My story of the day from yesterday turned into a more extended column, so I decided to post it this morning.
There has always been a conflict between the values of analytically inclined hockey people (let’s call them analysts) and old-school minds (purists). I’ve written before about how my experiences have allowed me to gain some insight into both schools of thought, and in many ways fuse them into my views on the sport, but I wanted to bring up a couple of issues with regards to the questions surrounding Michel Therrien and his coaching following an absolute drubbing — both in terms of possession numbers and the score — at the hands of the Calgary Flames. Friend of the blog Andrew Berkshire wrote a great recap of the game and criticized Therrien for the team’s possession struggles, which is justified. But some other analysts tend to simplify the game down to a variety of semi-predictive statistics without considering other circumstances. I wanted to use this situation to share some more general thoughts on the use of analytics.
First of all, there is an analytics debate, but it’s not the one that so often gets fought on message boards and social media sites. Using analytics is the proper way to go about maximizing a hockey team as much as using a trainer is a proper way to prevent and treat injuries. Not “buying into” analytics isn’t an acceptable view because analytics aren’t something for you to “buy into” any more than, say, the colour blue or the concept of gravity. As John Oliver would say, asking whether you buy into analytics is like asking do you buy that 5 is greater than 15? Or that owls exist? Or that there are hats? Analytics involve scrutinizing data to find trends. Not buying a certain metric or conclusion is fine. Not buying analytics means being satisfied with incomplete information. It’s not a valid position.
But there is an analytics debate, and one that will continue to be important surrounding all the numbers that have begun and will continue to be thrown out there. That debate is over which numbers have meaning, and maybe more importantly, how do you reconcile what meaningful numbers tell you with the underlying — and usually valid — truths about the sport which are as yet impossible to properly quantify?
It’s a relevant question with Therrien because if it were up to a number of analytics writers online, coaching would involve starting the season with lines that, drawing on with or without you numbers, would maximize past Corsi percentage, and then letting the players go out there and play all 82 games and only making changes if the Bayesian Corsi outlook shifted dramatically.
The problem is that such a thing isn’t practical. We’ve already seen the dangers of committing too much to maximizing principles when dealing with humans rather than material products in baseball, and in hockey, where chemistry is so important for player confidence and compete level (drink!) there is a downside to managing players like assets rather than humans.
For example, slumps can turn into massive droughts if a player begins to clutch the stick too tight, or stops going to the dirty areas. Those are material factors that hang outside the realm of variance, and sometimes players need to sit a game out or to try a fresh look with new linemates in order to catalyze the process of slump-busting.
These principles also come into effect when it comes to personnel decisions. The Astros have gotten into trouble on a reputational level by failing to promote talented youngsters in order to preserve contract years or even to pressure those players to sign for cheap long-term. Ryan Lambert wrote an interesting piece on why NHL teams should do something similar. While I agree that NHL teams ought to pay more attention to maximizing assets and doing data-focused cost-benefit analysis, the idea that the Buffalo Sabres should send an NHL-ready Connor McDavid back to juniors should they earn the privilege to draft is a case of a failure to consider other implications.
Twenty-nine teams would start the 2015-16 season with the phenom on their roster, and in Lambert’s hypothetical the Sabres wouldn’t. What NHL prospect would ever want to get drafted by the Sabres again? I’m sure the Erie Otters would be thrilled with the decision, but how about every other team in the CHL? Suddenly, the Sabres aren’t so popular around the league that, you know, produces more than half of the future NHL talent. Another Eric Lindros moment wouldn’t be out of the question, and suddenly by trying to maximize the returns, the Sabres have managed to handicap themselves for years to come.
It’s an element of cost-benefit analysis that most analysts ignore because it is very difficult to quantify, but that doesn’t make it any less important.
So back to Therrien, who has fans asking both why he has Dale Weise on the top line, and why he isn’t making more changes to the lines for a shake-up. Well you can’t really have it both ways. Sure, the opening night lines look like those you might want to start the playoffs, but that doesn’t mean you can go 82 games with them. Coaches don’t get enough credit for knowing their players — their personalities, their motivations, their complexities —and how to bring the best out of them.
Now of course, this has nothing to do with Therrien’s ability to draw a good possession team out of a talented roster, it’s just a general lesson that it isn’t always as straightforward as reading (admittedly impressive and important) numbers off of a chart. Nothing, when dealing with human beings — and a game as complex as hockey — ever is.
I think another important question is how direct a line can be drawn between coaching decisions and results. It may not be desirable (or even possible) to make lineup and deployment decisions based solely on analytics data but, if decisions are made that are at odds to what analytics would suggest and success does not (soon) follow, the question naturally arises as to whether the coach should try it the way the analytics suggest. I think if you were to ask Therrien, he would say that he has to take a long view and his goal is to have a team that plays the way HE thinks a team should play to have success. To me, it’s the articulation of that goal that needs to be evaluated through an analytics lens, not the individual coaching decisions the coach makes on a day-in, day-out basis which, as you point out, have to take intangible factors into account.
Pingback: AP Hockey Story of the Day: March 4 – On the Chicago Cubs, Kris Bryant and the Human Element | AP Hockey