I’ve been writing a lot more about defencemen over the past year than I ever have. I think we’re starting to get a lot better at talking about them with data, which is a very good thing. As a guy who follows the Oilers pretty closely, I’ve been fascinated by the contrast between Jeff Petry and Justin Schultz over the past year or so, given that Petry seems to get better results but Schultz seems to get a lot more love from the pretty much everyone who works for, reports on or follows the Oilers.
Craig MacTavish talked to Jim Matheson today and made some comments in keeping with that trend:
Q: Do you need a first-pairing defenceman, through a trade or free-agent signing, who can play with Schultz? Like a left-handed Dan Girardi? That would be ideal right? Or a Brent Seabrook?
A: I can’t mention any names. I think Justin has enough game. He’s not too needy out there.
Huh. Not too needy. Well, that’s an opinion.
As some of you will know, Corey Sznajder is engaged in a project tracking all of the zone entries from the 2013-14 NHL season. He’s about halfway through it right now and is generating some phenomenal stuff. If, after reading this post, you want to support his work, please consider a donation via Corey’s funding mechanism.
What Corey does is track every zone entry that takes place during an NHL game. I’ve taken Corey’s data and blended it with the NHL’s data for who is on the ice at a given point in time. This enables me see how the zone entries for the Oilers and their opposition differed when Petry was on the ice vs. when Schultz was on the ice. As mentioned, Corey’s provided me with about half a season’s worth of data – I’ve got the entries through the first 38 games of the year.
Why do we care about zone entries? If you’re new to them, I encourage you to read Eric Tulsky’s work here. The long and the short of it is this: the more you carry (as opposed to dump) the puck into the opposition’s end, the more shot attempts you’ll generate. The less often you let them carry the puck into yours, the fewer shot attempts you’ll allow.
What we’re trying to do here is to break the game into smaller pieces. Ultimately, if analytics has a place in hockey, it needs to be able to satisfy the Hall Test: you have to be able to look at the numbers and say to a player “If you do this better, your numbers will get better and we’ll win more.” We’re getting closer to that and this is a step towards it I think.
Zone entry data lets us consider a couple of things. First, by looking at the ratios of zone entry attempts for/zone entry attempts against, we can see whether or not a team is winning the neutral zone battle – are they generating more zone entries than the opposition. Second, they let us see how a team is doing in terms of generating carries – the best type of entries to generate – and preventing carries – the worst type to allow. Finally, although Eric’s work has shown that generating shot attempts given a certain zone entry type isn’t a repeatable skill, it lets us do a check to see whether or not that’s a problem.
When we do this stuff, we’re dealing in Fenwick: shots and missed shots. Blocked shots are excluded. This is the way Eric’s done it and Corey’s building off Eric’s work and I’m building off Corey’s, so I’ll do it the same way.
Here’s how things looked 38 games into the season:
Unfortunately, it’s actually pretty even. Even though Petry ended up slaughtering Schultz in the Corsis, it didn’t occur in the first 38 games of the season. Here’s what the last 44 games looked like.
Oi. Both Schultz and Petry see the rate at which they allow Fenwicks rise dramatically, with Schultz’ rate being much bigger. Petry at least managed to maintain what he was doing in terms of shot attempts generated when he was on the ice; Schultz took a hit there too. The big gap there is responsible for the big gap between them this year.
That’s kind of unfortunate because I’ve just got zone entry data for the first 38 games but I think it’s still pretty intriguing stuff which gives some tantalizing suggestions as to what Schultz’ problem might be and where additional data could help us get to the point that we can pass the Hall Test. I’ll do the Fenwick Against in this post and then leave the Fenwick For to be dealt with in a subsequent post, as this is a fairly lengthy post.
Corey breaks the entries into five types: Carries, Dump-Ins, X (which I assume are unintentional entries), Failed Entries and Faceoffs. The table at left explains the breakdown of entries into the Oilers end at 5v5 in the first 38 game with Petry or Schultz on the ice. I’ve organized it by both raw numbers and then the percentage of each type of entry as a percentage of the whole.
