This is part of a series looking for reasons for the Oilers Corsi% collapse in 2012-13 by examining things on a shift-by-shift basis. Part 1 can be found here. Part 2 can be found here. Part 3 can be found here. Part 4 can be found here. Part 5 can be found here. Part 6 can be found here. Part 7 can be found here.
Way, way back on February 13, 2013, I wrote about the disconnect between the Oilers scoring chance numbers at the time (decent) and their Corsi (poor), with a particular focus on Ales Hemsky and Sam Gagner’s abysmal Corsi numbers. Here’s the key part of the post.
Hemsky and Gagner both have a track record of posting respectable Corsi numbers as seen above, so I’d be surprised if this continues. That’s why I’m inclined to believe more in the chances at this point than I am the shooting numbers – I figure they’ll both start to post better Corsis. I do wonder why their Corsi For numbers are so low though and wonder if some of it has to do with Yakupov. Yakupov has a similarly low Corsi For number, although he doesn’t have the big Corsi Against number that Hemsky and Gagner have, which is likely to do with him sitting out some defensive zone faceoffs and late in games when the Oilers are protecting a lead and not pressing.
I haven’t sat down and ground through video or anything, but I have been kind of struck watching the Gagner/Yakupov/Hemsky trio that Yakupov seems to sort of wander around the ice at ES, not really consistently getting into places where the puck might come to him, whether to shoot it or to keep the play going in the offensive end of the ice. There are moments where he finds holes and makes a chance happen but they aren’t as frequent as they will be.
At the time that I wrote this, I didn’t have access to the data that I do now. When I started to pore over Hemsky’s season, I started by looking at how he did in open play. Something quickly caught my eye: his season kind of breaks down nicely into three chunks: Games 1-11, 12-19 and 20+. The piece that I’m referring to happens to dovetail well with the first of those segments. Let’s look at Hemsky’s performance on eight types of shift: Open Play (no faceoffs), shifts with one faceoff, of which we can create six types (OZ+, OZ-, NZ+, NZ-, DZ+, DZ-) and multi-faceoff shifts. There’s too many permutations to really sort those effectively, so I just lump them under one heading. In 2012-13, 10.3% of Hemsky’s shifts were multi-faceoff shifts.
So here’s the data for those eight shift types for games 1-11. Funny thing: Hemsky was actually performing at a virtually identical level to last year in open play situations: 53% to 52.5%. When it WASN’T an open play shift, he and the Oilers were getting murdered. If you look a little more closely, you can see an even more specific problem: the Oilers couldn’t win any bloody faceoffs in the offensive or defensive zone. Hemsky had 8 OZ+ and 21 OZ- shifts. He had 9 DZ+ and 19 DZ- shifts. Gagner winning OZ faceoffs at a 27.6% clip and DZ faceoffs at a 32.1% clip – it turns out that that’s a problem.
Two brief digressions. First, on faceoffs. I, and other analytic types, tend to downplay the importance of winning faceoffs. In my case at least, a large part of the rationale is that the range between guys who are good and bad tends to be small and that it’s not worth getting too worked up about. I am not talking about guys winning 27-32% of their draws in the OZ/DZ when I say that. Those are horrendous numbers and far beyond the range we usually see.
The second is more to do with context. One of the things that’s great about data is that it lets us prove whether or not things we believe to be true are, in fact, true. If they are, it then kind of lets us quantify the difference. Let’s take a look at the difference in Corsi% for the six single faceoff situations I mentioned above and in open play shifts. I’m going to do this by simply taking an average of the Corsi% for eleven forwards who played on the Oilers in 2011-12 and 2012-13: Taylor Hall, Eric Belanger, Jordan Eberle, RNH, Ales Hemsky, Ryan Jones, Ryan Smyth, Shawn Horcoff, Lennart Petrell, Magnus Paajarvi and Sam Gagner.
If I asked you to guess which types of shift would have the highest Corsi%, I bet you would have come out with something exactly like the order that appears in this table. It’s always nice when theory and practice converge like this. If you had this data on a league-wide scale, you could do some great benchmarking – there’s no reason that a clever team couldn’t build a system that would serve as a sort of problem area identifier, directing the attention of the coaches to those areas of the team that are deficient at 5v5 play.
Bringing this back to Hemsky, Gagner’s inability to win faceoffs (and Hemsky played mostly with Gagner) meant that Hemsky tended to be playing less favourable shifts. That will hurt you. This was certainly exacerbated by the Oilers having a shocker, as my British friends would say, when Hemsky was on the ice for a DZ- shift: a 9.1% Corsi% on those shifts!
That right there is a decent data point against my February 13 muttering that it was all stupid Yakupov’s fault, unless he was the guy responsible for the abattoir on DZ- draws. Interestingly, the coaches may have come to a similar conclusion about Yakupov: Yak lost his place with Hemsky and Gagner at about this point. It would be interesting to know what, precisely, gave rise to this decision. Whatever complaints I had about Yakupov looking lost, it doesn’t look like there was a lot of merit to them on open play shifts.
Let’s flip to the second segment, games 12-19. This is an eight game segment, most of which Hemsky played with Gagner. This segment was the cancer at the heart of Hemsky’s season. You can see that he did much better on his non-open play shifts: Gagner started winning some draws, although the Oilers struggled to do much with it (fun question to contemplate that might be answerable: did the Oilers change their faceoff plays to provide Gagner with some assistance?) Hemsky’s non-open play Corsi% through this segment wasn’t terrible – 45.9%.
