• NJ and PHI: Explaining the Devils’ Post-DZ Faceoff Win Results

    by  • May 5, 2014 • Hockey • 11 Comments

    There’s a guy by the name of Corey Sznajder who runs a blog called Shutdown Line who is working on a project that will provide an incredible amount of data to people like me that results in posts like this. Corey is watching all 1230 games this year (yes, even the bad Oilers games) and recording data for zone entries. He hopes to have it finished by July.

    As you can imagine, gathering this data takes an outrageous amount of time that Corey, a university student, could be devoting to more sensible pursuits. I’ve done tracking work before and, candidly, it sucks. You can generate phenomenal insights once the work is done but actually generating the data is unbelievably tedious. This is particularly true given the tools that NHL.com has – basically, a bad video player that’s incredibly buggy. It will result in phenomenal insights but it’s awful to generate.

    If, having seen the post that follows, you think that this is something you’d like to support, please consider donating a few bucks to Corey through his funding site at Go Fund Me. It’s an initiative well worth supporting and, as I’m hopefully about to illustrate, Corey’s efforts are going to lead to a lot of cool analytics work, regardless of what team you support.

    The New Jersey Devils are a fascinatingly weird team. Their last two seasons have featured phenomenally high Corsi% (55.9 and 54.4) with no playoff berth to show for it. Players go there and turn into possession stars. Players leave New Jersey and implode. There are very few shot attempts at 5v5 in Devils games: their 90.4 SAF+SAA/60 over the past two years is 30th in the NHL. 29th is Minnesota at 99.8. New Jersey’s playing a different game.

    Over the weekend, I wrote something looking at how teams did after they won defensive zone faceoffs at 5v5 in 2013-14. One of the things that jumped out at me was the sheer lack of shot attempts for both teams in Devils games after they won a defensive zone faceoff. New Jersey generated 0.111 shot attempts per DZ faceoff win and allowed 0.106, a total of 0.217. That 0.217 is the lowest in the NHL. Chicago was 29th at 0.318 and closer to the team in first (Winnipeg at 0.410) than they were to the Devils.

    So it’s sort of the same old story with New Jersey. Nothing happens in their games. Why? Well, here’s where Corey’s data helps. He gave me the data for the Devils and the Flyers. Philadelphia was in the middle of the league in terms of shot attempt volume following a defensive zone faceoff win, with 0.360 shots per defensive zone faceoff win. They’re a sensible club to compare the Devils to in order to understand what makes New Jersey different. Corey has 28 Devils games done and 30 Flyers games, which gives me a pretty good taste of what’s going on.

    At this point, if you’re not interested in the guts and the reasoning of this but are interested in the conclusions, feel free to skip down to the second set of asterisks.

    * * *

    What Corey does is note each zone entry attempt, the time it occurred, how it occurred, which team makes it and how many Fenwick events (shots on goal + missed shots) occur off that zone entry. So, for example, if the Devils dump the puck in with ten minutes played in the first period and generate a shot on goal and a missed shot before the puck is cleared or there’s a whistle, Corey notes that. I (or any other half clever monkey) can then spin that data into something that tells us about what was different between New Jersey and Philadelphia.

    New Jersey won 159 5v5 DZF in those games and Philly won 208. My method of looking at the impact of defensive zone faceoffs involves looking at what happens in the 21 seconds after the faceoff is won because that’s how long I’ve found that there’s an impact on shot volume and who gets the shots.

    The first thing I want to look at is the shot attempts given up before the puck was cleared from the defensive zone by the Devils and Flyers. Even when you win the draw, you’re going to give up some shot attempts before clearing the zone, because you’re going to turn the puck over sometimes. New Jersey gave up 11 shot attempts from those 159 5v5 DZF wins. For Philly, it was 23 shot attempts allowed off of 208 DZF wins. So Philly’s giving up 60% more shot attempts than New Jersey from the moment the puck is won until the zone is cleared.

    OK, so now the puck’s in the neutral zone. Just as a bit of a sanity check, I wanted to see how many total attempted zone entries there were following a defensive zone win for the Devils versus a defensive zone win for the Flyers. The puck has to be somewhere, so these numbers should be roughly the same – it’s not like Devils games just involve the puck sitting still in the neutral zone while ten guys stare at each other and 4,000 fans doze in the stands. Sure enough, it works out: there were 0.76 attempted entries in the 21 seconds following a Flyers defensive zone faceoff win and 0.83 for the Devils.

    It’s actually going to be closer than that in reality – my method involves cutting things off if there’s an intervening faceoff in the next 21 seconds. There are more shot attempts in Flyers games, which means more frozen pucks, which means more faceoffs, which means things are closer than they appear.

    The Devils have a big edge over the Flyers in terms of their share of those entries. For the Devils, it’s 67%. For Philadelphia, it was 58%. This doesn’t matter from the perspective of the combined shot attempt rate – whether the puck is in New Jersey’s end or their opponent’s, there are chances for shot attempts. It does tell us that part of the reason that the Devils posted a better Corsi% than the Flyers in these situations is that they did a better job of winning the neutral zone.