That’s one half of the package. The other half is the shot attempts that happen per entry in the five groups we’ve identified. As I mentioned, Tulsky’s pointed out that when he’s looked at teams, he’s found that the entry type tells us how many shot attempts there tend to be. If he’s right, we shouldn’t see significant differences between Petry and Schultz in terms of how many shot attempts they allow per given entry type.
Here’s the data. Tulsky’s theory holds up pretty well. Petry and Schultz allowed shot attempts at an identical rate when the puck was carried in. They were virtually identical in terms of allowing shot attempts following DZ faceoffs. The samples of X are too small to take anything from.
It maybe doesn’t hold up as well when we look at dump-ins. Schultz is basically conceding an extra six shot attempts per 100 dump-ins. I can’t say with any degree of certainty whether that’s noise or whether there’s something there but it explains about half of the difference between Petry and Schultz in terms of shot attempts allowed. Anecdotally, the internet’s Woodguy grumbled about Schultz all year on the grounds that he’s an ambler who moseys back for the puck and sometimes gets into trouble doing so. This would be a data point that supports that, although I caution that I would take care in evaluating it, given Eric’s broader finding that there don’t seem to be big differences between players once the puck is in the zone.
Still, it’s there and it explains about half the difference in the rate at which the Oilers allowed Fenwicks with Schultz on the ice and the rate at which they allowed them with Petry on the ice in the first 38 games.
Let’s go back to that first table. I’ll print it again in case you’ve forgotten.
You should notice that the split in terms of entry types for Schultz and Petry is different. The opposition carry the puck into the Oilers end more frequently when Schultz is on the ice. Those are the most dangerous kind of entries. Dump-ins and faceoffs have close to the same value, in terms of expected shots against when you’re on the ice for a possession that starts with one of them and they had similar totals of those – both saw around 58% of the opposition’s o-zone possessions when they were on the ice start with a dump-in or a faceoff in the Oilers end.
I note, as I look at this, that a greater percentage of the possessions when Schultz was on the ice started with a faceoff in the Oilers end in the first 38 games. This is a bit of a red flag to me. Relatively speaking, an opposition possession that starts with a faceoff is less dangerous than when they’re coming up the ice at you. Moreover, it’s probably a bit of a fluke. If Schultz ends up with a lower share of opposition possessions when he’s on the ice starting from faceoffs, the share that come from dump-ins and carries and will rise, which means the rate at which the Oilers allow shots when he’s on the ice will go up. I note that this is consistent with the observed spike in his Fenwick Against rate in the final 44 games.
Petry makes up for his lower rate of carries allowed pretty much entirely with more failed opposition entries when he’s on the ice. Obviously, these have an expected Fenwick against value of zero, so more is better.
If you do the math, you find that we can attribute about half of Schultz’s higher Fenwick against rate in the first 44 games to allowing more shots against off dump-ins and half of it to the puck being carried in to the Oilers’ end more frequently when he’s on the ice than it was when Petry was out there.
This seems to me like information that you can use to get a lot closer to passing the Hall Test. It tells me that video review as to why the opposition are able to carry the puck in more effectively with Schultz on the ice may be useful. It also tells me that I may want to review opponents’ dump-ins with Schultz on the ice to see if there are improvements that can be made – maybe the man is, objectively speaking, an ambler.
This also, I think, points towards additional information that would be useful. Tulsky’s started getting into tracking entries based on which side of the ice the entry takes place, which would be helpful here, I think. As you get further into more detailed stuff, it starts to get a little trickier to track – I have a few ideas about things that could be tracked to deal with the dump-in thing but, ultimately, I think that you might simply have to spend time in a video room to really get a handle on that.
It’s a step in the right direction though. Looking at this, I think I have a much better grasp on why the opposition got more shot attempts with Schultz on the ice than with Petry over the first 38 games. There’s a lot to be said for figuring out what matters in a hockey game and then systematically gathering information that lets you evaluate how players are doing at something specific.Email Tyler Dellow at email@example.com