1. Underlying or Under, Lying?
Check in with your favorite statistical modelers and you might wonder why the NHL is even bothering to play the Caps/’Canes second round series:
And it’s not hard to see why Carolina is such a heavy favorite in those models. In terms of expected goals over the course of the regular season, they were the better team (both in the theoretical matchup of the Caps offense against the ‘Canes defense (top) and vice versa (bottom):
The Caps are a very good team at five-on-five; the ‘Canes are the best team at five-on-five (at least by their underlying metrics):
Carolina also held an edge on special teams, with both teams in the same neck of the woods on the power-play, but Carolina posting better underlying numbers on the penalty kill:
Those heat maps and much of the bar graph are based on expected goals, which are black-box formulae based on underlying shot metrics. Actual goals, on the other hand, are based on pucks that cross goal lines and wind up on giant scoreboards that hang precariously above hockey rinks. And they tell a slightly different story:
What we see is that over the course of the regular season, the Caps outperformed their expected goals both for and against at five-on-five and the ‘Canes underperformed theirs. This suggests that Washington has superior finishing and goaltending talent (something we’ve touched on recently), but there’s also plenty of room for some regression, particularly on the Caps’ shooting percentage and the ‘Canes’ save percentage – and 82 games may seem like a large sample, but it’s not enormous.
On special teams, the Caps actually underperformed on the power play (maybe not a shock to Caps fans who has seen the paucity of opportunities they have created, especially recently, but here we’re actually talking about underperforming on the chances they did create), and both teams were exceptional on the penalty kill, with Carolina logging the best goals-against rate in the League over the regular season on the fourth-lowest expected rate and the best penalty-killing rate in the League. Oh, and they didn’t allow a power-play goal-against in their 15 times shorthanded in the first round while scoring a shorty, giving them a comical 106.7 net penalty kill percentage (i.e. they were better than 100 percent effective on the kill). Given what the Caps’ power play brought in Round 1, don’t be surprised if Carolina outscores Washington on the latter’s power play.
So what’s the point? Why do we even care about expected goals? Expected goals don’t win games, actual goals do, right?
Of course that’s right. In theory, expected goals are more predictive – they tell you more about what’s likely to happen; actual goals are more descriptive, they tell you what did happen. But a model is only as good as the extent to which, over large enough samples, the two align. So you wouldn’t be wrong to take models that have consistently underrated the Caps (and overrated the ‘Canes) with a grain of salt. (As The Athletic put it, “Anyone who is even fleetingly aware of hockey analytics as a concept knows that the above probability was predictable from both sides. Extreme disrespect for the Capitals coupled with unbridled adulation for the Hurricanes? Double whammy.”
How much salt is up to you (please consult a physician before dramatically increasing your sodium intake). As The Athletic wraps up their overview, “The good news for Washington? The Model has literally never been right about them. Ever. A 33 percent chance has the Capitals right where fans want them.”
2. Picks
Not everyone has their own proprietary model and instead pull their picks directly out of… the ether. So here are some prognostications from the experts, starting with NHL.com, where six of 15 writers are going with Carolina.
At Daily Faceoff, it’s an even five-five split.
Welcome to the resistance, ESPN:
And, rounding us out, The Athletic:
That brings the total to 31 of 56 (55%) going with underdog (by betting odds) Washington. So it’s man and woman (these experts) vs. machine (the models in Round 1, which, of course, were designed by men and women even if they act as if they have no control over the outputs). Guess we’ll know who wins this round soon enough.
3. Game On
The Caps haven’t played a game this month. The ‘Canes have been off even longer. Everything has been analyzed, everything has been said. It’s time to play Game 1… so we can all wildly overreact to an even smaller dataset. Game on.
