Back in the fall, the Toronto Maple Leafs were considered the latest, greatest test case for analytics in hockey.
They had won games in odd ways in the lockout-shortened half season, with high shooting percentages and low possession numbers, two of the most telltale signs that they would struggle to maintain a 97-point pace over an 82-game campaign.
The best summation of all that came early in the year over at Grantland, where Sean McIndoe wrote a piece that hit on all of the various high notes in the debate and even came up with a clever title for it all: Collision Corsi.
The day that piece was published, the Leafs won for the fourth time in their first five games as part of what would turn into a 10-4 October.
And, despite various peaks and valleys over the next 54 games, there they were in mid-March, on pace for 96 points and past the toughest part of their schedule.
With one month to go, the analytics appeared to have seriously misfired.
Fourteen games (and two wins) later, the Leafs had only 84 points and the eighth overall pick after the draft lottery.
This was a very strange season to cover this team, no matter what side of the argument you were on. There were debates in the press box (and sometimes the pub) almost every week over this aspect or that of the team, and they were split based upon if you believed there was some basis to the analytics or not.
What ended up happening was that the way the Leafs won games this season tested the limitations of what some of these statistics can tell us.
Early on, Toronto won a lot of shootouts, had great goaltending and streaky scoring, which combined to put the Leafs on two 14-game hot streaks – one to start the year and another leading into the Olympic break.
Very little of their success had anything to do with how often they had the puck, which wasn’t very often at all.
“How high in the standings could a terrible possession team finish?” was never a question I’d contemplated before this season, but that’s exactly the kind of strange conversation this team and its playing style started to generate.
In the aftermath of the team’s collapse, my sense is the way many view these new numbers has changed. I’ve received a lot of e-mails and tweets from fans the past two weeks wanting to learn more about analytics and a lot fewer filled with rants over why they are bogus.
I’m not going to attempt to recreate all the legwork others have done here in a couple of sentences to convince the skeptics – perhaps that would be a worthwhile project this off-season – but what I will say is that I understand the reluctance to buy in.
For me, it wasn’t until the 2011-12 season – when analytics folks accurately projected the Minnesota Wild’s collapse, the Los Angeles Kings’ Stanley Cup win and several other key trends – that I saw the undeniable evidence that this analysis has value.
Despite reading about these stats for years, I had to see things play out a few times to fully appreciate their value, and it really wasn’t until that point that I began using Corsi, Fenwick, PDO and others in my work.
And I’ve always gone to some lengths to try and avoid overstating how much they can tell us.
The good thing is that a lot of these numbers are not very complicated. Hockey remains behind other sports when it comes to analytics, and it’s such a hard game to predict that what we’re frequently dealing with is simple ideas like probabilities and regressions.
You learn quickly when you start following this on a closer basis just how much luck factors into results in the NHL. The margins are slim from night to night, the parity is so great and whether a goalie has a good or bad night (or season) can be very hard to forecast.
But what we can pull out of that chaos is some meaningful conclusions about what works and what doesn’t – both over a long season and in the playoffs – and go from there. And, ultimately, the right conclusion on the way the Leafs were built and the way they played was one analytics helped us reach before the season ever started.
Some of it is counter-intuitive at first; some of it can be a little arcane. But these numbers have pointed us in the right direction far too often of late to ignore them.
So we won’t.Report Typo/Error