Trying to explain how NHL teams use analytics

It's still a blackbox of data most of the time. But we at least have learned a little bit more.

One the frustrating things about trying to explain hockey analytics is the lack of real-world examples of use by NHL teams.

Teams are stingy with that information. They don’t, and won’t, share it on purpose, and as analytics becomes more and more involved in NHL decision making, protecting proprietary information becomes even more vital.

When Prashanth Iyer and I had Carolina Hurricanes assistant general manager Eric Tulsky a guest on “Expected By Whom?” last month we asked him about the base level of analytic usage in the NHL.

“I hate to do this to you, but I really don’t know,” Tulsky said. “There are teams that I think have research group that I respect a lot, that clearly has a lot of horse power, and I imagine they wouldn’t be making that investment if they weren’t using it. But I honestly I don’t know. For all I know they aren’t listening to that group at all.”

(If I embedded this video right, it should be right around that part in the clip below…. if I didn’t embed right, it’s right around the 49-minute mark.)

Occasionally we get bread crumbs.

A coach will mention that a player’s underlying numbers were pretty good, an executive will mention analytics playing a role, but when pressed for more details, things clam up.

For example, in Detroit, we know there are certain analytics and underlying numbers that Red Wings coach Derek Lalonde prefers for player assessment. I’m pretty confident it’s related to scoring-chance creation, but whether it’s expected goals, high-danger chances, or something else, we aren’t completely sure and I’m only using context clues.

I sat down with Dallas Stars director of analytics Matt Rodell for “Expected By Whom?” at the NHL Draft last month in Nashville, who in the past season was essentially asked to re-form how the Stars approached analytics in his new role.

The Stars had used analytics before the 2022-23 season, but Rodell’s move from assistant video coach to director of analytics made it a larger part of the day-to-day operation.

Rodell’s recent background as a video coach allowed him to better connect the visual side of the numbers, how to explain data with video, and his elevation from within the team ranks, made adoption and understanding easier according to some others I’ve spoken to within the Stars organization.

On the day-to-day side, the Stars use analytics like any other team when it comes to scouting, identifying certain tendencies — analytics are just stats, stats are just tendencies, if you really boils it down — but it’s also heavily project based.

What those projects are, the Stars won’t share, but many times they come directly from Pete DeBoer and sometimes have been driven by players. Those projects may have been implemented in a game, they may not have, unfortunately teams won’t share that information.

The other thing about analytic-usage comes from a roster management and player acquisition piece. In a hard cap world (at least for a couple more seasons) teams have to find ways to capitalize on market inefficiencies, especially in free agency.

For example, when Valeri Nichushkin was bought out by the Stars after not scoring a single goal, or committing a single penalty, during the 2018-19 season, a lot of the scouting view got caught up in the zeros. The lack of goals, the lack of penalty minutes, the fact he was often an unnoticeable forward, popped for scouts that this player wasn’t going to cut it.

Now, to be clear, the Stars buyout made sense, they needed that move to sign Joe Pavelski — that’s a move Dallas doesn’t regret. But the Colorado Avalanche, according to league chatter, took their bet on Nichushkin by looking at some underlying numbers, about how he suppressed offense for other teams and how even a minor normalization in offense would make him a low-risk, high-reward signing.

This offseason the Stars rebuilt their fourth line, signing Craig Smith and Sam Steel to come in an replace an old philosophy that was more built on a pure checking line. Smith and Steel, on pure counting numbers, aren’t big names, but data-wise both improve the Stars.

Using public-facing data like HockeyViz, Steel grades well in offensive creation, in general the team has been better with him on the ice.

(I like using Micah’s work because he does a good job visualizing, hence the name, player impact for those, like myself, who need it laid out to them like they are a toddler. Below red is good, blue is bad in the offensive zone. Blue is good, red is bad in the defensive zone, for the simplest explanation.)

Same with Craig Smith, who admittedly is also a scouting favorite, and has been for much of his NHL career.

We sometimes get into dangerous trap of giving someone all the credit for everything good and none of the blame for anything bad. So assuming the Stars only signed Steel and Smith for analytic reasons is faulty, but it is noteworthy that this type of signing came in the first full offseason with a full-time director of analytics.

This brings me to the communication and cohesion element of an analytics department.

Every NHL team has an analytics department of some sort. It’s not uniform and in some cases, like with the Winnipeg Jets, it’s primarily handled by the video coach. In other cases it’s handled off-site.

But as the league moves forward, having an analyst on site, in the room where decisions are made, is becoming more common. It’s an attempt to normalize the extra data, sort through the noise, and like some other NHL elements — shiny new practice facilities, for example — it’s going from a surplus advantage to a basic necessity.