save percentage than another – over 5500 minimum shots – was only 58% likely to retain that over the next 2000 shots. Getting 58-59% confidence in a statistical measure in sports is pretty good!
I’ll look at the + or – distribution by team. Also, I should run this for defensemen – perhaps their impact is more obvious; they also take a lot more penalties so we’ll have a larger sample size on the “off”.
Good points Hawerchuk. I still wonder if there are some other factors that would group certain players together. The defencemen suggestion is good too, especially since very few teams employ more than four regular PK defenders whereas many teams use six forwards.
by Scott Reynolds on Oct 6, 2009 1:21 PM EDT up reply actions
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Certain photos copyright © 2012 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
