• Big Oilers Data VII: Sam Gagner II

    by  • May 9, 2013 • Hockey • 13 Comments

    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.

    So, in the last post in this series, we learned that Sam Gagner really didn’t have results that were that different from his results last year on 5v5 shifts that didn’t feature a faceoff and that the gap would probably disappear if this nagging new issue that we’ve identified with difficulty in generating multi-SAF shifts were to be disappear. If you’re familiar with the dark algebraic arts, you’ll intuitively know that if we’ve identified 45% of the game or so where Sam was pretty similar to last year, there must be a really bad spot somewhere else.

    In other words, if we know how Gagner did overall and on shifts without faceoffs, the magic of subtraction lets us see how he did on shifts on which there were faceoffs. I’ve used such techniques to produce the table below.

    There really isn’t much to say other than “Wow.” Gagner’s results on shifts on which there was a faceoff changed from those of a guy sawing it off in the best league in the world to those of a guy who would struggle to stay in the NHL. The Oilers had an overall Corsi% of 44.5% this year after posting a 47.3% last year. If Gagner had again posted a 49.2% Corsi% on shifts that involved a faceoff and all other things had stayed the same, their Corsi% would have been 45.9%. Basically, there’s half of the Oilers’ decline, all found in 12.2% of their 5v5 minutes.

    An aside: I assume that this is obvious, but in case it isn’t, there will be many instances that you can point out as being a large part of the decline; there will be declines and gains and the real issue when you’re trying to improve dramatically is how it all nets out. In the Oilers’ case when you add up the good and the bad at 5v5, it was a net negative this year. I’m interested in identifying the particularly dramatic contributors to that. Like Gagner’s weird collapse on shifts featuring a faceoff.

    We know both that Gagner’s share of faceoffs in the defensive zone increased this year and that he had trouble winning draws in the neutral and defensive zones relative to 2011-12. Part of what we’re seeing undoubtedly reflects the fact that the ice was tilted against him more this year than it was last year and that he didn’t do particularly well in the neutral zone and the defensive zone faceoffs in comparison to last year.

    As it so happens, I’m pretty sure that there’s more to it than that though. One of the cool things about examining data this way is that I can look at players in very specific circumstances. Let’s look at Gagner’s shifts this year that featured a) exactly one faceoff and b) that faceoff took place in the offensive zone. In fact, let’s subdivide it further and look at two kinds of shifts: a) exactly one faceoff, b) that faceoff took place in the offensive zone and c) whether or not the Oilers won or lost that faceoff.

    I’ve summarized the relevant data in the table above It’s kind of amazing. Basically, in 2011-12 when Gagner was on the ice for a shift where a) there was exactly one faceoff, b) that faceoff took place in the offensive zone and c) the Oilers won the faceoff, the Oilers got 71% of the shot attempts on that shift. This year? 50.6%! What if we change one of the conditions and look at just shifts featuring a single offensive zone faceoff that the Oilers lost? Similar drop: 62% of the Corsi events to 47.5%.

    What could cause this? Three things. The first, randomness, gets tougher to suggest with changes this large. The second would be a change in personnel this year on the Oilers and in terms of who Gagner was playing with. He took some shifts in the offensive zone on the wing last year with Belanger. This year Yakupov was new. I’m not inclined to accept this as an explanation for a couple of reasons. First of all, there was just a teamwide collapse in terms of Corsi% – it’s hard to blame Yak for other lines, with whom he played less, disappearing into the vortex. Second, the Oilers did kind of move Yak down the batting order at some point and it didn’t really solve Gagner’s problems (which, by extension are kind of Hemsky’s problems too – they played together a reasonably large amount but we’ll get to Hemsky eventually).

    That leaves a third possibility: the Oilers were doing something different in the offensive zone with faceoffs. This requires watching some video, which I’ve started to do and I have some ideas, but it will take some time to answer. I think that this is probably the most likely outcome because, frankly, it makes more sense than basically everyone, young and old, getting worse at once.

    Let’s take a peak at our old friend, shifts with X SAF and shifts with X SAA and see what it has to say about all of this. We’ll start with shifts that included 1 (one) offensive zone win.

