“What the hell is John Hynes doing?”
This is the general sentiment I’ve seen on twitter with regards to who from this supercharged Devils roster have been seeing the most ice time in the most important situations. And it’s not that difficult to see why. For instance, there was the much-maligned decision to change the lineup while up 3 goals in the season opener (more on that later). And even just over the past two games, this is the 5v5 time on ice for the 12 Devils’ forwards:
Hayden/Coleman/Rooney at the top, and Hughes/Bratt/Gusev at the bottom — just how we drew it up in the beginning of the season, right? I mean, Hayden didn’t make most people’s starting night lineup, and most people didn’t even think Rooney should make the team. Meanwhile, the latter trio was the near unanimous 2nd line in the offseason. And it’s not just those ornery Devils folks noticing — in his coverage of odd usage across the NHL (paywall), Dom Luszczyszyn was confounded by deployment as well. So what gives? This usage of the unglamorous veterans over skill players is the first lineup decision that I’ll be looking at. We’ll focus specifically at what separates those top 3 from Jesper Bratt (focusing on Bratt because Gusev and Hughes have played very little in the NHL so we have less evidence for them).
The “Rooney/Hayden” Era
Well, some of this is due to special teams — of the 120 minutes in the last two games, only 74 have been at 5v5 — which will cause some weird deployment. But Kevin Rooney and Blake Coleman are penalty killers while Hughes, Gusev, and Bratt all play on the powerplay, so that actually doesn’t get us much closer to an answer. Why is it that, with boo birds raining down on him, with the pressure to produce mounting, Hynes has chosen to ice guys that may not clear waivers, over some clear pillars of the franchise future?
In a word? Defense. Below are the player isolates from Micah McCurdy’s hockeyviz (Ex: Coleman). If you like what they have to show, please consider donating to his Patreon. These maps show how the player impacts the ice, holding score, venue, teammates, opponents, coaches, etc. all constant. They are “chained” so that each new season is informed with a prior established by their past production. The percentage you see is the threat percentage — think of it as the percent increase/decrease on the expected goals.
As you can see, despite the fact that only Coleman (btw, how good is Blake Coleman!?) had an offensive impact greater than that of Jesper Bratt, all of these guys reduced the threat of their opponents by over 9%, whereas Bratt increased it by 3%. In other words, with Bratt on the ice, the Devils were more likely to score, but so were their opponents. With Hayden, both the Devils and the opponents are less likely to score.
So how does this impact decision-making? Coming off the Florida game, the Devils were fighting a reputation that they could not hold a league. Now, some of that was the goaltending — as I talked about in last week in my discussion on who to blame. But some was on the Devils skaters as well. The expectation here is that these guys can slow the game down and maintain the lead. Lo and behold, when leading, the Devils are allowing the least scoring chances per 60 when Kevin Rooney is on the ice. And these impacts appear to be independent of model — Evolving-Hockey’s RAPMs also agree that Coleman/Hayden/Rooney have better defensive (and, in some cases, overall) impact than Bratt.
Notice those last two bars? It’s all positive for Rooney/Hayden/Coleman, and all negative for Bratt. The overall impact does not seem too dissimilar, but Bratt’s impact is offensive whereas the others are defensive. That makes the deployment in a game where the Devils are ahead a little less befuddling.
Full disclosure, it still seems like the weighting towards Hayden/Rooney is a little outta whack. Special teams have conspired to lower some of the 5v5 ice times, but even if we look at total ice time the last two games, Zajac has played more than Hall and Rooney has played more than Palmieri. Yes, there’s something to be said for protecting a lead. But you know what’s better than protecting a lead? Extending it. It’s one thing to keep the unproven Nikita Gusev off the ice, but Kyle Palmieri has an amazing shot and an underrated defensive game and should not be out-minuted by Kevin Rooney. Taylor Hall is an Hart Trophy winner, and shoud vastly outplay Travis Zajac. I understand the logic behind the deployment, I don’t necessarily agree with it.
But these are fundamental philosophies, at least for the time being. Hynes and the Devils have come under fire for their treatment of the lines intra-game. Sepecifically when to change them, and when not to change them. I’l be focusing on the one big lineup change that got the most criticism.
Changing the Lines Against Winnipeg
This one was met with pretty fiery criticism. Some of it was more veiled and contextual, others were about as blatant as it can be. This criticism essentially boils down to one big assumption — everything was going pretty well through two periods, and the wheels came off in the 3rd. What I’d like to show here is that both of those statements are not true — things were not all hunky-dory after 2 period, and they did not get measurably worse in the third. I originally voiced these clarifications in this two-tweet thread. The lineup decision was made at the end of the 2nd period. It was not a specific decision that swapped linemates — I’ve seen some say it was — as much as it was an abandonment of the original lines. I think it’s helpful to look at it through the lens of Jack Hughes.