    Just for a comparator: if I look at the overall data for the Flyers and Devils, Philadelphia got 49.6% of the 5v5 zone entries in their games and New Jersey got 50.3%. So in this specific situation, when you might reasonably speculate that the game is going to be more structured, New Jersey does much better.

    Now – let’s get into entry types. One of the big findings of zone entry research is that pucks that are carried into the offensive zone result in more shot attempts than pucks that are dumped in. Corey breaks zone entries into four different types: controlled entries (ie. pucks that are carried or passed in), uncontrolled entries (dump-ins), failed entry attempts and something called X, which is a sort of grouping of things that don’t fit elsewhere.

    So this table summarizes things from an offensive perspective for the Devils and Flyers.

    Controlled entries should be the objective for teams. New Jersey and Philadelphia were virtually identical in terms of the percentage of their entry attempts that resulted in controlled entries: 33.7% for New Jersey and 32.6% for Philly. There’s a stunning difference in terms of the rate at which they’re generating Fenwick attempts off zone entry attempts though: the average zone entry for New Jersey netted them 0.27 Fenwick events. For Philadelphia, it’s 0.51.

    If you look at things by type, you can see that with both controlled and uncontrolled entries, Philadelphia generated many, many more shot attempts. That gap, although not as wide, appears in their overall zone entry data too. For whatever reason, it seems to be bigger here – might be noise, might be something. Either Philadelphia has players who are much better at turning zone entries into shot attempts (something that I’m skeptical of, given that Eric Tulsky has found that plugs and stars tend to generate similar volumes of shot attempts per entry), it’s randomness or New Jersey did something for some reason that suppressed the rate at which they turned zone entries into shot attempts.

    Looking at the defensive end of things might shed some light on this, so lets do that.

    We know that limiting controlled entries is better from a defensive perspective because it limits the likelihood of the opposition getting a shot attempt. Sure enough, New Jersey was much better at this than the Flyers. 36% of the zone entries against New Jersey were controlled zone entries. For Philadelphia, that number was 45.5%. Limiting the number of controlled entries will drive down the shot attempt volume.

    The samples are so small on the shot attempts allowed by entry type column that I’m not sure that there’s much we can take from them but still: wow. The Devils were kind of bad at limiting shot attempts off controlled entries and phenomenal at limiting shots off pucks that were dumped in. I note that, overall, New Jersey allows fewer shot attempts per controlled or uncontrolled entry than the Flyers do.

    * * *

    Welcome back, for those of you who sat out the reasoning part. Let’s now take all of this reasoning and discussion and put it into a conclusion. The question we are attempting to address is this: Devil games saw just 0.327 shot attempts/DZ win for both teams off entries within 21 seconds of a Devils DZ win. Flyers games saw 0.490 shot attempts/DZ win for both teams off entries within 21 seconds of a Flyers DZ win. To what extent can we attribute this difference to zone entry frequency, type and shot rate?

    So this is really three questions. First, does the frequency of zone entries explain why the Devils allow so many fewer shot attempts off of DZ faceoff wins? Well, there were actually more zone entry attempts in the 21 seconds following a Devils’ DZ win than there were in the 21 seconds following a Flyers’ DZ win. If Devils’ DZ faceoff wins resulted in the same rate of zone entry attempt as Flyers’ DZ faceoff wins, the volume of shot attempts following a Devils’ DZ faceoff win would fall slightly. So it’s not that.

    What about zone entry types? There are two major types we’re concerned with: controlled and uncontrolled entries. From an offensive perspective, there’s basically no difference. The Devils and Flyers generate controlled entries at the same rate. Defensively, there was a large difference. Only 36% of zone entries attempted against the Devils resulted in a successful carry into their end. For the Flyers, that figure was 45%. Those are the bad kinds to allow. The low amount allowed by the Devils explains some of the lower rate of shot attempts observed in their games.

    Finally, there’s the rate at which shot attempts occurred from a given type of entry. This is where the real differences can be found. I’ve put together a table that summarizes the Devils and Flyers rates here and the impact that that has on the volume of shot attempts in Devils’ games.

    So, for example, the Flyers allowed 0.11 shot attempts per DZ faceoff win before clearing the puck. For New Jersey, that was 0.07, a difference of 0.04. If you take into account the number of DZ faceoff wins for the Devils, it works out to 6.6 fewer shot attempts than we’d have seen if they had the Flyers’ rates. The “Impact” column lets us see what’s most important here.

    You can see that there are four big drivers for the low shot rate that we observe following Devils’ DZ faceoff wins in addition to the Devils forcing uncontrolled zone entries: the Devils generate few shot attempts off controlled entries, they generate few shot attempts off uncontrolled entries, they allow fewer shot attempts off DZ faceoff wins and they posted a silly number in terms of shot attempts allowed off uncontrolled entries.

    A reminder is probably in order here: not all of these are bad things. The question I’m asking is why neither team gets very many shot attempts following a Devils’ defensive zone faceoff win. Some of that is good: you don’t want the other team getting shot attempts. There may well be tradeoffs in what the Devils are doing. For example, it’s pretty easy to imagine that the Devils aren’t attacking aggressively when they escape their zone, which hurts them in terms of generating shots but pays dividends in terms of forcing the other team to dump pucks in and makes it less likely that the opposition recovers them. Hockey’s not baseball (or so I’m occasionally told); offence and defence are inextricably tied together.

