Chasing Perfection: A Behind-the-Scenes Look at the High-Stakes Game of Creating an NBA Champion

by Andy Glockner

  All this said, bargaining issues will work themselves out. In Berger’s column, NBPA executive director Michele Roberts said, “To the extent that the team and the player could come up with a better advantage based on the information, who’s got a problem with that?” The technologies also will continue to improve, and there will be a huge amount of value (and money) for the teams (and companies) that figure out the best ways to tie all of this data together into a comprehensive monitoring plan for players.

  A key figure in all of this is Kopp, who built up SportVU before joining Catapult in 2013. Early in his tenure at STATS, he hired Paul Robbins as the company’s director of elite performance. According to Kopp, it was Robbins who came to him early on during the SportVU experiment and convinced him that SportVU’s on-court data would be the primary source of load-measuring during games, and that that feature would be a big selling point to potential clients.

  “What he basically told me was, ‘I do all of the stuff in training,’—so kind of like P3 and Marcus Elliott. You mentioned those guys, they do amazing work in a lab,” Kopp said. “And there’s other things, like Catapult and others that you can do in practice, but the missing piece was anything in games. You do all the stuff in labs, you do all the stuff in practice, [then] you go to a game, and there’s nothing. So he convinced me this could be the game part of that equation, but as part of that, not only do we want to create things like speed and distance, but being able to approximate player load using x, y data. [That] isn’t true player load, but in a sense, the way we sold it was, ‘Well, currently there’s nothing. It’s better than nothing,’ which isn’t the best sales pitch, but it worked there for a while.”

  Kopp understands the current critiques about the usefulness of Catapult’s data. He hopes the players’ union will relent in the next series of labor negotiations because, unlike sleep monitoring devices and blood samples, Kopp reiterates that Catapult devices are to be used “in a workplace setting.” He then used an example from a sports league in Catapult’s home country to make a point that significant advancements could come to the NBA once the devices are approved for game use.

  “The best example we have—it’s really the only example in the world that you can point to—is what happens in Australia with Aussie Rules football,” Kopp said. “They use these devices in games . . . on the sidelines, and they actually—it’s kind of like hockey with its dynamic substitutions—they make substitution patterns based on what they see in the data. And they use it in practice in terms of monitoring guys and how hard they push.

  “And they also use it in the Combine for guys that are coming up through the system, and what they found is, it’s actually eliminated injuries in the Combine—because they used to have all these injury problems with guys coming to the Combine and pushing themselves too hard and they get hurt.

  “It eliminated those injuries and they use it in a way to create that baseline for players. And it’s all viewed as a positive. It’s not viewed as a negative. It’s not viewed as something where we’re going to make a decision based on, ‘Oh, hey, you’re not trying hard enough.’ The fact of that matter is, can you use it to say someone’s not trying hard enough? I guess, but is that a bad thing?”

  In the interim, SportVU data remains the most useful on-court load tracking data available, and Robbins is a key figure in its interpretation. As of early 2015, he said he was working as a consultant to seventeen different NBA teams to help them review performance data to aid their maintenance programs. According to Robbins, five NBA teams were already working with a Canadian company called Kinduct, which provides a technology platform designed to tie all of these different data sources together, but the quantity and preciseness of data simply isn’t there yet to properly piece everything together.

  Robbins, who comes from an exercise science background and was a strength coach for many years, works directly with the strength coaches and trainers for the teams that engage STATS for this service, which is an interesting intersection with the on-court analytic data provided by SportVU. He provides weekly reports to his clients that help them consider issues regarding their players’ physical workloads.

  “I will start by giving some predictions on basically what kind of workload a player might have this upcoming week, and that’s based off of how did that guy play against whom he’s matched up coming up this week, so [at first] I’m using last year’s data just to give me a feel of what a player might be going through,” Robbins said. “Later in the season, I’ll be able to use more of this year’s data, obviously, in comparing players.

  “So I’m looking at predictions. What happens to a player, who’s covering Kobe Bryant, how much load is that player getting, what does Kobe typically do to a player? And I’ll do that for all the teams, and basically the top five, six players on the teams. So I’ll give them prediction of the load—that’s the first thing I’ll do for everybody in the week.

  “And the second thing, I’ll review last week or the trends that are going on in players right now. If it looks like their intensities are dropping—it could be because of fatigue, a minor injury, it could be just the matchup he’s playing against—I will highlight anything I think is interesting to look into.

  “And again, somebody could come back and say, ‘Oh, it’s just a matchup, no big deal,’ or, ‘we blew those guys out by twenty points, it really doesn’t mean anything,’ but what I’m really trying to do when I send out these reports and I’ll go through each player, and anyone I’m concerned with, it’s to engage the strength coaches and the athletic trainers, and if there’s something that they’re like, ‘Well, I’m not sure why the intensity’s dropped in this guy. Paul, can you go and look deeper?’ And that’s when I go into the data. So what I’m trying to do is bring up red flags and question them, ‘What do you need me to look into?’”

