There’s been no shortage of great non-hockey content here at DBD. But I’m not wired that way. Yea I love breweries. And yea, I love video games. I’ll still rock out to that killer Castlevania IV soundtrack. But I love hockey just a little bit more. So I’m gonna pretend like nothing happened. Nope, nothing major at all. Let’s pick up where we left off, shall we?
Dallas has been defined by the kind of losing streak that can be turned into a punchline — and like any (good) punchline, the truth is penetrating. Teams ‘backing into’ the playoffs historically don’t do well. Two years ago when Dallas collapsed in March under Ken Hitchcock, that team was in a frighteningly similar position, with both the 2018 and 2020 team stuck at 82 points after game 69. The warning signs were there, in 2018.
The warning signs are here, this season.
But it’s not all bad. The Stars have been doing something really well under Rick Bowness. Something kind of important. They are 10th in the NHL in high danger shot attempts-for (classified as a shot taken around the crease and slot area of the ice).
Sounds good, right? Like something you can build on? Here’s my spicy take: I’m not so sure. Obviously, in an absolute vacuum, shooting from closer to the net is preferable. After all, the closer a player is to the net, the more likely they are to score a goal. Two plus two equals four, right?
But a high danger shot on its own is just that: a shot generated around the crease/slot area of the ice. ‘Danger’ is a relative term. If you had to choose between Blake Comeau firing a wrister from the low slot versus Denis Gurianov firing a slapshot from just inside the faceoff hashmarks, which would you choose? The high danger shot (technically Comeau’s), or the scoring chance (officially Gurianov’s)?
Here’s an interesting contradiction: while Dallas is 10th in the NHL in high danger shot attempts generated, they are 17th in expected goals generated (or xGF per 60). Why aren’t the Stars expected to score more if they’re constantly in the right area to score? The explanation is more obvious the more you think about it: while high danger shots are more likely to go in, it doesn’t mean non-high danger shots won’t, and Bowness’ Stars don’t generate much of anything else.
The Great Quantity vs. Quality Debate (...is not what you think it is)
Whether deliberate or accidental, the focus under Dallas’ organization has seemed to be on quality at the cost of quantity. If that’s Bowness’ strategy, he wouldn’t be the first NHL coach to argue its merits. As I said, they’re good at getting close to the net. But that’s the only thing they do.
- they’re very poor at generating low percentage shots, yet are much more likely to score on them in proportion to what they generate.
- they’re moderately poor at generating medium danger shots, yet are less likely to score on them in proportion to what they generate.
- they’re very good at generating higher danger shots, yet are significantly less likely to score with them in proportion to what they generate.
Before we go any further, let’s look at another unremarkable graph. If we look at the types of shots, will that give us insight?
They’re way above league average when it comes to deflections and wraparounds. They’re getting to the dirty areas. It’s gotta be bad luck. Right? Not so fast. In terms of general insights over the years, there’s something worth noting about high danger shots:
- generating high danger shots is a skill teams can repeat, but it’s not as repeatable as other shot attempts.
- generating high danger shots is a skill that teams can repeat, but the link between turning those shots into goals is a modest one.
- wraparounds and deflections are the least consistent of shot types, and tell us little about whether or not they can contribute to future goal-scoring.
This is a really important point. Not only is Dallas focusing their strategy on attacking an area that is harder to attack consistently, but they’re attacking with the type of tactics that are hardest to replicate. The former point has actually played out in realtime. From October to January, Dallas was actually better at generating high danger chances, when they ranked 4th in the NHL (again, currently 10th). The latter point is one I’ll get back to.
Tales From the Tape
Let me digress for a second. My eyes don’t see everything. Not all at once. Sometimes I have to use them again. And again. Not just because I have bad vision without corrective lenses, but because I’m not always 100% focused. Analytics remind me when to use them again; sharpening my vision not just from a visual standpoint, but from a value standpoint.
I bring this up to prove a point about the Dallas Stars organization. What we don’t know is a strong argument against the lazy perspective that ‘analytics provide value when you need it, and noise when you don’t.’ Who decides what’s valuable to begin with? Stats aren’t meant to be static figures of what happened the night before. They’re meant to be dynamic representations of interacting scenarios. As scenarios evolve, so do the models trying to capture the different patterns that teams exhibit from game to game. And as the sample size grows, we get a clearer picture of what events are contributing to those patterns. That to me, is the real value: being able to capture patterns — lack of shot quantity — before they emerge to become results — lack of goals, which ultimately lead to lack of wins.
