After an unexpected hot start, my blogging activity has finally regressed to its mean of zero. Sure my post percentage peaked for a while, but you could tell from my career average that it was completely unsustainable and due to fall apart. Ladies and gentlemen, advanced stats!
Self-deprecation aside, chases3 on MHH recently mentioned some idle curiosities over c6hor8’s recap of the Senators game that elicited a response. Before I knew it, my reply had ballooned to approximately ten bajillion words and I realised that it’d be a great excuse to start posting over here again after months of neglect. So without further preamble, here’s the quote.
c6hor8: Ottawa had a bunch [of shots] down low but I am not sure how many of those were legitimate rebounds. I am no suggesting this is the best way to win games, but look below.
chases3: Really? Because it sounds like you’re suggesting this is the best way to win games. The Avs have done it at least a dozen times this year. I know – we’ve had this debate before. And personally, I don’t have a side. I’m just befuddled by the fact that the Avs keep winning games while being massively outshot and out-Corsi’d. I’m waiting for the statistically-proven regression, but it hasn’t happened yet, and there’s nothing (other than Corsi itself) to suggest that it’s going to happen anytime soon. I’d be very frustrated if someone on the pro-Corsi camp used an early-round series loss to, say, Chicago, as proof that the Avs are outplaying their expectations, because that would be such a small sample size, and pro-Corsi people know better than that…I think. I just think there are too many variables on a year-to-year basis to use a hypothetical bad year as “proof.” Besides, in the end, we’re in the playoffs, so doesn’t that discount Corsi’s effectiveness at explaining “good” teams anyway?
Like I said, I have no side in the debate. But I sure as hell would like an explanation for why the Avs keep getting badly out-Corsi’d and yet win so many of those games – many of which aren’t even close contests.
Conventional wisdom says it’s that Semyon Varlamov has been spectacular, similar to how the Bernier/Reimer tandem has been driving Toronto to the top of the East. The other rationale that people (including the Avs themselves) are championing is that Patrick Roy’s defensive system keeps opponent shots to the outside and limits the quality of chances against, which is entirely possible. Analytics research indicates that teams are generally unable to control shot quality, and that quantity (a proxy for possession) and quality (should intuitively require possession to set up) should go hand in hand anyways. But c6hor8 and others have argued that Roy’s system is a unique situation in terms of being able to control shot quality against, which bears its own investigation.
I’ll say one thing right off as one of these nebulous analytics people – the shot quality chart that c6hor8 has drawn for us is uh … kind of the very definition of skewing data. Several Ottawa shots that are right outside the border of the ‘high risk area’ with everything near the crease selectively excluded does not a good indicator of shot quality prevention make. What this tells me is that Ottawa just outshot the Avs from every area of the ice, including from in close. They played well enough to limit the Avs to 6 or so shots from quality areas – the difference is that 3 of those went in while Varlamov was a brick wall on the other side. Hey, those sound like goaltender effects to me!
Ottawa also missed the net on a tonne of transition plays (the ones that presumably create shot quality for teams that get hemmed in often), which will also account for some of it. I don’t think that had anything to do with the system, which mostly controls zone play, so much as a bad shooting night for Bobby Ryan et al. Could be wrong on that note though, I’m not by any means knowledgeable there.
Also, statistical regression assumes that shooting is a percentage that can’t be controlled. There are exceptions, and the most prevalent one is that your goaltender is just playing out of this world, which is easily capable of buoying a team’s save percentage. I should also mention that while regression is a proven trait for percentages like these, that is not the same thing as being guaranteed. If a player who’s shot 12% on his career hits a 18% clip over 20 games, he’s still most likely to shoot at around 12% over his next 20. But that doesn’t mean he can’t just stay hot and continue on at 18% – it’s been shown to be predictive, but not determinant. (That said, the idea that hockey has too many variables to analyse with stats feeds into this – an overabundance of variables means statistical noise means randomness.)
Anyways, now to answer the actual question.
My first point to make is that the Avs’ possession over the entire season hasn’t been as bad as advertised – they were a 50% Fenwick close team for much of the season, until a nosedive around game 30, which is an interesting anomaly: they’ve played a chunk of the season with even possession and then a chunk where it collapsed and went to shit, and kept winning – so is it that the players have gotten worse, Roy’s system has trended towards giving up more shots, or some other phenomenon? To me the winning aspect at least speaks to quality of goaltending, staying consistently excellent even as the team’s possession game has deteriorated.
Colorado’s Fenwick% close also trends down because they have played 794.4 minutes leading by 1 and a scant 480.6 minutes trailing by 1, which is important because teams tend to collapse with a lead and press hard when they need to catch up. When down 1, the Avs control 53.5% of Fenwick events (shots on goal and missed), which is 13th in the league. Leading by 1, their Fenwick% is abysmal at 43.1%, 27th in the league. With the amount of time they spend leading games, of course their Fenwick% close (close as in up 1, tied, or down 1) is going to be pretty terrible. With that in consideration, they aren’t an elite possession team but certainly they aren’t sitting with the Oilers, Sabres and Leafs of the world either.
It’s a question of whether this is an artifact of Roy’s system, and there’s some evidence towards yes – trailing by 1, the Avs’ team save percentage is .917, good for 18th in the league. When turtling to protect a lead, Varlamov and Jean-Sebastien Giguere have combined to stop .955 of shots, which is just barely behind the Rangers for second in the league and a patently ridiculous number. With this degree of difference, the stats do show that Roy’s system protecting a lead could be protecting his goalies by giving them easier shots.
With a 1-goal lead, the Avs have allowed a categorically absurd 822 shot attempts. Next is the Leafs, with 747, and third is the Habs, with 692. Speaking of being up by 1, the 794.4 minutes the Avs have spent leading by 1 are tops in the league by far as well – the Bruins are a distant trailer at 694.9 minutes. You can probably tell where I’m going with this – the Avs take a lot of one-goal leads and absolutely bleed shots against in those situations, but hover around league average otherwise – they can win games despite ostensibly poor possession stats because these stats are highly inflated by score effects. It’s pretty fair to say that this points to a situational element in their poor Fenwick% close – namely, whatever adjustments we make when protecting a lead.
So is it a suppression of shot quality that’s allowing the Avalanche goalies to feast on easy saves? It’s difficult to say this – some great work from Jibblescribbits indicates that the Avs in fact allow more shots from close in than most teams, a far cry from the idea accepted among their fanbase that they’re excellent at keeping shots to the perimeter. Also troublesome is that they still rank near the bottom when it comes to Fenwick% tied, which removes score effects from the equation and also accounts for a further 1169.3 minutes played.
There are a lot of things that make the answer inconclusive, and realistically, it’s probably some combination of all of the above. As for the Colorado’s winning ways, there’s more to life than Corsi – an understated point is that possession stats only account for 5-on-5 play, which means strong special teams can be a great equalizer. The Avs boast an excellent power play (20.8%, 3rd in the league) to supplement an average penalty kill (81.1%, 19th in the league), symptomatic of their elite forward core and mediocre back end. Chalk another one up for why the Avs’ inexplicable record isn’t so inexplicable.
Anyways, I’m tired and it’s late, so off to bed. Hope this sates a few curiosities out there.