    So in 2011-12, 60.6% of his shifts with one offensive zone win featured at least one SAF. This year, that dropped to 50.8%. 51.5% of shifts with at least one SAF in 2011-12 were multi-SAF shifts. This year, it was down to 37.5%.

    The Oilers didn’t allow an SAA on 66.1% of Gagner’s shifts that featured at least one offensive zone draw in 2011-12. In 2012-13, that number fell to 50.8%. They were more likely to allow a multi-SAA shift this year as well, although I’m not sure I’d attribute that to some sort of defensive zone failing. I’ve been re-watching Gagner’s shifts and by, eye, the Oilers just seemed to go up and down the ice more on Gagner shifts that featured an offensive zone faceoff this year. Less offensive zone pressure, equals easier to get down the ice, get a shot attempt, Oilers recover, move puck up, less offensive zone pressure, another shot attempt against. If they were better at keeping the puck in the offensive zone, I suspect there’d be less of that. The data isn’t granular enough to know precisely what causes that bump.

    We see similar trends if we look at shifts that featured exactly one offensive zone loss. In 2011-12, the Oilers got at least one SAF in 53.4% of their shifts with Gagner on the ice. This year, that number fell to 38.8%. They were even worse at turning those shifts into multi-SAF shifts – from 45.5% of shifts with at least one SAF becoming multi-SAF shifts in 2011-12 to 25% of those shifts becoming multi-SAF shifts in 2012-13.

    The SAA angle here provides a little bit happier news. Some weak, cold light from an otherwise abysmal set of numbers. With Gagner on the ice and exactly one offensive zone faceoff in 2011-12, the Oilers allowed at least one SAF on 51.4% of their shifts. This year, that number fell to 43.2%. So that’s good. Unfortunately, if there was at least one SAA, there was a multi-SAA shift 21.1% of the time in 2011-12 and 37.5% of the time this year. So that’s bad.

    I’m gonna cut off the data part of this post here – I want these to be readable rather than being undigestible solid masses of numbers and I suspect I’ve pushed the envelope a bit on that with some of the posts in this series. That doesn’t mean that I won’t talk more broadly about things for a minute though.

    Now, keeping in mind that I haven’t found an explanation for why this happened with which I’m satisfied yet (Did we need more Ben Eager with Gagner?), I’d be awfully interested to know whether or not the Oilers were aware of this. Nobody that I know of is breaking down things like this in the public domain. The magnitude of this change is pretty massive – I mentioned previously that Gagner fell from a 48.9% Corsi% in 2011-12 to 43.1% in 2012-13. The change in Corsi results on his shifts involving exactly one offensive faceoff alone account for 2.7% points worth of Corsi. That’s a massive change.

    How massive is it? Well, in 2011-12, he played 75 games, which included 1053.87 of 5v5, non-goalie pulled time. 914 SAF and 956 SAA. Now drop that by 2.7 points from a 48.9% Corsi% to a 46.2% Corsi percentage. That gets you 864 SAF and 1006 SAA. A reasonable rule of thumb is that 50% of SA are actual shots. So at 48.9%, we’re talking about getting outshot 478 to 457. At 46.2%, it’s 503-432. Let’s assume that 8% of those are goals. At 48.9%, you get outscored 38.2 to 36.6. At 46.2%, you get outscored 40.2 to 34.6. Four goal difference. Off one tiny part of one guy’s game. That’s pretty big.

    It was apparent that something unusual was happening with Gagner and Hemsky pretty early on in the season. I wrote about it on February 25, 2013, using a variant of the shift-by-shift approach that I’ve employed here. This is what I said at the time:

    To me, there’s something not right going on from the Oilers’ blue line north when Hemsky/Gagner are on the ice, although I’m not sure what it is. I’ve joined in with others in kind of hammering Gagner for his defence and praising RNH and I acknowledge RNH has played tougher minutes, but it’s tough to find a lot of difference between them in the defensive zone when you look at is this way.

    My theory as to the offensive troubles of Gagner/Hemsky is tied to Yakupov, mostly in terms of him being 19 and needing to learn how one goes about creating ES offence in the NHL as opposed to in junior, where first overall draft picks can create it all on their own. In Yakupov’s defence, it’s not like things have gone any better without him, although it’s a smaller sample that mostly consists of d-zone faceoffs and periods when the Oilers were defending a lead.

