This upcoming season, tracking of player movement will be available in the stadiums of all 32 NFL teams, the site of the Pro Bowl in Orlando and in international arenas where NFL games are played in London and Mexico City. With all of this new data available, NFL teams will begin the mad dash to gain any sort of upper hand or slight advantage. The question is: Are all of these new stats the future of sports analysis or just a red herring?
In a sense, there are very few areas that shouldn’t be explored when it comes to advanced analytics and statistics. Teams have budgets well into the millions, so hiring an analytics department isn’t going to ravage a front office’s bottom line. It’s foolish to write off any sort of scientific progress in the name of tradition, so any organization choosing to remain steadfast in its approach is on a path to finding itself with a lot of ground to make up. This is especially true in such a performance-incentive-heavy field as sports.
So it was hardly a shock when the league announced the tracking partnership with Zebra Technologies, a multi-billion dollar international company. The league will be in possession of proprietary data that initially wouldn’t be available anywhere else. But instead of letting teams toil over the implementation and keep it close to the vest, they’re basically crowd sourcing anyone with an interest in numbers.
It was just a few years ago that the league started to get serious about player tracking. Eric Petrosinelli, the general manager of Zebra Sports, says the process began with an exhaustive study by the league.
“[It was] sort of what I’d consider to be a bake-off of different technologies and vendors back in the 2013 season, testing different technologies, ” Petrosinelli said.
In addition to its technology, the league was looking at infrared and global position systems solutions, but ultimately chose Zebra as the exclusive provider. Over two seasons, they began fully implementing their technology in all stadiums.
“Really, our focus is providing operation visibility to assets that people normally wouldn’t have sightlines to, and it’s allowing them to analyze that data to make better and faster and smarter decisions,” Petrosinelli said.
To achieve that focus, Zebra embeds tags in the shoulder pads of players that then transmit a signal to 25 receivers in the stadiums.
“Then we process that location information and translate that information into what I would consider to be advanced statistics: speed, distances traveled, accelerations, etc.,” Petrosinelli said. “So what we typically do, we gather all that data and provide it to the NFL, and then the NFL distributes.”
All of the technology and processes seem arduous for the average, non-tech whiz fan. After all, it’s a game and not a master’s class in computer engineering. With all of this data at its disposal, the league now has to get fans to care, and it’s up to the individual team to solve that conundrum. For some teams, the focus is improving the fan experience, according to Petrosinelli.
“You may see that data across different implementations, in broadcasts, digital mobile, gaming, fantasy or in stadiums,” he explained.
Microsoft, for example, will start showing these next-generation stats on their NFL app available on XBox One and Windows 10. So fans could theoretically watch Odell Beckham Jr. bust down the sideline and beat Josh Norman cleanly in a foot race to snag the ball, then immediately know how much faster Beckham was running than Norman. Or after Tavon Austin takes a punt return 76 yards for a score, using sideline-to-sideline, fleet-footed dekes, you can see the overall distance traveled in real time, not just the yards gained on the play.
“The data that allows that product to exist is the next generation stats that come from the Zebra system that’s deployed at NFL stadiums,” Petrosinelli said.
Fans are everything in the NFL, but the on-field product matters more than the packaging. Knowing a player’s top speed is nice information for a team to have, but if a player is slogging up and down the field, and now there’s additional evidence of that lack of hustle, it could get embarrassing for the player. So it’s up to the teams to find a way to best use all this new data.
Petrosinelli said that teams are using it from a strategy, tactics and performance standpoint.
“This year, they receive it on the field and the next day after the game, then they use that information, whether it’s for coaching, scouting, or from a performance standpoint with players,” he said.
Specifically relating to player performance, Petrosinelli said, training staffs and strength and condition coaches are really using the data to create plans aimed at keeping players on the field for all 16 games.
“They’re looking at the data to see, in essence, cumulatively the workload and the stress that the players are exhibiting on game day and what their performance numbers are,” Petrosinelli said. “Let’s take for example, a specific player: If it’s game day, then [the team is] knowing how many yards he is actually running in the game, not just the yards that are shown up on a traditional stat sheet.”
