The mood was tense and frustrated on an NBA-wide team analytics call in late July.
The call was ostensibly meant to discuss all the exciting features and workflows of the NBA’s new partnership with Hawk-Eye Innovations, a company known for its motion capture work across several sports. Quickly, though, team analytics staffers realized a big problem: Many of the building blocks of the motion tracking system previously used by the league, which teams have been relying on for everything from scouting and analysis to coaching game-plans for most of a decade, were missing entirely.
Worse yet, members of the NBA league office seemed oblivious to the issues this caused.
“It became very apparent to all of us in the league that they didn’t even understand how crucial [those features] were,” a senior source in an NBA analytics department told the Guardian. “It was like, ‘Oh, those are important to you guys?’ And unanimously, all of us were like, ‘Yes, they’re embedded into everything we do.’”
Hawk-Eye’s presence in the NBA stands poised to revolutionize the game much like it has in tennis, baseball and cricket, where in a short period of time it’s gone from a fun curiosity to a vital part of the way these sports are viewed, understood and even officiated. Many tennis tournaments, including the US Open, rely solely on Hawk-Eye’s tracking for all line calls; MLB’s much-discussed impending move to “robot umps” for balls and strikes involves Hawk-Eye in a central role. The prospective uses of this technology in a sport as dynamic as basketball are as vast as they are fascinating.
But with days to go until the NBA regular season, team analytics experts remain frustrated. Multiple team sources describe the last few months as a series of missed deadlines and delays in receiving basic, vital data with which to inform their coaching staffs and scouting departments, plus regular communication breakdowns. The NBA counters that the arrangement has been clear since the Hawk-Eye partnership was first announced, something multiple team sources dispute.
Could this game of broken telephone stain what should otherwise be an overwhelmingly positive new era of data in the NBA?
***
Hawk-Eye’s technology will be visible to players, coaches and fans alike from the jump this season. There’s no doubting its immense promise.
The system will use 14 cameras placed in standardized locations around arenas, tracking 29 points on players’ bodies (arms, legs, hands, etc) plus the ball in full 3D and near-real time. This replaces a prior system most recently run by tracking company Second Spectrum, which used six cameras in arena rafters to track player torsos in 2D, or the “dot” system, which created just a single point for each player at any given moment. From a data and visualization standpoint, this is akin to upgrading from a Super Nintendo to a VR headset.
The wealth of resulting data will be found across all corners of the game; officiating will be the most immediately public-facing.
The NBA Replay Center has already been outfitted with Hawk-Eye data feeds. This tech will play an active role in goaltending reviews in the 2023-24 season, using algorithms that track whether the ball was on its upward or downward trajectory when touched by a defender (or had contacted the backboard first). The league is also demoing both on-the-line and last-touched out-of-bounds systems, which are expected to roll out next season.
“Our eventual goal is, anything that is quantifiable, black-and-white, and just hard for human eyes to discern – we are hoping to use Hawk-Eye to help on those calls,” says Tom Ryan, the NBA’s vice-president of basketball strategy.
The league is taking a cautious approach early on. Goaltending algorithms will only be in use for coach’s challenges or last-two-minute reviews in the fourth quarter this season; expect the same sort of trial run for out-of-bounds or other call types in the future. As confidence in the system grows, though, don’t be surprised if the machines take a larger and larger role in a few years – communicating calls to on-court refs via watches or earpieces, or even creating an in-arena buzzer notification setup.
Viewers will soon see digital renderings of these calls on broadcasts, much like the graphics produced during in/out challenges during tennis events. That’s the iceberg’s tip when it comes to fan interactions.
Tennis and baseball broadcasts showcase simple live stats like serve or pitch speed; NBA fans can expect things like sprint speed, pass speed and shot arc just as a taster. Another layer deeper will be insights drawn from 3D pose data (made possible by the tracking of 29 bodily points) on everything from jump height to shooting form or defensive positioning.
Full fan immersion via virtual reality is just around the corner. Imagine watching digital recreations of a live game from the perspective of your favorite player; think of toggling a command and getting a 3D rendering of, say, Steph Curry’s average shot arc from a given game.
“It’s going to be an exciting next few years,” Ryan says.
A similar data explosion is inevitable in the world of basketball statistics. Tracking 29 bodily points instead of one creates an ocean of potential where there was previously just a large pond.
Synergy Sports, an entity well-known to NBA analytics staffers and enthusiasts alike, is under the Sportradar umbrella. Seventeen NBA teams are currently contracted for access to Synergy’s platform, per Sportradar representatives, which combines its longtime manual play tracking with Hawk-Eye’s pose data.