    This is a pretty important question though. Why aren’t the Oilers comfortably north of 50% in terms of Fenwick Close and looking pretty good for a playoff spot? Hemsky/Gagner are getting hammered. Between 2009-12, it was a sort of interesting theoretical issue, something that, like about ten other things, was acting to keep them out of the playoffs. That’s no longer the case.

    I didn’t take the next step when I did that analysis and start breaking things down into shifts with faceoffs and without. Frankly, it didn’t occur to me at the time. With the power of a spreadsheet that has all of Gagner’s shifts date and time stamped, I can do that now. His numbers were pretty roundly awful, albeit in small samples: 43.1% Corsi% on shifts without a faceoff, 38.7% on shifts with at least one. If I’d been in a position to compare that to the year before (48.5% and 49.2% respectively), that might have caught my eye.

    I concluded that post with this:

    A technologically advanced team could do analysis like this a lot more quickly than I have and then pair it with work being done on the qualitative side of the operation. It took me a while to assemble this data – a team that invested in having a proper piece of software built could pull this sort of thing up with a few keystrokes and then compare Gagner/Hemsky this year to themselves in past years and see if they could identify the issue. They could also make a pretty informed guess about whether this is just noise (which is probably tied with “something to do with Yakupov” on my list of theories) or something more important. They could work with the video coach on what he sees when things have gone well and poorly and try to find a way to get their numbers back on the rails.

    I’ve got no doubt that the Oilers have talented people looking at video and such. What I’m less sure of is that you can pick out a problem like this by watching, no matter how well you understand hockey. Bill James pointed out many years ago that the difference between a .275 hitter and a .300 hitter is one hit every two weeks; good luck catching that with your eyes, no matter how good you are, unless you’re watching every game. Even then – there’s so much other stuff going on that it kind of distracts the eye. Having data like this that you can incorporate into things gives you a much more solid base from which to begin – you can identify specific issues of the game where things are working or aren’t working. It’s kind of mind boggling to think of having this kind of data on a league wide level, for the entire era in which the NHL was gathering the information necessary to crank this out.

    I don’t know what sort of data the Oilers have available to them or how good they are at manipulating it to find something like this. I’ve started reviewing some video of Gagner’s shifts with offensive zone wins to see if I can see anything. I’ve done something like this before with the San Jose Sharks PP and it’s amazing how stuff leaps out at you when you watch video like this – Gagner had 63 shifts with exactly one offensive zone win in 2012-13, just over one per game. It kind of slips through your brain when you’re watching games.

    When you sit down and just cue them up, you immediately notice things, like how Hemsky takes the inside lane on every draw (at least early in the year) and Yakupov was lining up closest to the boards. When the puck was dropped, they’d both enter the circle, usually with Hemsky retrieving the puck and ending up on the wall, with Yak in the slot and Gagner dropping low into the corner to provide support to Hemsky. It’s obvious when you watch it this way; not so much when you just review a hockey game without looking for anything in particular.

    Here are some screen shots of what I’m talking about. This is one of the few faceoffs Sportsnet showed this year (just barely). This is with 8:35 left in the third period of a game that the Oilers are leading 2-1 against Vancouver. It’s the thir and they’re up, so Yakupov has been replaced on Gagner’s wing by MPS – Hemsky is still there. As the puck is dropped, Hemsky and MPS crash the circle.

    MPS picks up the puck and Hemsky traverses through to the boards.

    MPS is now in the slot, with Hemsky on the wall. Gagner hasn’t really moved yet.

    MPS wheels and shoots. The rebound is kicked into the left corner and Hemsky’s on it.

    That’s the gist of the faceoff play that the Oilers were running early in the season – I’m up to about game 233 and they’re pretty consistent in running it, although Gagner usually heads to the corner more aggressively. We’ll see if they got away from it as the season went on – that might be a sign that they had identified this problem.