  In 2015, Robbins estimates that SportVU’s on-court cameras are only providing him with maybe 30 to 40 percent of the data he would ideally need to track player loads. He assumes at some point, the league will allow on-court tracking monitors during games, which will close that data gap. Until then, there’s no issue with him providing consulting advice to more than half of the league, because Robbins says there’s a surprising amount of information sharing going on between training staffs at the team level. A lot of new team trainers and strength coaches are coming from the same placement sources, and a good number of them know each other. According to Robbins, there were two teams in the league that absolutely will not share data or any findings or techniques with other teams, but he guesses that as many as two-thirds of the strength coaches in the league are friendly enough to trade some basic tips.

  While advancements continue to be made across the board, the ability to truly know what the game-to-game loads are for players, and to track those trends against the off-court training, diet, and sleep data, would lift this discipline to an entirely new level.

  “The whole thing is managing loads for a player,” Robbins said. “Just make up some numbers here, but a player may have a load of ‘5,000’ in a week. That’s what he could do. I’ve got to manage that load. Is 3,000 of that load coming from games? Well, that means he only has 2,000 on all the other things going on. Now can they get some points back if they sleep well? Can they lose more points if they don’t sleep well? And the same thing with eating. They eat poorly, maybe it’s not 5,000, now it’s only 4,500. So I can manage that. And then the players start understanding the big picture.

  “So that’s where we’re really going. And then that will then help start connecting that with injuries because there’s a lot of injuries due to overuse. And that’s the load—if you have too high of a load, then that’s overuse, and there’s going to be injuries. So if I can manage those injuries . . . There’s other injuries that I’ll never be able to manage, but if I can manage overuse injuries? That’s the ultimate goal. But we’re nowhere near that yet.”

  Nine months later, back in Santa Barbara, P3’s busine
ss is booming. Just during the summer of 2015 alone, Elliott estimates that they did one hundred new or continuing assessments of NBA players. The company’s Twitter feed periodically posts photos of the players who do training there, and the 2015 batch included Detroit Pistons center Andre Drummond, Memphis Grizzlies wing defensive specialist Tony Allen, Charlotte Hornets big man Al Jefferson, Chicago Bulls center Joakim Noah, Atlanta Hawks center Al Horford, and many other well-known NBA veterans. The company also did a lot of pre-draft work for incoming rookies, with players like Jahlil Okafor (the No. 3 overall pick to the Philadelphia 76ers) and Emmanuel Mudiay (No. 7 to the Denver Nuggets) also training there.

  Word of mouth clearly is driving the expansion of P3’s business, but it’s not specious to also draw a connection between the burgeoning client list and the testing the company did at the 2014 NBA Draft Combine. After P3’s technicians analyzed the results of that testing, the company provided the NBA with a list of fifty-eight to sixty players in order of their likelihood of getting seriously injured. Then the 2014–15 season unfolded, and as noted earlier, a huge number of high-profile rookies suffered serious injuries. Like Mike Clark, Elliott is not at liberty to divulge exactly what risks were assigned to which players, but he did say, “We didn’t just say who was likely to be injured, but we said where they would be injured. And it ended up being very predictive.”

  The accuracy of scale and specificity of injury really spooked the NBA, which ran the 2014 exercise as a closed test, but invited Elliott and Clark back to the 2015 Combine, where Elliott says P3 examined all but about four or five of the attendees, even though they were an opt-in process and not required. And the NBA subsequently changed its stance about allowing P3 to share its prognosis projections with the players.

  “This is going to be a great project with the NBA,” Elliott said about his company’s now-annual involvement at the Combine. “There is going to be so much value that comes out of this. This is the first time a league has taken something like this on. We collect thousands of data points on each one of these guys, and we make predictions about them, and then we get to see what happens.”

  This new disclosure created a much bigger talking point than the actual test results, though, which relates back to the collective bargaining agreement and the players’ union’s stance on biometric data: Who owns the data that companies like P3 are collecting?

  The league has engaged P3 to come to the Draft Combine. The players are being tested, and it’s their bodies being diagnosed. The teams are the employers of these players and are investing significant amounts of money in them. It’s a very difficult and nuanced question that raises some of the same issues as blood-testing and sleep-tracking data, as some players will greatly be helped by early diagnosis of structural issues while others could suffer economic consequences from the leak of such data. As Elliott notes, there’s even debate as to what category of data this analysis falls under: Is it more akin to the bench press at the Draft Combine (which is released publicly) or like an MRI, which has much more distribution protection on it?

  “Originally, [the NBA] planned to give the data to the team, and they’ve come around to which is what I think is our position—well, I know it’s our position—that the data should rest with the players,” Elliott said. “The players control it, it’s their bodies, and they should have control of this information. And we’re going to encourage the players—especially those who have significant pathologies—to share the data with the teams, but there won’t be a mandate. The teams will only get this information if the players requested to share a copy with them.

