clock menu more-arrow no yes

Filed under:

Lies, Damn Lies and Data Viz on TV

New, comments

The NHL is trying to modernize how it collects and presents statistics, but they’re adopting technology the rest of us already find obsolete.

Colorado Avalanche v Anaheim Ducks Photo by Victor Decolongon/Getty Images

At the All-Star Game this weekend, the NHL is going to give television viewers an opportunity to preview their new puck and player tracking data that will become available to their gambling partners and for some media use in the next season. There has been no clarification yet if any of the data will ever be public.

The broad strokes timeline of the introduction of tracking data was announced along with the partnership with MGM in October:

The preview of the tracking technology on display this weekend is described this way:

The technology will enhance the storytelling, bringing an unprecedented level of precision to description. For more than a century of NHL play, people have had to rely on adjectives for speed and estimates for distance. Now the NHL has the tools to give exact measurements.

”I think we’re going to leave it to them to decide what to show and when just based on the natural flow of the skills competition and the games themselves,” said David Lehanski, NHL senior vice president of business development and innovation. “Obviously they’re the experts in storytelling. They’re the best at dealing with short stoppages and unique twists and turns during the game.”

This puts this new technology into perspective nicely. For as long as the NHL has put players on the ice, they’ve provided data. Over the decades the data has grown, not in complexity so much as in simple quantity of things that are counted up and tracked. This increase in data points tracked in-game by the official scorers culminated in the public availability of that data for the last decade or so. This is the source of most of the statistical information about hockey that has come to be called analytics. Corsi, WAR, whatever — it all starts with what the official scorers track manually.

During the ASG, we’ll learn which player skated faster or how far they travelled in each game. It’s safe to say the broadcasters storytelling with this information won’t ever think to ask themselves why it matters, and they won’t be giving any contextual meaning to the viewers. They might show a nice chart with sophisticated visuals and player images only they’re allowed access to while they give you statistics on who shot the puck the hardest. And lots of viewers will infer meaning from this data because faster is better, right? More is good, and the guy on top of the chart is better than the one on the bottom.

For now, we won’t have enough data to provide context to the descriptions. OK, McDavid is skating X miles per hour. We knew he was fast, we know how fast he is now, and that adds to our level of appreciation of him as an athlete and of hockey as a sport. What we still don’t know is how that compares to his record speed, his average speed, his slowest speed, his last shift and so on, let alone how it compares to other players.

You won’t get context. And judging by this, no one is ever likely to care about significance, either. Without understanding which measurements matter and which don’t, they might just as well tell you the combined plus/minus of the players on the ice together. Tracking information presented this way will provide TV broadcasters with more numbers that seem important while leaving the viewer no wiser.

The meaningful aspect to tracking technology is the refinements it will provide to the existing data on shot location, the chance to add passing data to the pool of information available, and a possible increase in reliability of the data itself.

For decades, we’ve mostly used two statistics: goals-against average and save percentage. In recent years, we’ve refined that somewhat, but even measuring, say, save percentage on high-danger scoring chances has been imprecise. Now we’ll be able to know exactly where each shot was taken, how fast it was moving, whether it was deflected and so on, and we’ll be able to sort by game situation.

The increased precision and detail on shot types will be welcome if this tracking data is ever made public, but the NHL is promoting its use to upgrade a statistic on goaltending that’s been around for something like a decade and has a very limited utility. They’re pleased with their new ability to more accurately bin their data while the rest of us want, at a minimum, to have that increase in shot detail lead to better, more significant, more predictive expected goals models, or even just a better version of something like this:

I saw that graphic on Sportsnet during the Washington - Toronto game. They popped this up on the screen after Alex Ovechkin had scored from the right dot, and nothing emphasizes how unusual that shot was like a good heat map of his power play shot locations. I only discovered after the fact that they’d actually used a HockeyViz graph as the base of their image because that’s a teeny-tiny font on the credit line, there, Sportsnet.