Hughes was with Bratt and Gusev in the 2nd and third, and was with Hayden-Gusev, then Gusev-Zajac, then Hall-Hischier, then didn’t play the entire final 8 minutes of the game. So, first of all, was this warranted? How was the Hughes line doing? And second, did the move benefit the team as a whole? Well, according to the NST summary of the game, that was the only one of the 4 most common lines to be a negative shot share unit, controlling only 36% of the shots. The weren’t getting completely buried, so it might not have been worth screwing around with the rest of the lines. So ... did it work?
As you can see from the graph above, the 5v5 shot-share was plummeting starting from a little before halfway into the 2nd period. The lineup change seemed to stop the bleeding somewhat. In fact, the Devils considerably in every on-ice metric except goals.
Yes, the Devils were up 4-1 after two. Yes, Hynes changed the lineup shortly thereafter. And yes, the Devils ended up losing. But to say only that and ignore the actual pace of play in the game is to have an incomplete picture of the game. Every metric other than goals improved for the Devils after the change, but the young goalie coming off the bench in the first game of the season let two through, and Brossoit didn’t.
Of course, it wasn’t all on Blackwood. The team struggled on zone exits pretty egregiously (Corey Sznajder is building a Tableau this year on which you can see evidence of this). This deficiency largely falls on the defenders, who have had some interesting deployment of their own.
I’ll start off with the one that we, here at AAtJ criticized the most — good ‘ol 25-28 — Mirco Mueller and Damon Severson. Why? Well ...
In only 23 minutes, that pairing has managed to get outscored 0-5. But, did they deserve it? Yeah, totally — they were out-chanced 4-17 and have pretty clearly been awful. Which is, by the way, why after only 23 minutes, they are no longer a pairing?
The question should be, was it fair to think they’d work originally? Of the 31 pairings the Devils have tried since 2017, Severson-Mueller was 6th in xGF%Rel. The only pairing with a higher xGF impact and more TOI was Butcher-Lovejoy. That’s including the terrible start to this year. More subjectively, Severson’s been worst when playing with guys that were poor in their own zone like John Moore, and best when with defensively responsible guys like Andy Greene. I think Mueller was a fair bet to make, but it really didn’t work. I’m not ruling it out forever, but going away from it was the right move.
What about the other decisions here? It’s tough to isolate some of it, because Greene was injured, and now Butcher, but the first top pairing was Subban-Greene which worked very well. After Greene was injured, Severson got the nod on the top pair which had a negative goal result, but came out on top everywhere else. Vatanen-Butcher had some pretty disappointing results given Butcher’s long history of positive on-ice impact so that may be fair to question as the top-minute pairing, but most of the rest of what we’ve seen seems pretty defensible.
Other than the fact that Vatanen is a strange pairing choice, I’ve been pleasantly surprised with the usage rate of Will Butcher. He’s injured right now, but we don’t currently have reporting indicating anything long term so assuming he returns soon, this will be something to watch. Since joining the team, he leads all Devils skaters in on-ice shot impact, and trails only Taylor Hall in expected goal impact. The two defining features of skater impact in recent years have been that Taylor Hall dominates impact among forwards, and Will Butcher dominates impact among defenders. It’s not immediately clear to me if the coaching staff has finally noticed what they have in him, but he leads all Devils defenders with 17.4 minutes per game at 5v5. This is probably a little bit due to the fact that he’s not playing big minutes on special teams, but 1) his overall minutes per game are up a minute per game from last year and 4 minutes from his rookie as well, and 2) he’s most effective at 5v5, so that’s where we want to see the biggest jump anyway.
Something negative to watch as the year progresses is the play of Damon Severson. While his expected goal and possession numbers aren’t terrible, we’re being outscored 4-11 with him on the ice. Normally you’d defer to the less erratic CF or xG metrics, but Severson has allowed more goals than expected (GA > xGA) in every single season of his career — so it’s reasonable to surmise he’s doing something other than what’s calculated in xG to allow the opponent to score on a disproportionate amount of their shots. The fact of the matter is, the Devils are getting outscored 16-20 at 5v5 in total, which means that they are actually outscoring opponents 12-9 when Severson is not on the ice. Given how this season has gone so far, that is a fairly shocking statement. His deployment may require continued experimentation over the course of the year. I’m not holding out hope, though, given that this is the same staff that tried Severson-Moore for 1600 minutes.
No decision is made in a vacuum. The Devils have had a pretty poor start to the season, suffering in pretty much all situations at some point. There are a lot of new faces on the team that have not played much together. There are some players completely new to the league that clearly have a way to go before having a sound 200-ft game (Hughes, Gusev). There have already been 3 injuries to key players (Hischier, Greene, Butcher).
When accounting for all of these things, sometimes you need the offensive of Jesper Bratt and Jack Hughes, sometimes you need the defense of Kevin Rooney and John Hayden. Sometimes you need to make a move in anticipation of future results, not in reaction to prior ones. And sometimes, you need to experiment with different pairings fairly consistently.
What do you guys think of these decisions? Do any of them seem indefensible? Is there anything that you’re a big fan of? What would you like to see moving forward? Leave your thoughts on these questions and anything else you may wish to share in the comments section below. Thanks as always for reading!