Last Friday morning, Karl Alzner stated that he doesn’t believe in hockey analytics, and it was the least important thing that happened on a day that nothing important happened.
From Isabelle Khurshudyan’s piece:
“I don’t believe in any of the advanced statistics,” Alzner said. “I don’t believe in a single one of them.”
Alzner then brought up that the Capitals had 80 percent Corsi, a statistic used to measure shot attempt differential while at even strength play, in a recent 5-2 loss against the Rangers, “so it doesn’t matter.”
Alzner certainly isn’t the first player to voice his antipathy for advanced statistics. Most recently former NHL-er and renowned fist-holster Paul Bissonnette made public his doubts that analytics could tell anything that a good scout couldn’t see for his or herself. Alzner and Bissonnette certainly won’t be the last. Analytics in hockey have quickly achieved ubiquity. Put another way, something that hockey players have spent most of their hockey-lives without now permeates their profession, and sometimes their livelihoods.
Change is often greeted by scorn.
But here’s the thing. How players view analytics is completely irrelevant. In hockey, analytics are used to measure what a player has done, and predict what they are likely to do. What analytics don’t do is directly inform a player how to play.
We pointed out last summer that Barry Trotz does seem to possess a functional awareness of analytics, and a willingness to incorporate them into his strategic thinking.
“I found out, through stats, that we had a defensemen who would dump a little more often than he probably needed to. So then we really tracked it, on a one on one basis, and found that he had more time to make better quality plays.” (Barry Trotz)
While here is an example of analytics dictating a strategy change, it’s gotta go through the coach first, which is pretty much the entire thrust of my point. By the time Trotz, or Todd Reirden, or whomever, tells our mystery defenseman, “hey, you’re dumping the puck too early, don’t be afraid to hold onto it for an extra second and look for a better play,” the player probably isn’t aware that the information was extracted from a spreadsheet somewhere. He’s just being coached, same as he always has.
That leads to another interesting question: do you even want your players to be aware of analytics, lest the Hawthorne effect wind its tendrils round their impressionable minds?
Although hockey analytics encompass a broad spectrum, their most common mainstream applications are understanding puck possession, good and bad “luck” (which is actually just good or bad goaltending or shooting), and productivity on a rate basis rather than raw.
Of course, the most useful analytics are those that can be actionable. For example, we recently looked at how possession metrics can inform how forward lines and D-pairing would be most effectively deployed together. Zoomed in a bit further, they can also inform the ideal makeup of the forward lines or pairings themselves. Analytics also offer insights into effectiveness by zone starts, or by how the puck is carried into the offensive zone.
Speaking of zone-entries, a few years back Nashville Predators‘ analytics consultant Eric Tulsky provided some great insight into the relationship between data, coach, and player in an interview with the Nashville Post. Here’s a particularly pertinent excerpt:
So when Karl Alzner says, “I don’t believe in analytics” it’s defiant (particularly considering his coworker is the dude who invented Corsi), but harmless. Alzner’s ire smacks even more of irrationality because he cites a single game as evidence that the system’s broken. Karl Alzner is never going to make any decisions because of his own belief or doubt in “advanced” stats (or at least he shouldn’t). That’s his coach’s job.
Paul Bissonnette’s overall point—and a valid one—is that possession is a product of every single player on the ice, but is often applied to understand a single skater; basically the same argument for why +/- is a bad unit of measurement. The information comes with caveats, with a grain of salt. But that’s simply called context, and professional hockey clubs should (and certainly do) have the wherewithal to understand that. To claim that because the information is subject to the effects of variables the entire enterprise is rendered implausible is absurd (and to Bissonnette’s credit he does concede that the data has a place in the game).
You’d like to see Alzner acknowledge a statistical movement that undeniably has a place in today’s NHL, but ultimately his staunch rejection of it is not unlike a warehouse laborer scoffing at a metric used by finance departments as a best practice across his entire industry; it has no impact on the bottom line. The role of analytics in hockey is to provide insights. It’s up to the organization to decide to what degree these insights will be valued, how they will be obtained, and at what level they will be obtained. The coach is undoubtedly involved in determining which insights can be translated to an on-ice change or optimization, and in an event where something is in fact deemed actionable, he is responsible for developing that action.
The player’s role is unchanged. It remains “do what the coach says”, whether or not he believes in the philosophy or data that informed the instruction.