I digress about analytics not just because I’m gonna use them, but because Dallas doesn’t seem to forecast well as an organization. I worry we’re looking at yet another trend they’re not gonna see until it’s staring them in the golfing face. So is Dallas getting the most value out of these high danger shots? I don’t believe so. Here’s what I consider to be a typical Dallas high danger shot. A forward takes possession of the puck down low, shifts it up high, and then down low again.
Below you’ll see Stephen Johns receive the puck from down low, then unload a bomb from up high. Radek Faksa and Blake Comeau try to grab the rebound down low to register a high danger shot. Just the meat and the potatoes.
There’s nothing wrong with this play. It’s the FCC line. The puck goes at the net, they go after the puck, and then pray to the hockey gods that a kerfuffle can slide the puck past the goaltender.
But you can see where the danger might be considered relative. The “high danger” shot is preceded by a low danger shot in mechanical fashion: Edmonton goaltender Mikko Koskinen has the whole play in front of him. There’s only one location that Dallas is threatening from. Not only does Koskinen see the angle, but so does Edmonton’s defense.
There’s something in analytics called Flurry Adjusted Expected Goals. A whaaaat? Consider the following scenario. The likelihood of Stephen Johns scoring from a point shot is statistically slim. But the likelihood of Radek Faksa scoring a goal by picking up a rebound close to the net as a direct result of Johns’ slapshot? A lot better. However, Faksa’s high danger chance only existed because of Johns’ low danger chance. That’s where the flurry adjustment comes in: to help clean up some of the noise. As expected, once you account for these shot flurries, Dallas’ expected goal rates are lower by a not-insignificant margin.
The play below follows the same pattern. Jamie Oleksiak receives the puck from down low, chambers high, and Roope Hintz grabs a tip down low to register a high danger shot.
Same thing on the next play. Miro Heiskanen receives the puck from down low, shoots from up high, and Faksa grabs the deflection down low to register a high danger shot.
This was a game Dallas absolutely dominated, but still lost. It was a sloppy game. A lot stood out to me: not just the bottom six minutes Hintz and Gurianov got (keepers of the faith may be sick of this criticism but since we’re talking high danger shots, it’s worth noting that the players Bowness trusts are the ones leaving HD chances on the table), but also the way Dallas dominated. Did you ever feel like Koskinen, kind of just had Dallas’ game down to a science? That’s something people working on the best model for predicting goals have been trying to figure out. In a game that’s constantly evolving, at what point do efficiencies become...inefficiencies?
Rebounds themselves are a fascinating case. John McCool wrote a really great abstract on the biases of how expected goals are calculated, and whether using a model that takes inputs from the previous season to tell us about the current season creates statistical noise. Rebounds have had dramatically less influence over expected goal probability, experiencing a steep decline specifically during the 2013-2014 season, over the last decade: something Dallas has relied on a lot in the post-Ruff era.
Whether it’s goalies themselves adjusting or the teams, the fact remains. Let’s look at two clips. The first:
Then the second:
This seems like the ideal Stars forecheck to me. Two low, one rover mid-zone to break up exits. Benn does that.— however comma (@eyeversehen) March 11, 2020
But then you’re left with 3 high, so you can shoot through bodies or dump it in, where you’re now out-numbered. They split the difference with “miracle tip-in.” pic.twitter.com/NwtjDwV4oS
This, to me, is a classic shot attempt under Bowness; set up in an umbrella-like formation, throw the puck behind the net or across the slot hoping to grab a quick point shot, or generate some random deflection. It doesn’t always work. In fact, it obviously doesn’t work at all or else they wouldn’t be outscored by bottom feeders like New Jersey and Ottawa. But it’s something Dallas seems wedded too.*
Dallas’ lack of creativity and current struggles make intuitive sense when you break these plays down. They’re inside the right area, but they’re outside the right threat. The plays themselves seem easy enough to defend. The goaltender doesn’t have to move dramatically from one side to the next. The defenders have to fill out only a small list of to-do’s: get in lanes to block the shot, pressure the shooter, or grab the rebound before Dallas does. Where’s the passing on Dallas’ end? Where’s the ability to threaten from two locations, instead of just one? Where are the isolated shooters? Bowness himself can’t decide if wants more passing, or less passing.