    I’m pretty sure that I’m going to come up with something approximating the correct answer eventually as to what was happening (seems to me that the way they’re running faceoffs, they’re guaranteed to have nobody in front of the net if there’s a shot in the first few seconds but we’ll see), but for someone doing this with knowledge of what the Oilers were trying to do on offensive zone faceoffs in 2011-12 versus 2012-13, this would be an even easier task. It’s a heck of a thing though – I’m still trying to figure out precisely the right way to kind of organize the information I’m gathering from watching it to understand their difficulties with Gagner on the ice this year.

    This series has been, in a way, proof of concept. The concept works, I think.

    (Next up: either something on Gagner and shifts with faceoffs in the defensive and neutral zones or some discussion based on video. We’ll see what I get to.)

    Email Tyler Dellow at tyler@mc79hockey.com


    13 Responses to Big Oilers Data VII: Sam Gagner II

    1. bparsonsproject
      May 9, 2013 at

      You totally need to be the guy at their computers working their video. You’re passionate about the team AND you have a pretty good idea on how to break this down. I find it fascinating, hopefully the oil figure out the problem, I’m getting tired of watching them fall apart.

    2. Daniel
      May 10, 2013 at

      Thanks for the enormous amount of work you’ve been putting into this series!

      I did want to raise a concern with this post, however, which is that as you continue to subdivide the game of one player, especially over a shortened season, I think small number effects can easily creep up on you. For instance you point out that the percentage of OF+ shifts with at least one SAF dropped from 60.6 to 50.8, which is a big swing, but actually only corresponds to 6 shifts (32/63 instead of 38/63). If you had a coin that came heads 60.6% of the time and you flipped it 63 times, then you would get 32 or fewer heads about 7% of the time. Not common, but not uncommon (in particular if you ran this test for say 12 forwards, you would expect one of them to have a drop this big by random chance). Anyway, I don’t want to say there’s nothing here – I think you’ve showed that there is something going on with or near Gagner in this part of the game, but once the numbers get this small I don’t find the data as compelling. I wonder if there’s any way to look for other players in other seasons that seem to have had their stats suffer in a similar way, and whether any inferences can be drawn from that (maybe it tends to happen under new coaches, or with guys that are paired with rookies, etc).

      But, thanks again for the work you’ve put in!

    3. Mark-LW
      May 10, 2013 at

      Stauffer, can you lobby MacTavish to hire Tyler?

      • Mark
        May 12, 2013 at

        Stauffer’s using his clout to get Daum hired as an assistant coach. Besides, if Tyler works for the Oilers, what will I read after Lowetide?

    4. don_cherrys_conscience
      May 10, 2013 at

      Tyler, awesome work.

      Don’t fall prey to working for “The Man”….or in this case, “The Oil.” Your instincts will never be as pure as they are right now, in terms of pinpointing what went wrong this past season and sensing what data may prove it. When the salary starts rolling in, Uncle Craig will starting asking if you can find/shape data to justify a new contract for Lennart Petrell, or some such nonsense.

      Maybe a consultant, with you maintaining the rights to your work, defining most of what you want to research, but giving the Oilers an advanced peek prior to publishing?

      P.S. – Big ups on the Klopp reference in your previous post. Perhaps he is a German version of a young Arsene Wenger?

    5. don_cherrys_conscience
      May 10, 2013 at

      Tyler, awesome work.

      Don’t fall prey to working for “The Man”….or in this case, “The Oil.” Your instincts will never be as pure as they are right now, in terms of pinpointing what went wrong this past season and sensing what data may prove it. When the salary starts rolling in, Uncle Craig will start asking if you can find/shape data to justify a new contract for Lennart Petrell, or some such nonsense.

      Maybe a consultant, with you maintaining the rights to your work, defining most of what you want to research, but giving the Oilers an advanced peek prior to publishing?

      P.S. – Big ups on the Klopp reference in your previous post. Perhaps he is a German version of a young Arsene Wenger?

    6. Woodguy
      May 11, 2013 at

      The drop in SAF and increase in SAA after face offs is very startling.

      Its not just the lack of SAF after a win, but the ridiculous increase in SAA after a win too. Its massive.

      Their set plays much have a very low puck recovery rate.

      Either the wrong system or the wrong players.

      If its the wrong system and the coach didn’t adjust, then its the wrong coach.