  “This is about protecting the players, this rare resource that these guys are, and they didn’t have access to this before,” he added. “They just waited until their bodies broke. Now they have access and they can cut off some of these [injuries] and it’s amazing for the players. If the Players Association is going to rebel against this, it kind of undercuts the whole project.”

  As all of this develops, teams are getting increasingly interested in working more in concert with P3, which is focusing more on the collection and analysis of the data, and less on having physical training actually at their own gym. The corrective work, with the proper oversight and equipment, can occur anywhere, and as the salaries paid to individual players continue to rise rapidly thanks to the league’s new TV rights deal, teams are more and more concerned about protecting their investments to the best extent possible. Players and their agents are increasingly interested in this, as well, as tens of millions of dollars can be at stake on their side, too—even as everyone continues to try to figure out what it all means, and who exactly should know what about whom.

  “Just the idea of trying to perfect athletes in the NBA, it’s so timely right now,” Elliott said. “It’s unlike any of the other sports, and it’s changing so fast. Honestly, I think it’s the biggest sort of imperfect market that there is in basketball still, this piece that hasn’t been optimized yet. They went from having no real training culture to saying this is maybe the most important piece that we have. They’ve gone from zero to sixty in like three seconds.”

  CHAPTER 6

  The Tricky Art and Science of Turning Data into Wins

  It really starts with leadership’s appetite for that kind of information . . . the owner, the head of team operations . . . the head coach. . . . If anyone of those, particularly the head coach, is . . . not open to this, it’s useless. It really comes down to where the rubber meets the road and where the decision makers are willing to legitimately give this stuff weight.

  —Tom Penn, ESPN basketball analyst

  An hour or so before a mid-November 2014 home date with the New Orleans Pelicans, Sacramento Kings general manager Pete D’Alessandro was standing against a wall just off the tunnel that leads from the home team’s locker room to the court at Sleep Train Arena. At the time, D’Alessandro was in the early stages of his second season with the Kings after coming over from Denver, where he was an assistant to Masai Ujiri (who then became the general manager of the Toronto Raptors), and things were going unexpectedly well for his team.

  Six months earlier, the Kings had been the talk of the league for off-court reasons when they openly engaged in a crowdsourcing analysis exercise ahead of the 2014 NBA Draft, in which they held the No. 8 overall pick. The Kings were open-minded to all sorts of analysis and ideas that came from outside their management team, and entertained numerous strategies about which player to take with the pick or whether they should try to trade up or down in the first round.

  Eventually, ESPN’s Grantland site documented the team’s experiment in a short film, which showed a variety of interesting and awkward interactions among Kings officials. For basketball fans, the juiciest part of the film was the disclosure that Sacramento was very much considering Louisiana-Lafayette point guard Elfrid Payton against their eventual selection, Michigan shooting guard Nik Stauskas, but ultimately decided they needed more quality shooting to help space the floor around blossoming star center DeMarcus Cousins.

  By November, though, the Kings were making news on the court, where they were one of the early surprises of the season. They had compiled a 6–4 record against a very difficult schedule, and were coming off a home win over the defending champion San Antonio Spurs, during which Cousins had his way with all-time great Tim Duncan down the stretch. Furthermore, a widely questioned pair of summer moves—when the club allowed diminutive scoring point guard Isaiah Thomas to leave via free agency, and then signed the older, less dynamic Darren Collison for similar money—was paying early dividends. Collison was bringing much more stability and defense to the point guard position, in addition to a better understanding of Cousins’s role as the team’s star, rather than needing the ball in his hands as Thomas did to be effective.

  The day before this particular encounter, news had surfaced that another Kings’ gamble had paid off. Eleven months earlier, the team had traded for small forward Rudy Gay, whose low-efficiency offensiv
e game made him one of the analytics community’s poster children for how you didn’t want to play. Gay was on his third team in a year after moving from the Memphis Grizzlies to Toronto in a previous trade, and was more or less considered to be an albatross, on a contract that was paying him over $17 million a season. The conventional thought process was Gay was being paid far too much money for what he was producing in a league where salary caps and luxury taxes have significant effects on both team-building and roster preservation, if things are going well.

  Gay, though, was evolving his offensive game under second-year head coach Michael Malone, who had been brought in a few weeks before D’Alessandro by owner Vivek Ranadive in the summer of 2013. Malone was utilizing Gay a lot more in the post and deriving excellent early results. Gay was shooting the ball more efficiently, getting to the free throw line more than he historically did, and providing the Kings with some late-game mismatch advantages, as Gay often found himself posting up a smaller forward.

  The Kings moved quickly and got Gay to accept a three-year extension for about $40 million, which was much more appropriate market value for him, given his improved performance and the pending increase in the NBA’s national TV deals, which will see the cap increase from around $70 million for the 2015–16 season to perhaps $110 million for the 2017–18 season. Under the new financial structure for the league, Gay’s new deal approximates a deal worth $9 million or so in the 2014–15 environment, and even analytics folks are more than OK with that value.