How Hockey Is Embracing Big Data and Analytics

July 23, 2019 | 5 minute read
Michael Chen
Senior Manager
Text Size 100%:

The idea of applying analytics in professional sports has been around since the term sabermetrics came about for baseball in the early 1970s. This hit the mainstream with the phenomenon of Moneyball—the technique as used by the Oakland A’s and the ensuing book and film. However, all of the actual statistical tracking involved came from manual tracking, which made sense for a sport like baseball, given its rhythmic pace and defined start/stop points.

Today, professional soccer and basketball have started to use analytics, particularly as data tracking technology has moved things from manual entry to automated data. But when people discuss analytics in hockey, the potential seems limited at first glance—after all, players skate at speeds up to 35 MPH and the puck flies up to 100 MPH. To the uninitiated, the game can appear to be too chaotic to quantify. But while teams began applying new standards of advanced statistics in the past decade, the convergence of technology and sports is about to create the big data era for the NHL. The result is a revolution in the potential of sports analytics, and it’s going live for the 2019-20 NHL season.

How It Works

For the past decade, the NHL has employed a team of statisticians utilizing a tool called the Hockey Information Tracking System to log things like a player’s time on ice, face-off wins, and other stats. The key word here is team—statistics were tracked manually on a proprietary data entry application by multiple people.

Automated data tracking technology, though, has been in development for several years and first publicly deployed during the 2019 NHL All-Star Game in San Jose. Specially manufactured pucks contain tracking chips, and similar chips are placed into player shoulder pads. Combined with a network of sensors placed around the rink, suddenly the world’s fastest sport has big data being collected during every moment of the game: position, speed, direction, distance, and other key metrics.

How many metrics comprise hockey’s version of big data? Consider that the chips involved deliver data points at about 200 times per second, all across the categories stated above. Now combine that with fiver skaters and a goalie per each team and 60 minutes per game, and that’s a lot of data points, especially because skater data must be consolidated between who’s currently on the ice and who’s on the bench. In fact, crunching the numbers shows that 60 minutes of gameplay will create roughly 9,360,000 logged events per category in an NHL game that ends in regulation. That’s some pretty big data, and that’s even before it’s parsed into standard delineations for individual player shifts (e.g. consolidated numbers for total distance skated).

Collecting this raw data is an achievement in itself, but the NHL is leaning into the potential of emerging technology by processing this data through artificial intelligence. The AI is designed to flag specific hockey plays as it crunches through the data; for example, if a goal is scored on a 2-on-1 rush or if a goalie was out of position. As with all AI models, the more data processed and the more plays and details that can be flagged—all without human error and bias built into manually tracking such a fast sport—the more robust the model.

A New Level of Hockey Analytics

The first analytics revolution in hockey sought to provide context to traditional statistics such as goals and assists. This was done by applying manually tracked statistics such as shots on goal and face-off location in different ways, all in an effort to create deeper insight into player and team effectiveness. With completely new categories of statistics tracked in real time, the upcoming analytics revolution is poised to shatter what was previously possible. Now data-driven insights can quantify—or disprove—assumptions about players, strategies, and other coaching techniques. For example, players will naturally fatigue over the course of three periods, but how much does this affect top speed and reaction time? And is there a formula for distance skated or length of shift to maximize late-game speed? Details on this level were previously unheard of, but now they will all be part of a standard real-time data package.

That’s a lot of data points, and a lot of people to use that data in different ways. It’s still yet to be determined as to who will have access to what; the common assumption is that the NHL will follow the model set by other professional leagues in tiering the data so fans get some numbers, media gets another tier, and the league has the full complement. Who truly benefits from big data in hockey? Let’s take a closer look.

Teams: In pro sports, every single decision is about getting an advantage. We’re well past the days when roster and coaching strategies were based solely on gut feel; in sports, as with business, results can now be quantified. With heavy volumes of data, team statisticians and data scientists can do more to look at tweaking strategies, uncovering hidden free agent gems, and maximizing player usage. Companies use big data analytics for insights that power data-driven decisions, and this is the same thing: it’s all about using results to get better results.

Media: Last year’s All-Star Game provided a hint as to how a TV broadcast can be changed with real-time big data. For live game action, player statistics such as speed can be displayed during play to provide viewers with more data. Replays can be completely changed, as new information about distance, speed, and space provide context into the how and why plays happened.

Revenue Streams: The primary role of big data in hockey focuses on evaluating player and team performance, whether for internal strategy or in media. However, an indirect benefit of this data comes from new revenue streams. Licensing that data, sponsoring data milestones on broadcasts (e.g., “Tonight’s fastest skater presented by Hertz”), even seeing how data provides new gambling lines, all of that has been discussed by the NHL internally. So while there is an investment involved with implementing big data and analytics, it also provides a means to new revenue.

Fans: The more information fans get, the better. Whether that’s during a game broadcast, via an app with real-time data while sitting in an arena, or aggregated on sports websites after the fact, this data—and the ensuing embrace of new analytics models—provides fans with greater understanding of the how, why, and when of hockey. That understanding ultimately brings them closer to the game, which is what any professional sports team wants in building fan relations.

As with all technology advances, this represents just the beginning of the NHL’s big data era. In coming years, analytics models will improve, more tracking possibilities will open up, and media will have a better understanding of how this can all be optimally integrated into their presentations. Players themselves may even change, adapting training and nutrition to insights provided by big data. The scope of this revolution can be judged by the way big data has impacted other industries, such as the way manufacturing processes are now optimized for quality and procurement, or the way healthcare uses data to streamline the patient experience. For the NHL, this all starts with the first face-off of the 2019-20 season—a face-off featuring a microchip in a puck.

New to the concepts of big data and analytics? Subscribe to Oracle’s Big Data Blog or check out the following links to learn more:


Michael Chen

Senior Manager

Previous Post

Five Different Ways Businesses Use Big Data

Michael Chen | 5 min read

Next Post

How Big Data Powers the Internet of Things

Michael Chen | 5 min read