The Pittsburgh Penguins vastly outchanced the Montreal Canadiens in the second round, which would typically mean a Pittsburgh win. But goals and chances are not one and the same - shooting is a fickle thing, and sometimes players make their shots; sometimes they don't. Pittsburgh vastly underperformed its expected shooting percentage, while Montreal vastly outperformed its own.
These things happen, of course - if we model shooting as though it was random, the range of shooting percentages we see is barely narrower than the actual range we see from NHL teams. Similarly, while individual players have the ability to get themselves into positions where they can have a high shooting percentage, there's very little evidence of shooting talent across the league. In other words, there are hundreds of forwards who are just as capable of putting the puck in from a given spot, even if they're not equally capable of getting to that spot. When we see a particularly high or low shooting percentage, we tend to think that it's not a persistent talent; whether you want to call that 'luck' or a performance high, it's not deterministic.
Of course, when improbable things happen, people want to find deterministic explanations for why they happened. So I decided to go back and watch every recorded scoring chance in the Montreal-Pittsburgh series and make a few more annotations:
Pittsburgh had substantially more scoring chances overall, but their distribution was, as we know, different from Montreal's. Pittsburgh had more opportunities to set up on the PP, which is also reflected in the shot totals: 63-43. Montreal had just as many chances off the rush as Pittsburgh did, but basically conceded the puck along the boards in the offensive zone, while Pittsburgh forechecked to generate puck possession. One thing I found very interesting is how few scoring chances came off odd-man rushes - I counted just 12 (2-on-1, 3-on-2 or 4-on-2) between the two teams over seven games.
One claim I've seen is that Montreal put bodies on Pittsburgh's shooters, which reduced the overall quality of their scoring chances. Montreal did outblock Pittsburgh, 138-98, which made it seem like Montreal was in the way of Pittsburgh's shooters. But Pittsburgh also took proportionally more shots than Montreal, so each team ended up blocking 30% of their opponents shots. For every scoring chance, I noted whether there was some element of coverage on the shooter. The totals:
Coverage is clearly a judgment call, so I could be off by a few percent either way. But the key takeaway is that Pittsburgh was no more likely to have its shots heavily-defended.
Two other common claims about Montreal's opponents is that they didn't screen the goalie, get rebounds or otherwise put a lot of traffic in front of the net. I couldn't find evidence of that:
Again, judgment calls here (aside from rebounds, which are shots taken four seconds or less after a previous shot) and it tends to fall in Pittsburgh's direction.
One place where Pittsburgh definitely does worse than Montreal is in the giveaway department. Now, with Pittsburgh having the bulk of puck possession, it's not surprising that the Pens would turn the puck over more often, but I also believe they had more bad defensive zone turnovers that led directly to scoring chances:
It's ironic that Montreal wasn't really pressuring Pittsburgh but the Penguins still made more bad defensive miscues leading to a small number of high percentage chances. Pittsburgh also did a poor job of finishing the very good chances that they had:
We don't know a whole lot about in-game breakaways, but we do know that there isn't a lot of differences in true talent on shootout breakaways - there's no reason to think that Evgeni Malkin would blow so many 1-on-0 opportunities.
So...to summarize this data and previous results:
1) Pittsburgh dominated the shot counts and scoring chances
2) Based on shot location and the number of rebounds, Pittsburgh had the same caliber of opportunities as Montreal.
3) Pittsburgh generated more PP chances and more chances off the forecheck
4) Both teams had an equal number of chances off the rush, though very few were odd-man rushes
5) Pittsburgh and Montreal were both equally likely to have good coverage on the shooter
6) Pittsburgh had more screens and rebounds than Montreal, while Montreal had more deflections
7) Pittsburgh gave the puck away more often than Montreal, which is expected because they had the puck more often
8) Pittsburgh made more bad defensive zone giveaways than Montreal did
9) Pittsburgh had seven breakaways to Montreal's two, but failed to finish
I didn't note it anywhere, but the "shot quality" model predicts Pittsburgh to outscore Montreal by eight goals in this series. I think the last two items on the list might be worth 3-4 goals, which isn't quite enough to swing things in Montreal's favor, especially given that my shot model doesn't know that the Pens had so many breakaways.
Fundamentally, we have Jaroslav Halak playing better than his established talent level, Pittsburgh's shooters failing to make their shots at their career rates, and Montreal's shooters exceeded their own career numbers. Call it luck, call it personal performance highs - either way, it's not a persistent talent - it's something we all know might come crashing to the ground tomorrow. Of course, that means nothing is certain: nine times out of ten, Pittsburgh or Washington would be in the conference finals...This was time #10.