While insights are limited when it comes to passing data, some brave souls have tried to figure out how well a pass can influence the quality of a scoring chance. And yes, here again, Dallas is behind the curve.
Of all the passes to increase the quality of a shot, passes from low to high are the worst.
Getting Ahead of the Curve
Before my hot take gets twisted, let’s clarify something. The argument is not, high danger shots are actually bad. The argument is, the danger is relative given their predictability.
‘Getting to the dirty areas’ is not a bad philosophy, but it doesn’t account for the drastic difference between players. Standing in front of the crease trying to slam home a rebound might be how you want Faksa, Cogliano, and Comeau to score. But is that where you want Jamie Benn and Tyler Seguin to be, all the time? Isn’t their skill precisely in having time and space? Being able to threaten from multiple areas instead of just one?
Over the last three seasons, Seguin and Benn have been on the ice for 45 medium danger goals, and 28 low danger ones; all at even strength. Quantity has a quality all its own. By ignoring ways to exploit other areas — well, you get the point.
I suspect this is what makes Hintz and Gurianov stand out beyond their goal-scoring. Not only are they producing, but they don’t seem trapped by the system. They have enough speed to create their own space, and can therefore play beyond the dump and chase trappings. This is also why their individual xG rates are so much better than the others: most goals are scored within seven seconds of entering the zone, which is what Hintz and Gurianov specialize in; being able to break into the zone, quickly, for one-on-one opportunities.
I also suspect this has a trickle-down effect on the defensemen like Heiskanen and John Klingberg. Both have struggled to make the impact we know they’re capable of. Heiskanen’s numbers off the ice versus on are particularly stark under Bowness. Just as the system asks Benn and Seguin to play more like Faksa and Comeau, the system asks Heiskanen and Klingberg to play more like Lindell and Polak, which is why one gets rewarded with the other’s shifts despite no tangible justification other than Protect Our Identity (I hate to group Lindell and Polak together, but they share something in common: they are black holes in transition from zone to zone, which affects Dallas’ ability to generate offense at even strength, and their lack of creativity breaking out of the zone leads to what can only be charitably described as ’informal’ turnovers).
I want to go back to Ryan Stimson’s article on shot types. He also mentions an important point about one-timers; setting them up is a repeatable skill that is also more likely to lead to future scoring. Recall the chart on shot types: Dallas is actively bad at generating slapshots, so it naturally follows that one-timers (not synonymous, but closely linked) aren’t on the menu as much as they should. This is worth stressing because we can look at Dallas’ unlucky shooting percentage on their high danger shots and say ‘well it’ll get better’ and I think that’s true. But what if their low danger shooting percentage starts to fall where it should, in proportion to what they attempt? What if their medium danger shot attempts don’t improve? What if their high danger shooting percentage starts to go up, but their high danger attempts, start (read: continue) to fall? No matter what, they’re still in no man’s land.
So sure, it’s nice to generate high danger shots at a higher rate than most teams. But why stop there? It’s hard to buy luck when you’re only scratching off one ticket. If chance favors the prepared, then it follows that chance punishes the unrehearsed. Dallas better start rehearsing if they want to come back a stronger team, whether it’s in the summer, or next season.
*I have a crackpot theory on this one. In a game where jobs were literally on the line, Dallas scored 6 unanswered goals against Minnesota back on October 28th. Two of the goals in the 3rd period went off Wild players from Stars looking for random deflections near the slot. As you can see, they still do this from time to time. Nobody’s open, but a defending player is near the goal crease, there’s a teammate who maybe-sort-of has lane towards a rebound, but without certainty or probability on their side, Dallas throws it at the net anyway, ignoring other options. I like this theory, but of course, like with most crackpot theories, the boring version seems to explain more, which is: not being a good possession team means less confidence and efficiency when they do have possession.
All data taken from Natural Stat Trick.