Here's what we know about the Avalanche so far:
- They're a playoff bubble team, who might be able to make it with help.
- They are a poor possession team.
- They were quite unlucky early and have been getting ridiculous goaltending lately.
- Brad Stuart might be a black hole in human form.
Let's break those things out over time and see it visually. As always, the numbers presented are at 5v5 only, the possession stats are score-adjusted numbers, and everything comes from the marvelous War on Ice.
Colorado are pretty much a 44% possession team at this point, which is Real Bad. Few teams in the league sport a worse share of possession. They peaked above 45% around game 28, and have gone down since then. To explore that, let's look at the 10-game moving numbers to spot season trends.
First the variance:
I didn't multiply the PDO by 100 or 1000 because, as you can tell by my spot on this masthead, I'm fucking lazy. The team were really unlucky when the season began - which happens when it takes you 3 games to score a goal - and ever since game 10 have been about at the 1.000 mean. In fact, as they've won games recently, you can see the PDO tick up above 1.030; they've won these games thanks to Semyon Varlamov, which I would like to believe has been convincingly proven unsustainable at this point. He's really good and he's going to go on runs like this. You just can't count on it forever and need to be able to help him out to win long term.
So lets look at the stuff with more long term predictability. Yeah, Corsi again.
Look at how much they improved in the middle of the season! They have at no point had a positive 10-game differential, but that stretch from game 20-33 or so is where they come the closest to breaking even every night. At the time, I attributed this to the change from man to zone.
That sounded great back then, but unless we've missed out on news of a giant system change, the zone is actually performing worse right now than the man-to-man system ever did, allowing almost 180 more shot attempts against than the Avalanche produce every 10 games.
That's easy math to do: it's allowing almost 18 more shot attempts against than it produces every game. Based on this season's leaguewide numbers, that's an expected goal difference of about -0.75 every game. That's a lot and it's Bad.
So why did the zone defense system have such an immediate impact? And what has changed through the season that might be the reason it canceled back out?
Theory: Brad Stuart is a human black hole. If he is in your defensive zone, no puck that crosses the event horizon (blue line) will escape.
Okay, yes, this is a bold statement. Stuart certainly doesn't look like a star that exploded and collapsed in on itself. But let's think this thing through. What else was happening to the team when it was showing improvement at puck possession? Brad Stuart was out with an injury. The numbers check out: He got hurt and missed games 15-30. Check the 10-game above: it lines up.
I knew Stuart's relative possession numbers were singularly (singularityly?) awful. I attributed a lot of this to his deployment. Then I made this pie chart. This is not a normal WOWY. This is the entire team over spans of games. On the left, we have the team's possession in games that Stuart played. On the right, the team's possession between games 15-30, while he was out of the lineup.
With Stuart out of the roster, the Avalanche were a 47.4% Corsi team. That is bad, but merely bad. It's not Bad, or Awful, or Terrible Horrible No Good. But the rest of the year? With Stuart on the roster, the Avalanche have been 42.1%.
forty-two, point one, percent.
I checked. Colorado are 29th in score-adjusted CF%. Their score-adjusted possession is only better than Buffalo, who are a historically shocking 36.9%. But I checked all the way back to 2007. You know how many teams have been 42.1% or worse?
Two. Buffalo this year, and Buffalo last year.
So without Brad Stuart, the Avalanche are a bad possession team, but one within reason. When they have him in the lineup, though, they approach Historically Bad levels. (His dCorsi Impact, for the curious, is -125.25, which is TWICE as bad as anyone else on the team, AND WEIGHTED LESS THAN OTHERS BECAUSE HE MISSED 16 GAMES.) I believe this evidence supports my theory, that the blue line in front of Varly is an event horizon, past which pucks cannot escape if Brad Stuart is in the defensive zone.
I didn't set out to write this conclusion, but it's what I found. It was inescapable.
A confounding factor, mentioned to me on Twitter by Do Not Feed The Octopus, is Zach Redmond. He asked whether Stuart's games out correspond with Redmond's games in, and they do, quite well. Outside of his 4 other sporadic appearances, Redmond played games 15-36. But I don't think Redmond is the difference maker here. He is deployed more offensively than Stuart (more on this in part two). His impact is fairly minor due to lower ice time, and his positive impact on possession appears to be more in generating shots than in preventing them (he is right about at expected CA).
But hey, it might be the net difference when you add Redmond and subtract Stuart that make the Avalanche almost respectable at possession, not just the one guy. It's a factor that's difficult to control for. But I think the Black Hole Theory has some support.