So, I guess Ovi was right. The Capitals are not suck this year. With 59 points through 45 games at the start of their bye week, Washington sits in first place in the Metro and second in the East.
That accomplishment merits celebration and consideration. Many people, including me, expected the Caps to fall off this season (while still being a Good Team™). From an outcomes perspective (goals and wins), that simply has not happened. From a process perspective (smaller events like shot attempts), the Caps have dropped even farther than I feared.
I’m fascinated by this gap between outcomes and underlying process. I wonder what it portends for how we evaluate the team right now and how it should inform our expectations in the back half of the season. Because there are really only a few ways this will go down:
Or, put another way: what if I’m wrong about everything?
The best way to begin, I think, is a massive, convoluted, color-coded table that tries to cram as much information as possible about every team in the league into a single image.
Click to enlarge. There’s a glossary at the bottom of this article.
I pulled down team-level data from Natural Stat Trick for play at even strength, power play, and the penalty kill. I color coded each column (green is good, red is bad) and sorted by standings points. The effect is largely that good teams are at top (green) and bad teams are at bottom (red). The interesting stuff occurs when you see exceptions to what you’d expect. For example:
Those even-strength percentages have been the focus of concern this season. Since we’ve got a nice long break ahead of us, I looked up previous teams that are comparable to the 2018 Caps in those stats to see how those similar teams fared after mid-January.
WSH: 48.4 percent (22nd) of attempts belong to the Caps
The 2015 Flyers had this same shot-attempt percentage at this point in the season. At that time, they were on a miserable 76-point pace. They improved down the stretch but still missed the playoffs with 84 points.
The 2016 Wild also had a 48.4-percent shot share, but they performed closer to the Caps. They were on a 99-point pace after 45 games, but slowed in the back half of the season to earn 87 standings points, enough for a wild card playoff appearance.
WSH: 48.0 percent (22nd) of chances belong to the Caps
Here’s where it gets less encouraging. Two teams in 2012 had the same scoring-chance percentage as the 2018 Capitals: the Leafs and the Canadiens.
The Leafs were on a 91-point pace, but slowed to finish with 80, missing the playoffs by a mile – but still closer than the Canadiens, who were on a 75-point pace before finishing dead last in the Northeast with 78 standings points.
WSH: 43.2 percent (last place) of chances belong to the Caps
No team in the NHL has a smaller share of high-danger events than the Capitals this season, but here’s some company in previous years.
The 2015 Flames (43.1 percent of high-danger chances) were on a 90-point pace at part of the season, and they ended up making the playoffs with a late-season rally to earn 97 points. That same year’s Coyotes (43.9 percent) were less lucky. They were on a 70-point pace, but got just 56 after 82 games, winning dead last place in the west.
With 59 standings points already banked (a 108-point pace), the Capitals are in better shape than every one of their historical comparables. It’s basically unprecedented for a team this poor at 5-on-5 possession to be this successful in goals and on the standings board – especially considering how ordinary their special teams are.
Exceptions are fascinating. We get insight from looking what doesn’t match up to the pattern. For Washington, goals don’t match up with shot volume and shot quality, so we have to ask why. Volatile luck is the most convenient answer, and traditionally the most likely, but that doesn’t mean we should accept it automatically.
Shot volume and shot quality are the cornerstones of one evolving corner of hockey analysis called expected goals, where one predicts what should happen based on the profile of a team’s offense and defense, then compare it to what actually happened.
The Capitals are doing way better than models expect.
The “daggers” show the difference between a team’s expected goals for and against. Wsh surprises most as they look fine in the standings but atrocious here. pic.twitter.com/bBR0W0bbgJ
— Sean Tierney (@ChartingHockey) January 13, 2018
That could mean that the Caps are getting undue and unsustainable puck luck (by way of precariously high shooting and/or save percentages), or it could mean that the expected goals model doesn’t capture what makes them a good team.
One argument for the latter, and one recently expressed by Muneeb Alam of Japers Rink, is that the Capitals’ history of over-performing their expected goals percentage is steady enough that it might just be sustainable.