So again, take a receiver such as Antonio Brown. On a typical day, he may finish a game with 5 catches for 60 yards and a touchdown. Those are the stats you’d see in any box score. But throughout the course of a game, he’s run exponentially more than that. He runs a route or throws a block on every play, so that’s a high-intensity strain on the legs, whether or not he catches the ball or is even targeted.
Or look at a cornerback who is covering a real burner, someone running go route after go route. Until now, teams couldn’t measure how much that cornerback was running in a game. With Zebra, that will change.
Petrosinelli said that Zebra will be tracking not just that distance traveled and the workload, but also will be able to tell when the intensity turns up. For example,when a player is sprinting instead of just running. With all of that data, teams can then tailor a specific recovery plan for players and measure how much work they should get during practice to keep their bodies in peak shape.
“All with the goal of having optimal performance on game day, and with the goal of reducing the level of injuries that could occur over the course of a season,” Petrosinelli said.
He added, “When we’re talking about the health and wellness of a player … we’re really talking about their health and wellness in the sense of, ‘How do I train an athlete for a full NFL season and during the NFL season; how do I look at the acute – like in-week attributes of performance – as well as chronic, which is across all weeks?’”
A confluence of new technology and statistics can also be a bit dangerous. There’s so much data to parse that often teams or analysts can assign too much value to the wrong thing or maybe a number that doesn’t mean anything at all.
If a cornerback is sprinting more in a game, does that mean he cares more and is engaged more and trying harder? Or is it simply that he just isn’t as talented and has to put more strain on his legs to keep up with a wide receiver or stay involved in the play?
Petrosinelli says that Zebra’s background as a leader in enterprise asset intelligence solutions can help teams navigate that murky water by packing the stats that they believe are integral and getting them to the right people expeditiously.
“The challenge in bringing operational visibility is, how do you provide information in a very succinct and timely manner such that people can make better, faster and smarter decisions?” he said. “Inundating people with data doesn’t really accomplish the task; probably, it would frustrate them, especially in the vertical of sports.”
He added, “That’s where we’ve worked to refine and understand exactly what are the key pieces of information coaches, scouts and trainers need and get them information in the quickest manner and in a simple format.”
One of the people traversing the peaks and pitfalls of so-called “data-dumps” is analytics expert Keith Goldner. As a professional, Goldner is the chief analyst of a sports analytics company called numberFire. He’s done consulting work for three NBA teams and has written extensively about football. He has also presented at the MIT Sloan Analytics Sports Conference alongside some of the top analytics minds in the country. Speaking to player tracking as a general concept, Goldner sees an immediate benefit to teams.
“I would think the lowest hanging fruit would be using machine learning to automatically recognize and identify things like plays and routes – essentially automating the role of a quality control coach,” he said. “The next step would be to combine those plays with the higher level advanced stats to build efficiency reports on top of the frequency data.”
With a lot of the numbers freshly being made available, Goldner said he can’t tell at this point whether there will be any value in things like speed, contact point or total yards traveled, but there are certainly things that will be predictive about how players move. The angles of their routes, distance and separation between a defender and closing speed are all things that seemingly add value from a scouting perspective, but without having analyzed the specific data from Zebra, it’s too early to say how much.
For members of the media trying to use analytics in their analysis, building a predictive model can be extremely useful, and the same could be said for the analytics departments of teams. Goldner says in that case, the data needs to be actually predicative, which isn’t necessarily the case here.
“If you’re trying to make something predictive, it needs to actually be predictive,” he said. “The example of ‘distance ran’ in soccer is one statistic that, while interesting, is not particularly useful for teams.”
Despite the potential pitfalls of maybe placing too much emphasis on an interesting new stat that might not actually be useful, Goldner added that any new data is always a plus. But it’s a double-edged sword, because with that new data comes new responsibility.
“More accurate data is always a positive thing from an analytics perspective,” he said, “but the toughest part is always knowing what questions to ask and how to approach the data.”