This was a great way to use the visualization of data to back up the commentary around actual game events with facts of significance. The graphic illuminated the ideas they were conveying in a meaningful way.

This kind of analysis tool is all over the internet, and the official NHL tracking data presentation on TV is going to look simultaneously very slickly presented and very hokey to those of us used to looking at meaningful analysis tools.

The NHL isn’t just modernizing around tracking data either. They also announced a new tool they’re giving the coaching staff to provide them with statistical tools and visualizations.

NHL.com

Ah, the homeplate area marked out to show exactly how the data is binned. Watching the NHL go hard on this oldest form of measuring shot quality is like watching someone figure out the internal combustion engine just as everyone else is going electric.

The shot plot above, plus some other in-game data (all from the public gamesheets produced by the NHL for now) will be available to the coaching staff on an app they can access via their iPads.

“It’s all about the coaches, and it’s really the assistant coaches who are giving information to the head coach,” NHL director of digital business development Chris Foster said. “Specifically, it’s the coaches on the headset talking to the video room who are then giving the information to the head coach.”

Foster said the NHL, in consultation with video coaches from various teams through workshops last season and at the 2018 NHL Draft, built the app to have player usage (time on ice) and face-offs at the forefront because that information would be most valuable to coaches during games.

On the app, face-offs are broken down by success rate per zone and side of the ice, against specific players and a color-coded visual graphic for how a player has done in his past five face-offs.

For time on ice, there is a customizable threshold coaches can input for each player, essentially the amount of minutes he is expected to play in that particular game. If a player goes over the expected minutes, the graphic turns red so coaches can quickly see that information.

NHL.com

This app has the kind of slick presentation that comes from spending money on the form of the software while having access to licensed images. It also seems extremely basic, years out of date, and not much more useful than the tools these coaches have already had access to with an internet connection.

The Canadian junior leagues have this live-updating plot on their official websites:

Zachary Bouthillier is the goalie for the blue team here.

The Liiga in Finland has a version that is customizable to show the shots of particular players. Myriad private websites have been producing shot plots like this for years. Moneypuck’s visualizes their expected goals model but is based on the original NHL shot data. You’ve seen Natural Stat Trick’s in our recaps all the time. The NHL’s first foray into this commonplace visualization in private, exclusive-to-teams content they won’t provide to the public is less than what fans can already access easily.

The NHL is getting the bugs out of their new data delivery system with not very important information before they bring their tracking data online next season.

So far, the kinds of data the NHL is presenting to the public are less significant or likely to make you any the wiser than that one graphic Sportsnet lifted from HockeyViz. I assume the cheque is in the mail, there, Sportsnet, correct?

Please check out the Patreon link for HockeyViz if you liked that graphic on television and want more.

It’s impossible to guess if the NHL will present the same statistical trivia to their own teams that they want to saturate the public with or if it will be something better. And it’s up to teams to decide how to use what they’re given. Some teams have a full compliment of staff who can tell the coaches to use that iPad full of faceoff data for a coaster, and some might think it’s an amazing innovation.

There’s a lot of desire from fans and analysts to get access to the tracking data, and there’s a general belief it will revolutionize the non-NHL-sanctioned world of statistical analysis. It might, but first it has to be public, and then we’ll find out in time if the refinements make significant improvements to existing models or just open the door to more and more bar charts of rankings on various metrics without any more meaning behind them than what the NHL itself gives you now.

Is more data always better? That depends on what you do with it. So far, the NHL’s data delivery and analysis is all very much form over function. All it will take is one television broadcaster to go big on their own data analysis and presentation to blow the NHL’s initial efforts right out of the water.

Maybe I’m just holding on to a faint hope that someone will do that. How you measure and talk about hockey doesn’t really matter much in the overall scheme of things. But I can’t help but feel that what the NHL is doing — saturating their fans with insignificant statistical trivia for storytelling purposes, and giving their coaches the least meaningful information they currently produce — isn’t making people better at living in a data-saturated world. They’re just using it for set decoration.

I guess it’s not the NHLs job to make their fans smarter.