Player tracking data is next step in NFL's analytics revolution

Eagles general manager Howie Roseman put together the roster that won Super Bowl LII. Matthew Emmons/USA TODAY Sports

It's 2021, and the NFL is on the eve of free agency. A general manager in need of coverage help is watching tape, toggling between video of two cornerbacks he's considering pursuing, and comparing them.

The reel he's watching wasn't pieced together by his coaching staff, but rather by a tool on his laptop. With just three taps on his mouse pad, the GM selects man coverage, in-cutting routes and then common receivers. And just like that, his custom film playlist is set.

The GM isn't only working off the tape, though. He also holds numbers -- new numbers -- on the cornerbacks' coverage skills. Specifically, the percentage of routes receivers were open against each in man coverage over the past season. And better yet: the chance of a completion, the expected yardage and the expected points of every route the cornerbacks covered, regardless of where the opposing quarterback actually threw the ball.

It all comes from a computer that, for lack of a better term, knows football, after "watching" every moment and every movement over the past several years.

Player tracking data is the next analytics arms race in the NFL, and it's here.

In recent seasons teams have held player tracking information on its own players, but never opponents. Earlier this offseason that changed, and the NFL is about to embark on its first season and full offseason of clubs having the full league's worth of player tracking information. And at least some believe that the edge gained by those that embrace the data will be significant.

"Make no mistake, it is going to be a separator in terms of your competitiveness, both in personnel and on Sundays," said one NFL executive, who requested anonymity before discussing how use of the new data would affect teams' strategy. "My belief is it will drastically help teams compete if they can embrace it and integrate it. [And] I think it will be more of a separator early."

The information, which the league refers to as its Next Gen Stats, is collected via RFID tracking devices in every player's shoulder pads. Many believe the player tracking data will eventually reveal significant insights that can impact teams' strategies in the offseason and on game day, the way it has in the NBA.

Next Gen's introduction to fans has begun with integration into broadcasts on TV and on NFL.com, but much of that has been focused on speed, acceleration and context-less separation. The long-term benefits of the system can reveal much more actionable information for teams, once mathematicians have been sprung on the colossal quantity of coordinates over time. Machine-learning models will eventually yield information like predicted completion rates, double-team percentages, catches over expectation, and more. Each team may soon have a vast database of play types that can produce the average yardage of a flood route concept against the Cover 3.

"Scheme, and being able to predict the results from scheme," the executive said when asked where the biggest impact will eventually come from. He added that the data could allow for significant improvement in quantitative evaluation of individuals and could shine a light on individual matchup advantages, as well. Additionally, there is a health and injury-prevention aspect that could be valuable to clubs as they can track players' workloads, and some teams have installed the same system at their practice facilities. One analytics-focused former front office staffer posited that organizations might be able to use the measurement information to better understand aging.

"Teams have some real general ideas about what aging curves look like. But not every guy ages the same," he said.

Of course, in order for player tracking data to have a positive impact on any given team, that franchise has to have a desire to use it.

"I think there's interest, but there's always some healthy and some unhealthy skepticism about what it can do," said Dean Oliver, vice president of data science at TruMedia Networks, a sports analytics company. (Oliver is also a former ESPN employee).

But even beyond interest, successful implementation for most teams will require a willingness from upper management to invest in both the technology and people necessary to work with the data. One front office member floated the theory that the forward-thinking Philadelphia Eagles winning the Super Bowl could actually help convince hesitant teams that allocating more resources to analytics is a worthwhile endeavor.

Ultimately, approaches to player tracking data across the league will presumably vary a good bit from team to team. Based on an informal survey of several people in and around the league associated with analytics, a few teams are expected to be among the aggressive in their pursuit of useful information from player tracking data: the Eagles, 49ers, Patriots, Ravens, Falcons, Browns and Dolphins. One AFC personnel man said he expected "most teams" to be aggressive in trying to use the new data source.

But even among the more quant-friendly organizations, building up to a scenario in which Next Gen Stats can lead to truly advanced analysis would almost certainly take time and incremental steps. For a particular team, those steps might look something like this:

1. Some automation of tagging and filtering tape based on basic information, like which positions and players are on the field, may be possible early on. As far as insights are concerned (and putting aside workload monitoring, which some teams have been doing for years already), measurement information in the data could be useful for evaluating individuals. "What is a defensive end's burst off the line of scrimmage? How effectively can a linebacker not only drop into coverage but also change direction? Those types of things you'll be able to evaluate in a much more objective and quick fashion," the executive said.

2. Teams will build machine-learning models that will be able to classify aspects of football like route combinations, blocking assignments or coverage type. People can currently do this sort of task, but with player tracking the information should be perfectly objective, consistent and available in far fewer human hours. This should lead to more advanced tape automation.

3. With further research, those classifications will lead to new metrics and tools, like catches above expectation or a decision evaluator for quarterbacks.

4. Finally, teams may use all the aforementioned information to build predictive models -- anticipating playcalls or forecasting player performance given a type of play -- though admittedly right now it is hard to even fathom what teams' main uses for the information will be once their capabilities have reached this stage.

In other words, teams have years of analytics work ahead of them in this realm. "It's going to be a real challenge to get this stuff integrated into ... the traditional system," the executive said.

Presumably in part because of Next Gen Stats, the league office hired Michael Lopez to the new role of director of data and analytics.

"We're a resource for teams, and teams are going to be using it. So there needs to be a level of familiarity in the league office," said Lopez, a former statistics professor at Skidmore College.

Third-party companies will play a big role in some teams' process of turning player tracking data into actionable information.

TruMedia (also an ESPN partner) has signed on a double-digit number of NFL teams to work with player tracking data, Oliver said. Some of those teams had previously worked with the company, which is owned by Jaguars executive Tony Khan, when they had access to only their own players' tracking information.

Telemetry Sports now has 11 teams as clients for its player tracking tool, per co-founder Jeremy Hochstedler, noting that most came aboard this year once all of the player tracking data became available. Hochstedler revealed one metric the company has developed: Tackle Expectation, which determines each defender's chance of recording a tackle on a run play based on pre-snap positioning and formation and the intended gap of the run. Teams could then determine how many tackles above or below expectation a defender made in a given period of time.

Machine-learning company Second Spectrum (another ESPN partner) is also working with multiple NFL teams on player-tracking data, according to its CEO, Rajiv Maheswaran.

Second Spectrum is known most for its handling of NBA player tracking analysis. The company's models can identify pick-and-rolls and off-ball screens largely without any human intervention, and that information is distributed via its Eagle Tool, which it licensed to 26 NBA franchises last season.

Both Oliver and Maheswaran believe teams that move quickly to build an infrastructure and integrate the new data into their decision-making process will build a long-term edge.

"The real thing that will happen is that while there will be an impact in Year 1, the teams who start adapting early will have like 10 times more value in Year 3 and 4 than other people who are slower," Maheswaran said.

Ben Alamar and Brian Burke contributed to this story.