The open play Corsi% though? Ugly. As I mentioned, the Oilers switched up who played with Hemsky – there was a sort of rotating cast of wingers through this stretch, with Hemsky/Gagner being a consistent duo. The table at left summarizes the data during that period for people on the ice with Hemsky. Nothing worked. The funny thing – in a black way – is that Hemsky/Hall was a hell of a duo in 2011-12 at open play. They were +143 -102 in Corsi events, 58.4% Corsi% in open play. While we don’t know what the best players do at this because it’s a new way of measuring it, I suspect that that’s a pretty awesome number.
The thing with this sort of stuff is that it’s kind of tricky to pick out when it’s noise and when it’s signal. In this case, I don’t know if the Oilers were doing anything differently at this time that might have caused this. Young Willis is working on a project that classifies zone entries for the entire season as either dump-ins or carries. We have some strong indications that dumping the puck in means fewer shot attempts. He’s passed his data along to me and I’ll have a peek at it.
Back on that signal/noise point. As you start to gain experience dealing with this stuff, you sort of learn what the normal variations are. We’ve learned that players experience big, seemingly random, swings in on-ice S% and on-ice SV%. People who are in tune with the analytics don’t get too worked up if a guy has a run with an .875 on-ice SV%, for example. As we do more with this stuff and more of this kind of data becomes available, we’ll start to have a better idea about the size of the swings we should expect and when it’s noise and when it’s signal. For now, this is a bit of an open question, one that I’m going to think about a bit. I tend towards thinking that this is noise but I’m not married to the position.
Let’s move on to the third segment that I identified for Hemsky. It’s funny, the way that this lines up – this segment is from game 20 to 48 and Hemsky fractured his foot in game 23. He played the rest of his season on a dinged up foot. And yet…
The open-play Corsi% comes right back, almost to where it was in the first segment I identified. Hemsky’s centres are still getting murdered in OZ/DZ draws (39.6% and 32.7% respectively) but not quite as badly as in the first segment we looked at. If you think of this in terms of open play and non-open play shifts, Hemsky’s three segments went like this: 52.5%/34.6%, 31.3%/45.9%, 50.9%/44.4%.
Going back to what I wrote about Hemsky and Gagner on February 13, while it didn’t pan out, I think I have some ideas as to why. I mentioned in one of the Gagner posts that his numbers on open-play shifts weren’t that different this year from last year: 48.5% to 46.2%. Hemsky had a much wider spread: 53% to 46.7%. I suspect that playing another ten games would have done a lot to narrow the gap in his case. An already short season made shorter by an injury will amplify the impact of a period of struggle (or an unusually good period). Sample size and all that.
I think that there are some implications here for how we think about Corsi%. There’s a general consensus that extreme ZoneStarts will affect it and I think that’s right – I’ve reposted that table that I posted way up above how the Oilers Corsi% declined from OZ+ down to DZ-. You can infer from that that offensive zone faceoffs are better than neutral zone faceoffs which are better than defensive zone faceoffs, at least for the Oilers over the past two years, although I strongly suspect that this is a principle of general application.
You can also see that there is a real price to be paid in terms of winning a faceoff versus losing a faceoff. In 2011-12, it was about ten points of Corsi% in the OZ and NZ. In the DZ, it was about five points. This year was weirder – on a team level, the price you paid for winning versus losing on OZ faceoff was smaller and it almost disappeared in the NZ. In both cases, it wasn’t because the Corsi% associated with losing the draw changed much – it’s because of a decline in the Corsi% following wins.
In effect, I think there’s two things going on here with Hemsky: there’s a decline in the percentage of shifts that are faceoff win shifts as opposed to faceoff loss shifts and then there’s a decline in what was done with the win shifts. Here’s a table setting out Hemsky’s percentage of each type of shifts in 2011-12 and 2012-13.
You can see that Hemsky was on the ice for more OZ and DZ losses this year and fewer wins. That doesn’t help the Corsi%.
Then there’s the issue of results in those game states. Gagner and Hemsky are sort of intertwined this year because of how much they played together. Here are their numbers for each of the seven game states (I”ve left out multi-draw shifts; they’re a relatively minor component):
Huh. So they both suffered massive dips in their Corsi% on OZ+ draws. Gagner collapsed in terms of his Corsi% on OZ- and NZ+ draws as well. Hemsky experienced declines in his Corsi% in each state, notably on DZ losses.
Again, there’s a lot to tease out here – the samples are small and it’s a question of figuring out wether or not we’re seeing signal or noise. This process is, for me, one of examining data in different ways and then coming to conclusions slowly (which must be frustrating as hell if you’re reading this, looking for answers) but I’m pretty convinced that they made a change on offensive zone draws that hurt – I’m going to write about this in the coming days. I haven’t gotten into the NZ or DZ stuff yet. All to come.
If you listened to MacT at the end of the year, it sure sounded like Hemsky won’t be back next year. His answer to the Hemsky question sounded a lot like his answer to the Khabibulin question and you never want your future discussed in the same tone as that of the perpetually injured backup goaltender who represents one of the worst contracts in Oiler history.
I think Hemsky can still play and I don’t think that the Oilers will get much in the way of a return for him. For that reason, I hope that they keep him and look to move someone else. If he does have to go, I sure hope he goes somewhere decent – I think there’s a pretty decent chance he could put up some excellent numbers if he was playing with a guy who could win draws.Email Tyler Dellow at firstname.lastname@example.org