On the defensive side, one reason for the over-performance is obvious: Braden Holtby. Holtby has saved more shots than expected every season under Trotz, including a super-hot season in 2016-17, in which his 93.7 save percentage was way over the expected 92.5 save percentage based on the quality and volume of shots he faced. He’s demonstrably an above average goalie, and he saves more shots than one would expect.
On the offensive side, it’s possible that the model’s shortcomings (i.e. currently available data do not describe passes prior to a shot) “punish” the Caps for their offensive creativity. Put another way: maybe the overpassing fans and pundits hate so much is lowering shot volume and increasing team’s dangerousness in a hard-to-quantify way. Except, for this to be true, the added danger would have to be solely a function of puck movement before the shot instead of shot location. Otherwise, we would see the shot quality reflected in the team’s scoring chances, high-danger chances, and expected goals –all stats where the team is struggling (they’re among the league’s bottom third in all three).
(The Washington Post’s Isabelle Khurshudyan ran a piece about this very idea last month, and I offered a rebuttal of its quotes the same day here. If my rebuttal is wrong, this is how.)
Shot quality is a hot topic in the NHL right now. The gap between high-possession teams and low-possession teams (as measured in pure shot-attempt volume) has sharply dropped over the last three seasons, and teams are looking for an edge wherever they can get it.
If the Caps have found a way to be continually successful in outcomes (goals) without good shot volume, that would be rare. If they have found a way to be continually successful without good shot volume and without shot quality (as currently quantified using available data), that would probably be unprecedented.
This doesn’t have to be an article of faith. In my response to the WaPo article about shot quality, I noted that the Capitals generate a lot of medium-quality shots (top ten in the league by volume) and score on them better than anyone. It’s possible, and maybe even probable, that elite Caps scorers like Kuznetsov, Backstrom, Ovechkin, and Oshie could transform a shot that is currently considered medium danger into one we’d anecdotally call high danger. They could do this, for example, by making passes across royal road, or by firing the shot at an uncommonly inconvenient angle for the goalie.
An optimist would have to hope that’s it. The comparable teams I mentioned above mostly got worse by the end of the season. Most of them missed the playoffs. None of them went deep in the playoffs. Even if the Caps were a virtual lock for the playoffs (they are), their priors suggest they’re not going to be a championship contender, regardless of current standings. All that may be true for the Capitals unless we lack the vocabulary to describe (with numbers) the team’s offensive magic.
For us to believe that the Capitals will continue to be successful requires us to believe that the team has found a unique and reliable way of generating shot quality that no statistical model can account for yet (plus having Braden Holtby in net). That seems unlikely, but the limits of current models should be considered. We know where a shot comes from (location), what kind of shot it is (slap, wrist, etc.), and its general context (rebound, on the rush, etc.), but we don’t know if that shot is the result of a great cross-ice pass that forces the goalie to make a lateral shift. Until tracking data improves and is made widely available, there will remain some epistemological problems with expected goals. (To that point, Prashanth Iyer of Hockey Graphs has written a neat tool to crowdsource the collection of this data.)
The final 37 games of the Caps season can be a test. If the Caps’ underlying play continues and they keep near their 108-point pace, then perhaps that means they’re a sui generis team – not merely good in a way that an ever-evolving body of research doesn’t yet describe, but good in a way that is contrary to patterns of successful NHL teams over the last decade. If the Caps begin to founder, and their goals-for percentage trends towards their shot-attempt numbers, then maybe they’re not so revolutionary after all, and our understanding of process would be partially validated.
In real life, the result won’t be nearly that tidy. Hockey is probabilistic, and, really, the Caps’ success so far is most likely some unknown combination of both good luck and an esoteric kind of eldritch shot quality magic that I hope we’ll be able to describe better in the future once new data are available.
Russian Machine Never Breaks is not associated with the Washington Capitals; Monumental Sports, the NHL, or its properties. Not even a little bit.
All original content on russianmachineneverbreaks.com is licensed under Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0)– unless otherwise stated or superseded by another license. You are free to share, copy, and remix this content so long as it is attributed, done for noncommercial purposes, and done so under a license similar to this one.