The Canucks main analytical focus probably isn’t Corsi stats, it’s "fatigue," "readiness," and "peak performance."
On Monday afternoon, Mike Gillis appeared for two hours on the Team 1040. The segment was hosted by Matthew Sekeres who god bless him, spent a sizable portion of the show asking the General Manager and President of the Vancouver Canucks about analytics.
Read past the jump.
To some extent, Gillis parried Sekeres’ question regarding what stats the team pays close attention to in player evaluations. "It’s different for different positions. For forwards it’s a combination of shots on net, quality of shots on net, location of shots on net versus time on ice." Quality shots on net is another indication that the Canucks count scoring chances, and it’s good to hear Gillis qualify that they look at these numbers rated by time on ice. Shot location is a bit of a mixed bag, many are convinced that it’s hocus pocus, though I’ve long suspected that the Canucks pursue players who have been shown to suppress on-ice shooting percentage.
In any event, Gillis appears to be dubious about the ability of advanced stats to accurately build a predictive model for hockey (transcription via Cam Charron and the Province):
We do use advanced analytics to some measure. It’s more difficult in hockey than in baseball because baseball is a defined event. You’ve got 100 different things that go into player success. Who they play for, match ups they constantly play against. Their age. Injury history. So you’ve got lots of things that are determinant factors in hockey that can’t be properly analyzed just through analytics. In baseball you can.
The notion that hockey is too spontaneous and free-flowing to lend itself to statistical analysis is a popular one, but it’s a narrow and overly restrictive paradigm in my view. Accurate predictive models have been built to measure, better understand and predict the outcomes of things like the voting behavior of an electorate of 350 million people, or global financial markets. Yes those aren’t sports, but my point is that areas of macro-interaction that are significantly more chaotic than several seasons worth of on-ice events at the NHL level have been modeled accurately, so it seems odd to think that hockey simply cannot be. Corsi, Fenwick, Zone-Entries haven’t reached Nate Silver levels of accuracy yet, no doubt, but they’re steps in the right direction.
Where Gillis’ comments on analytics become more forthcoming and thoughful, however, were when he got started on the topic of zone-starts and player development. The Canucks manipulation of zone-starts to gain an offensive edge is well documented and unique, as is Alain Vigneault’s massaging of match-ups: "What we’ve done is look at things and try to design success based on where [young players] are starting, and who they’re playing with and what situations they’re playing in and the number of minutes they play."
It’s worth noting that as recently as the 2009-10 season, the Canucks zone-start deployment schemes and patterns were pretty standard when held up against the rest of the league. Alain Vigneault told CanucksArmy that his zone-start deployment patterns were a matter of "personal preference" rather than an example of some sort of organizational philosophy, but I’d be curious to learn more about whether or not he’s executing (with an impressive degree of discipline) a scheme concocted in concert with the Canucks front-office.
Finally, whenever the topic of advanced thinking, and how it informs Canucks management strategies is broached, the talk seemingly always turns to "Human Performance," and "fatigue." Yesterday on the Team, Gillis shared with the audience what he’d learned from looking at advanced analysis in soccer:
Well oddly enough we have looked at [passing efficiency] in soccer. And we put that in a very different context, we’ve looked at it relative to fatigue and conditioning and how you’re percentage of passing success is relative to your conditioning and the time in the game when you do it and how many minutes you’ve played. There are studies that we’ve looked at that indicate that passing percentage in soccer goes dramatically down depending on the time in the game or depending on the conditioning of the player…
We’re trying to define fatigue levels in those circumstances and as you know, a player usually gets hit twice when he gets hit once. He gets hit by the player and then hits the boards. How you can attribute that to success and how you attribute that to fatigue levels is instrumental in finding out when a player in the third period makes a mistake. And something happens and I think that as we’ve found, in a dynamic, competitive contact sport that fatigue levels are really a lot of the determining factor in success or failure.
Mike Gillis’ sleep doctors were one of the first "big innovations" that were instituted when he first took over as General Manager four years ago. This past season the team was subjected to a new type of readiness testing, in addition to their military type sleep schedules, "mind-rooms" and witch-doctor like "sleep bracelets" that monitor players circadian rhythms.
After the team slept through the final four months of the 2011-12 regular season, by Mike Gillis’ own admission, he spoke about the teams desperate need for an improved "Human Performance Plan" designed to help the club deal with a difficult travel schedule, the pressure of playing in Vancouver and the grind of playing extremely physical games (from April 24th):
"I’ve been working on a human performance plan to try and address those issues. I think fatigue was the first stage in it, about how you deal with the ups and downs. One of the things I’ve concluded is that compounded with the types of games that we play and the pressure – we have to find a better way to deal with that, and we have to find it quickly. That’s one of the things we’ve worked really hard on, it’s more complicated than you think and I’m hoping that we’re going to have that plan in place for next year."
While the Canucks are aware of and use possession stats, zone-starts and quality of competition, it seems that they’ve been primarily concerned with gaining an exploitable knowledge edge regarding their own personnel’s "peak performance." As Cam Charron points out in his excellent recap of Gillis’ 1040 appearance over at the Province, Gillis’ on-going analytical focus on fatigue has paid spectacular dividends during his tenure as GM:
But the big improvement for the Canucks in the last four years is in the third period. Pythagorean Expectation predicted 50 wins from the Canucks in the Gillis era per year (actual, 49.8) and 43.2 in the four previous seasons (actual, 43.3). That’s an increase of 16%.
But where did the increase come from? They actually lost ground in the second period, their win expectancy dipping by 8% in the middle frame in the Gillis era, but that’s propped up by a 23% gain in the first period and a 35% increase in the third period.
The median NHL team won an equivalent of 42.2 games pro-rated to 82 games in the first period over the last four years according to Pythagorean Expectation. In the second period, that dipped to 41.6 and in the third and overtime it dipped to 41.1, so perhaps the average team does get impacted by fatigue.
In March, Oakland A’s General Manager and Moneyball protagonist Billy Beane spoke at a Yahoo! event and suggested that a better understanding of player health and injuries was the next frontier in advanced analytics in baseball. The Canucks efforts to apply advanced thinking and statistical analysis to issues of fatigue and readiness or "peak performance," strikes me as a not so distant cousin of what Beane was getting at.
Clearly this is a topic that the Canucks have targeted with an arsenal of research dollars, and a willingness to implement it at the management level. That they’ve been so successful, and open-minded enough to apply lessons from other sports like soccer, is fascinating.