Virtually any basketball metric you can imagine will either be improved upon or created from scratch using 3D inputs, but some of the simplest ones showcase its value even to casual fans. Think of shot contests and “closeouts” to shooters, intuitively one of the baseline elements to consider for any team’s defense: With dot tracking, this was confined to crude approximations; how close was the defender’s torso to the shooter’s torso when the shot went up? 3D data, however, tracks things like player direction, jump height, arm extension and even hand angle – a true recreation of what’s happening on the court.
Before long, expect to see certain Hawk-Eye-derived stats up on public NBA.com archives.
Many in the analytics world also quickly envision a robust health and performance market cropping up based on 3D pose data.
“Biomechanics has been the next revolution in baseball,” says Bryan Spangler, director of engineering, player optimization and insights for Sportradar. “They’ve really looked at the pitcher’s arm angle and velocity … There will be [NBA] teams who go out and get a strategic advantage with that approach.”
That’s a ways down the line. For now, more immediate hurdles remain.
***
The NBA’s decision to bring the Hawk-Eye setup to Abu Dhabi for showcase preseason games in early October surprised many on the team analytics side due to the issues with its implementation to that point. Quickly, it just added to those concerns.
Team staffers expected a data dump following completion of games; none came. The next day, per multiple league sources, a leaguewide email informed teams that someone had turned the power off in the arena, stopping the data upload. Another update was promised soon. Full data from one of these Abu Dhabi games took roughly two weeks to become available, while data from the second remains incomplete as of this publication.
“I was willing to cut them a ton of slack on this,” said the same senior team source who described that contentious July call. “It’s a massive change, a whole new provider, it’s preseason.
“But the lack of communication is just crazy. We haven’t heard anything other than ‘Oh, the power went out.’ And then it’s crickets since then.”
At issue here are two forms of data: “Raw” inputs, which refer to x/y/z-axis location files for each of the 29 bodily points tracked; and “markings” files, or various basketball plays and events extracted from raw inputs – dribbles, passes, screens, post-ups and other simple basketball actions. All 30 NBA teams used markings as a shorthand for their basic operations under the prior “dot” tracking setup, helping them track and easily recall vital needs like coverages and game plans; some augmented that with their own in-house work using raw 2D data. Second Spectrum provided both these feeds in the old arrangement; Hawk-Eye, though, is exclusively a data capture company – they provide raw data only, leaving further insights to other entities.
NBA representatives maintain that the plan all along was for Second Spectrum to continue providing 2D markings files, allowing every team the exact same baseline data they had in prior seasons. They point to an expanded broadcast and team analytics partnership with Second Spectrum that was announced on the same day as the league’s move to Hawk-Eye.
Multiple team sources dispute this, saying the team side was a distant second priority behind broadcast initiatives. They say a markings provider was never made clear, and that their own pressure on the NBA – including that July call – forced the league’s hand to restore that role to Second Spectrum.
“At the time we realized, ‘Whoa, this gives Second Spectrum a lot of leverage,’” the senior team source says. “They’re the only game in town.”
Recognizing this, sources say, Second Spectrum raised prices for many teams – forcing the NBA to step in and cap potential increases. The NBA disputes this, again saying this two-pronged approach was always part of the plan.
“Ensuring minimal disruption to the team analytics community is a significant focus area for the league,” Ryan says. “We’ve made a ton of progress over the last several months as we’re going through the transition from Second Spectrum capturing games to now Hawk-Eye. And we fully expect there to be full continuity of team analytics workflows and the way they think about coaching analytics, player analytics. It’s going to be using the exact same data they had available to them, in the ‘22-23 season.”
Representatives from Second Spectrum could not be reached for this story.
Wherever the truth lies, things have been rocky even since the summer. Analytics staffers went weeks during preseason without data from their team’s games; data feeds mostly caught up with under 10 days to go until the regular season, sources say, but data for many preseason games remains flawed or incomplete. Team staffers are worried the same issues will persist when meaningful games begin next week.
Frustration levels range across the league. Some teams with entrenched coaching staffs or analytics teams are especially upset at being forced to uproot processes they’ve built for years; others with newer staffs and infrastructures are less concerned. Even among those most heavily impacted, some simply accept headaches like these as part of the job.
Bumps in the road are no surprise for such a cutting-edge new system. Some will surely even out with time. Certain league and team staffers even see the separation of raw and markings data as a long-term positive – new entities could enter the markings space over time, increasing competition and bolstering the end product.
And as 3D pose data moves out of its infancy and is understood at higher levels in upcoming years, we’ll see the best separate from the rest in ways never before possible. Teams with deeper-pocketed ownership will have more resources to invest in plumbing the true depths of 3D inputs; truly elite analysts will be in even higher demand.
“There are definitely going to be some teams that get a huge edge over others when using this data,” the senior team analytics source said. “That’s what we’ve seen happen in baseball. Teams fell behind because they were valuing the wrong things.”
The future of data in the NBA is tantalizing no matter how you look at it – provided teams and the league can play nice with each other long enough to get there.