      • David Supina
        May 11, 2013 at

        It might be a coaching/systems problem, but I think we need to give Krueger the benefit of the doubt… for now. It might be more of a problem with execution of a new system, due to a very short training camp, and compressed season that Krueger specifically chose to reduce practices during.

        Basically, it might be an effective system that the players simply have not had time to learn how to implement well.

        Though I do admit, hiring a more proven guy like Tippet is tempting.

    7. Tyler Dellow
      May 11, 2013 at

      I’m a Krueger guy, mostly because I think he’s a genuinely smart guy and I like smart guys because I think they tend to figure things out. Or, I suppose, they tend to make catastrophic mistakes that a dumb guy would never get in a position to make.

      I’ve got some more stuff coming and to me, it does raise some questions about the tactics and whether the OIlers have sufficient systems in place to identify this stuff in-season. Stauffer’s kind of pushing the idea of another coach on staff, a technical guy. I don’t know if it’s a coach that they need or data but they definitely need some sort of a process to identify when things are off the rails.

    8. David S
      May 11, 2013 at

      If I’m Krueger, I take a look down the bench and realize my D is my biggest problem. I can’t tell you the number of times I saw Sam execute a play that resulted in that first SA, then peel off and float above the FO circle. I also just about ground my teeth to dust this year watching as the Oilers were absolutely incapable of holding the puck in the O zone, resulting in immediate opponent rushes the other way while the D were on yet another inexplicable mid-play line change.

      I think there’s a couple of factors in play here. (1) The players were asked to make changes so radical they couldn’t fully implement (to the point of being instinctive) during a brutally short season with literally no real training camp. (2) The system was built to compensate for the weak D and was unsustainable down the stretch because it over-tasked the players.

    9. Lloyd B.
      May 11, 2013 at

      Nice breakdown and you say this was what they were doing consistently. Pretty easy to see why there would not be a second SAF on this play. None of the players are even facing the net until MPS shoots. When they finally do turn the right direction they are heading into the corner to get the puck! (In fairness though, it does appear to have confused the opposition, as none of them are heading that way.) The Oilers are all above the face off dots until after the shot ! No one heads to the front of the net to cause a screen or get a rebound. More and more I’m leaning toward the system the new coach is trying to implement. But can it really be …. head in the wrong direction to confuse the opposition and then take a long range shot to score on a surprised goalie? Nah…couldn’t be.

    10. JS
      May 14, 2013 at

      This is a great effort. A couple of thoughts …

      ” Like Gagner’s weird collapse on shifts featuring a faceoff”. Wonder how much of this can be attributed to opposition? When there is a draw, at least half the time (road games) the opposition can define who you play against. And in home games, if they want to they can often quickly make the switch. I don’t follow the Oilers closely enough to say whether or not Gagner rates that kind of attention, but it might mean something. In baseball, if Josh Hamilton batter over half the time against #1 or #2 pitchers, who were left handed, he”d be in A ball. Baseball can’t do matching at the same level as hockey, other than one AB per game. On the other hand, if matching was part of the issue with Gagner, I would surmise it would be true of other players who rate matches. Actually, I think that similar data on Henrik Sedin would be fascinating, because the last several years I doubt any centre saw a wider gap in matchups between home and away. AV consistently got him the best starting spot on the ice, and in road games, because the Canucks had no other centres, he ALWAYS saw the best opposition. Although Gagner is not the #1 centre, is there possibly some type of effect from this? Dunno. There’s a lot of meat on this bone.

      The effort here is wonderful. I bought James first Abstract back in ’76, and own ad still read all of them. I was going to say that the state of hockey-metrics (fancystats?) was about where he was with baseball circa 1978. A way off from national understanding and acceptance, but noticeable to those with an interest. He got big after publishing the 1980 Abstract. Having read this series, though, you’ve taken it a year or two further. It’s one thing to be counting Corsi and faceoffs and everything else; it’s a whole separate ballgame to apply the numbers in this way. In the fullness of time, this kind of data will be available for all teams, just as baseball eventually started to provide useful information. It takes forever, but the end result is awesome.

      Keep up the excellent work.

    11. Pingback: Sam Gagner: The more things change… | Edmonton Journal

    Leave a Reply

    Your email address will not be published. Required fields are marked *