Update On The NHL Data Divide and The “Haves” and “Have-Nots”

We’ve been discussing the approaching data divide that is barreling towards the public analytics crowd for several years now. In a nut shell, NHL teams will be receiving exponentially greater amounts of data and data types from the implementation of a new puck and player tracking system. Data that will be unavailable to the general public.

In a recent interview with Sportsnet, Sam Ventura, director of hockey operations and hockey research for the Pittsburgh Penguins, was asked how big he perceived the chasm to be between publicly available data and the data that teams can get their hands on.

”It’s pretty big and it’s getting bigger. The league is going to have player-puck tracking data for the first full season in 2021 and we had it for the end of the playoffs last year. And it is a game-changer in many ways for us as an analytics department. And it definitely widens the gap between what’s out there publicly and what’s behind closed doors. It widens the gap in terms of the types of tools and methods of analysis that you’d have to use to really extract the most information out of these data sources as possible.”

Each NHL team will get a “fire hose of data.” Players will produce 200 data points a second and the puck will generate 2,000 data points a second. More than the sheer quantity of data, it’s the type of information that’s exciting analytics folks.

[From Garden Hose To Fire Hose: NHL Teams Prepare To Receive, Store and Analyze a Flood of New Data]

Ventura was asked if he would like more data from the league to be made public and accessible to street analysts.

”(Laughs) selfishly I would rather everything stay private, but I guess in the spirit of people educating themselves in terms of how to use data and statistical methods and everything, it certainly would be nice to see that same boom that has happened in other sports in terms of people doing independent sports-analytics research with tracking data, like what’s happened with football, basketball and soccer. It’d be great to see that happen in hockey as well. I’ve actually tried to stay up to date with what’s been happening, for example, in the NFL by doing research with some of my former colleagues to get some exposure to the kinds of methods that would be necessary to work with tracking data, practising with the different sport-tracking data to prepare for the current scenario, where we’re going to have player-puck tracking data for the first time.”

So What does this mean for public analysts?

In it’s simplest form, beginning this season teams will be able to generate information and make evaluations that may seem questionable to analytics folks on the streets, but because of the data divide, there will be no way to fully replicate, validate or understand what information teams are using, or why certain decisions were made.

For example, elementary advanced analytics like ‘expected goals’ or basic possession metrics (Corsi) may say that Player A has better success than Player B in a certain scenario. However, new data elements available from puck and player tracking systems (unavailable to the public) may say the exact opposite. Public analysts may be left scratching their heads by lineups or moves made by the team, and maybe even decree organizational mis-steps, not realizing they don’t have key information utilized in a decision.

Puck and player tracking data is a generational game changer. The new data points and data sets available to teams will add a third and fourth dimension to elementary, second generation stats like Corsi and Expected goals. All players on the ice, player locations, shot speeds, shot lane characteristics, shift times at time of shot, passing trajectories, player trajectories, are just a few of the many data characteristics unavailable to the street that antiquate current analytics calculations.

Healthy Steps Forward

We see it all the time, extreme extrapolations and inaccurate conclusions developed from a very thin layer of data, calling for a player to be canned from a single graphic. Ventura had very healthy advice for public analysts.

“One of the most important things you can do as an analyst is to know the limits of what you’re able to say with the data. So being careful about how you present results like that is important. I do think that understanding the limitations of what you can say with data is important when you’re communicating results to people who actually have to make the big decisions.”

In this day and age, we need to be cognizant that public analysts have less and less of the data to make an informed decision. But the league and the teams also risk losing value and insight.

Teams risk losing the ‘Wisdom Of the Crowd” that was extremely helpful during the early days and early growth of advanced analytics. Crowdsourcing analytics and insight will also become irrelevant if there aren’t measures made to include street analytics groups.

Teams will continue to establish private partnerships with universities and third party analytics companies, but values attainable through open development of analytics and apps will never fully be realized, unless public inclusion is achieved.

Hack-a-thons and other open data events with anonymized data pools would be a good first step, or making day-old bread (Data) available to the streets also a potential remedy. It’s understood that some of the new data sets related to player bio metrics should never really be made public, but the data from the game should always be made available to the public.

It’s unknown if the league or teams ever plan to include the general analytics public in future use of the next-gen data, as the league has essentially sidestepped direct questions about the use of next-gen data.

It’s up to the teams and the NHL if they want to bring along the public analytics community, or leave us on the side of the road. What’s best for all? Liberate the data.

By Jon Sorensen
Follow @nova_caps