    With a dataset for the entire NHL, someone with an interest in getting as much as humanly possible out of an NHL team basically has an index to the entire league. You can identify where teams deviate from norms and then go to the video and ask yourself what they’re doing. It enables us to boil big questions (“Why are the Devils a Corsi powerhouse?”) down to their essence (“The Devils generate their Corsi advantage from A, B and C. What do they do differently?”).

    Corey’s work lets us get closer to where this is ultimately going: identifying things teams and players should do differently. If you’ve got a few bucks to toss into the pot to support him in doing this, please consider doing so.

    Email Tyler Dellow at tyler@mc79hockey.com


    11 Responses to NJ and PHI: Explaining the Devils’ Post-DZ Faceoff Win Results

    1. AJ
      May 5, 2014 at

      This is good stuff. Seems like the Devils play a strong neutral zone game which has kind of been a hallmark of their success for years. Also, the fact that they are able to recover dump ins and clear the d-zone is very interesting. They have a few smooth skaters back there but Salvador and Volchenkov don’t exactly seem like the type of players to make a quick outlet. I’d expect to see a lot of guys who make quick decisions and move the puck. Any case this is very thought-provoking, thanks.

    2. Woodguy
      May 5, 2014 at

      I love where all of this is going.

      Props to you and Corey, to whom a contribution is on its way.

      2 things really stand out to me.

      NJD with .47 F/E with a controlled entry. That seems pretty low on the surface (without seeing too many more besides PHI)

      I wonder if they are coached to cycle the puck into a corner rather than take a low percentage shot?

      We’ve hear the apocryphal stories of Hitchcock and Lemaire teams being coached to dump it into corners on 2-1 if they are late in the shift.

      This isn’t that bad, but it stinks like Lemaire.

      The second thing that jumps out is NJD Opp getting a minuscule .07 F/E on uncontrolled entries.

      We all know the stories about Brodueur and his ability to play the puck, that cannot account for this given his gp and how small that number is.

      Do they get away with mugging the first forechecker every time?

      That’s really interesting and look forward to hearing more about it from you.

      Thanks again.

      • Triumph
        May 6, 2014 at

        No, they don’t get away with mugging the first forechecker every time. The Devils have been at or near the bottom of the league in power plays against for ages upon ages.

        Schneider did say that there is a much larger reliance on goalie puck playing in NJ’s system and indeed he was out of his net playing the puck a whole lot even though he’s not very good at it – it’s something I felt like The Contrarian Goaltender wasn’t really able to quantify with his study on Brodeur and puck handling. Maybe opposing goalies don’t play the puck enough? Or maybe they play it enough and the effect of playing the puck so much in depressing shots on goal against isn’t mitigated by the number of goals you give up on misplaying the puck?

      • May 6, 2014 at

        Just by sight, I can speak to that the Devils have tended to play the puck in deep to get started on offense this season. It’s one thing to see it from a fourth line; but it’s another to see it from the top unit. What made Jagr so amazing for me to watch this season wasn’t just that he was playing a lot and putting up points, but that he was starting and winning battles down low as if he was a big, healthy 20-year old. I can believe that they’re being coached to get it in deep and move it around before finding a seam. Also because if they lose possession too quickly, two or three forwards are easily caught behind on a counter-attack. This is just observation but I wonder how someone tracking entries would see it. Maybe they’d see a pattern?

        Incidentally, the Devils don’t really mug their forecheckers. Or if they do, they’re especially good at not being penalized for it for several years.

    3. Triumph
      May 6, 2014 at

      This is great work, but it seems like we’re running into sample size issues, no? The error bars on ‘true talent’ for all those things should be pretty high, I imagine. Anyway, as a leading purveyor of grains of salt and not much else around the stats community these days, just felt I should point that out.

      • Tyler Dellow
        May 6, 2014 at

        Yep, I don’t disagree. At the same time, given that we’ve observed the NJ low shot attempt phenomenon over an entire season (and, IIRC, it was around last year too), I’m less concerned than I’d otherwise be. We’re just drlling down on a small portion of things but we know that the effect into which we’re drilling is bigger.

    4. DD
      May 6, 2014 at

      This seems to be a Devils-specific thing rather than a coaching strategy as Peter DeBoer’s Florida teams didn’t have low shot totals, while the Lemaire-coached Devils were near the bottom of the league as well.

      • DD
        May 6, 2014 at

        Just to clarifly, it clearly is a coaching strategy, but not some specific philosophy that DeBoer holds on to since he doesn’t seem to have implemented a similar system in Florida.

    5. Johnny ringo
      May 6, 2014 at

      Just wanted to say this is an amazing article and I will be checking in with you guys frequently now. Thanks!

    6. Pingback: New Jersey Devils hire one-time Edmonton Oilers blogger as new Director of …

    7. Pingback: How Can the Rangers Replace Anton Stralman? | Save By Richter

    Leave a Reply

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