clock menu more-arrow no yes

Filed under:

Duncan Keith’s viral Corsi joke raises interesting point about value of NHL’s proxy stats

Keith, asked about his 15 shots over the Blackhawks’ last two games, joked that he was “getting the Corsi up, so you guys think I’m good.” In doing so, he raised a good point about Corsi’s key weakness.

Duncan Keith’s individual shot attempts shot way up in the Blackhawks-Blue Jackets series.
Duncan Keith’s individual shot attempts shot way up in the Blackhawks-Blue Jackets series.
Jeff Haynes/AP

Defenseman Duncan Keith was just trying to make a joke.

But in doing so, he raised an interesting point about the merits and shortcomings of hockey’s proxy stats.

First, some context: Keith has visibly made an effort to shoot the puck more often recently. Blackhawks coach Jeremy Colliton noted it Sunday, and the data back him up.

Keith took five shots in the Hawks’ first game against the Blue Jackets, then 10 in the second game — leading the team with six on goal — after taking no more than three shots in any game this season.

There’s a real reason why, too.

“It started on the power play,” Keith said Monday. “We didn’t really have too much [going], a couple of entries didn’t work out, and I just figured that we needed to start getting some shots. I shot a few, and we were able to get a little momentum, and then a few other things opened up. . . . It’s important that if I have the shot there, I have to use it.”

But before giving that thorough answer, Keith made a quip that went viral in the hockey world and ignited a years-old argument on Twitter, Reddit and beyond:

“Getting the Corsi up, so you guys think I’m good.”

Corsi, of course, is the statistic that measures shot differential (it’s like plus-minus, but for shots instead of goals) while at even strength. It’s purely quantity-based.

In recent years, analysts have somewhat shifted away from Corsi and toward scoring chances and the more complex “expected goals.”

Those stats incorporate quantity and quality and give a more accurate, comprehensive evaluation of performance.

Yet Corsi remains the buzzword for the entire family of statistics, making it the most recognized and the most contentious.

Keith is far from the first NHL player to criticize Corsi for its overuse. Fellow Hawks defenseman Calvin de Haan memorably did the same thing last season in a far longer rant. (De Haan chimed in again Monday, tweeting, “Dunks isn’t wrong.”)

Keith is, however, perhaps the first NHL player to imply — even clearly jokingly — that a desire to maximize Corsi could theoretically affect decisions on the ice.

That brings up the philosophical maxim colloquially called Goodhart’s Law: “When a measure becomes a target, it ceases to be a good measure.”

It would certainly be possible for an NHL player to slightly manipulate Corsi if he so wished.

Consider a scenario in which he has the puck on the perimeter of the offensive zone. He could shoot, minutely improving his Corsi ratio but with almost no chance of scoring, or make a pass, forgoing the guaranteed shot but potentially leading to a better scoring opportunity.

And while there’s no documented evidence of a player consciously manipulating his Corsi, there would be plausible motivation.

Corsi and other similarly “advanced” stats are often brought up in contract negotiations — and players are keenly aware of which stats are and aren’t, as Nikita Zadorov evidenced Monday when told he ranked in the top five in hits per game.

“I don’t think that stat’s going to . . . give an extra $5 million,” Zadorov joked. “It’s not points or goals or anything.”

Therein lies Corsi’s problem.

As a proxy for possession rate, it has definite value, at least until the NHL’s new player- and puck-tracking data become publicly available. But the data points that serve as its foundation — shots — are not, depending on context, always valuable themselves.

Conversely, there’s no question scoring chances are always valuable. A player who “manipulated” his scoring-chance ratio would have simply become a better player — the very thing proxy statistics intend to measure.

Keith probably didn’t think about any of that Monday. But his viral quip raised a good point nonetheless.