Expectations for tracking data at the World Cup of Hockey (Trending Topics)
It was announced on Wednesday that the NHL will be using player and puck tracking data at next month’s World Cup of Hockey, and allow ESPN and Sportsnet to tap that information for use on its broadcasts. On the surface, this is a good thing.
It’s better to have data than not have it, but what is actually collected from the puck tracking, and the chips sewn into players jerseys, is not yet totally clear. John Matisz of the Toronto Sun reported that some of the information will include puck speed, skating speed, puck trajectory, spacing between players, who’s on the ice, TOI, and so on.
The issue, then, is how it’s used, and what that information can tell us about the game that we didn’t already know. For example, having the numbers will be nice because we can say, “That Shea Weber shot went 96 miles an hour,” and “Erik Karlsson carried the puck through the neutral zone in 1.2 seconds,” and that’s nice information to have. But it doesn’t tell you anything you didn’t know generally, right? We know Weber can shoot the puck hard and Karlsson is fast.
Similarly, puck trajectory data will probably be able to tell us in real something like real time when a shot was deflected out front, if only by a few inches. But again, if you’re used to watching this sport on a regular basis, you’re usually going to have some idea of whether something looks like it hit a stick or a skate on the way through a scrum.
Even player distance information might be more useful as a factoid than in actually evaluating the sport. Fans know a soft gap when they see one, and they know that if you give certain players — like, say, Johnny Gaudreau — too much room they’re going to waltz by you like you’re not even there. So if we look at a highlight and say, “Pietrangelo gave Gaudreau eight feet of space on this goal,” that might not be enough information to actually tell us anything useful.
The problem with “micro-stats” in hockey these days — the kind that comes with player tracking technology — is that you can roll it out all day long without actually having it amount to anything. In this way, it’s more like trivia (“The capital of Zaire is Kinshasa”) than useful knowledge (“This is how to fix a carburetor”). And to some extent, we have this already in the NHL. Companies like Sportlogiq provide reams of data in the NHL about controlled entries and exits, successful passes, and so on, but don’t usually go too far out of their way to explain how that translates into goals for or against. That is, or at least ought to be, the ultimate point of data analysis in this sport.
The reason people have taken so much to tracking possession through shot attempts is that there is a clear A-to-B-to-C you can work through. You want to know what leads to wins? It’s the ability to both score and prevent goals. So what leads to scoring and preventing goals? Shooting the puck and not allowing your opponent to do so. Over a long enough timeline, your winning percentage, goals-for percentage, and corsi percentage become the same number with very few exceptions (i.e. the Rangers will have a better-than-they-deserve goal differential because Henrik Lundqvist breaks the fundamental rules of the sport).
So the question is: What does puck or player speed, the distance between players, and so on actually mean in terms of scoring and preventing goals, and subsequently winning and losing games? We don’t know, and can’t know at least until the broadcasts begin. It’s just not data we have. And frankly we don’t know whether it ever will be.
Will the NHL make this player- and puck-tracking data available online, much as it does all other stats it currently collects? And with respect to SAP’s advanced stats data on NHL.com, will it even be accurate? Clearly the league is using the World Cup as a bit of a test run, to see if this is a tracking system that can be rolled out to all 31 teams at some point in the indeterminate future.
There are potentially plenty of practical applications to this data that could, in theory, be sussed out if the league allowed the general public to see it. For instance, you could determine whether shooting hard from the point is necessarily going to result in more goals than shooting with a little less velocity and potentially a little more accuracy. NHL teams could then theoretically use that information to make their power plays more likely to score, or tailor their defensive structures to defend against that information.
This is the same issue with Sportlogiq data. We don’t actually have public access to it — and fair enough, it’s a company trying to sell a product. So while we intrinsically understand that controlled zone exits are better than dumping the puck into the neutral zone, the public doesn’t get to know how the actual math behind what that does to teams’ goal differentials over the course of a season.
That fact has led to some stats-minded people criticizing this kind of data as “snake oil,” though I wouldn’t go that far. Again, if you’re selling something it makes sense that you not draw direct lines between micro-stats like zone entries and macro ones like goals and wins. To put it in high school terms, it doesn’t behoove their bottom lines to “show their work” to people who aren’t paying for the privilege of seeing it. It doesn’t mean there’s no correlation between these things, but it allows for a situation in which some people can swear the data is useful, and other people can dismiss it as inconsequential to win percentages.
The NHL is in no such position. If league-wide tracking becomes A Thing this season or next, there’s no reason it shouldn’t post the data it collects on NHL.com and let people figure out the A-to-B-to-C-to-D of that data, if indeed there is a D worth examining at all.
Another issue here is that all this information is quite simply too new to really put into any useful context right now, but that likely won’t stop people from trying. The issue of snake oil comes up here, to some extent, because people in the media or working for teams can use early-stage data here and say to their audiences, “This leads to that.” The sample is likely to be too small to be meaningful, and probably would be even after 82 games.
Again, it’s better to have information than not, even if it allows us to throw out that information as useless to actually analyzing the game. At that point, it’s just fun to know a fact. And that’s fine too.
Ryan Lambert is a Puck Daddy columnist. His email is here and his Twitter is here.
