这篇文章来自 wired.com。原始 url 是: https://www.wired.com/story/how-data-saved-the-golden-state-warriors/
以下内容由机器翻译生成。如果您觉得可读性不好, 请阅读原文或 点击这里.
“Well, you’re going to have to look down if you want to see the camera.”
Travis Schlenk didn’t give off the impression of someone you’d care to defy. Beefy and bald, with a penchant for wearing cowboy boots instead of designer shoes, Schlenk was wrapping up his seventh season with the Warriors when we met on April 13, 2011. It was his second year as the team’s director of player personnel. Two months later, he would be promoted to assistant general manager. It was an impressive rise for the Selden, Kansas, native who caught on with the Golden State Warriors after working four years with the Miami Heat as their video coordinator. He then paid his dues as a video scout and assistant coach with the Warriors. Schlenk is, perhaps due to his midwestern upbringing, reserved in conversation—“one of those guys who flies under the radar,” according to Jerry West—and that reticence served him well in becoming the Warriors’ resident stats nerd in those first few months under the Joe Lacob regime. Someone who’s ostensibly in charge of your team’s nascent analytics buildup is not someone who should be gabbing to the press when it’s not his place.
从 BETABALL: How Silicon Valley and Science Built One of the Greatest Basketball Teams in History by Erik Malinowski. Copyright © 2017 by Erik Malinowski. Reprinted by permission of ATRIA BOOKS, a Division of Simon & Schuster, Inc.
But Schlenk was doing me a service this day, as we walked along the catwalks suspended some 80 feet above the Oracle Arena flooring. A few hours before the Warriors’ final regular-season home game of the 2010–11 season, a dance team down below was squeezing in last-minute practice before the 7:30 p.m. tipoff against Portland. The Trail Blazers, sitting on 48 wins, were looking ahead to the playoffs, where they’d face the Dallas Mavericks, while the 35-win Warriors were looking ahead to a summer of golf and general uncertainty about what changes might befall the franchise.
For now, Schlenk was willing to show me what I’d asked to see. As Lacob prepared to head off to Boston for the MIT Sloan Sports Analytics Conference several weeks earlier, I’d read a news story that the Warriors had installed motion-capture cameras inside Oracle Arena as part of a data-gathering effort called SportVU. The hope was that this budding technology might usher in a movement that would revolutionize how basketball is analyzed, and how the game is taught to the next generation.
For some, this represented nothing less than the future of the NBA.
Despite its name, SportVU didn’t originally develop from anything sports-related. The firm was founded in 2005 by Miky Tamir, an Israeli entrepreneur who earned a PhD in physics and worked for the Soreq Nuclear Research Center in Israel for 10 years. His specialty was advanced optical recognition and image processing, and Tamir founded the company based on missile-tracking technology he had developed. And in the way you can track a missile based on motion capture, you can also track a player or a ball during a sporting event, and that seemed like a more pro table and practical application.
SportVU started analyzing soccer games by recording ball and player motion more than a dozen times per second. The footage would be piped through SportVU’s proprietary algorithms to plot the data on a three-dimensional grid. What resulted were not only custom animations and tailored scouting reports, but also virtual reams of spreadsheet data that could be further parsed by the teams themselves, provided you had a stable of quants who knew what the hell to do with all the information.
Most American audiences probably learned about SportVU for the first time during the 2008 US presidential election, when it partnered with a graphics firm named Vizrt to provide CNN with its Jedi-like holograms of anchor Jessica Yellin during the network’s coverage. A month after Barack Obama was elected president and the need for political holograms had dissipated, SportVU was acquired by a company with three decades of expertise in sports and statistics. Stats LLC, which was jointly owned by the Associated Press and Rupert Murdoch’s News Corporation, bought out Tamir and other minority investors. For a year, the Chicago-based Stats had been looking to move into motion-capture technology if it could acquire the right company.
That company was SportVU.
MORE OF THE TOP TECH BOOKS OF 2017
The first task was deciding which sport to target. Baseball was cornered by a Silicon Valley company named Sportvision, better known for its “glowing” hockey puck, the virtual first-down line on NFL telecasts, and (more pertinent to Stats’ interests) the location and speed for every pitch thrown in every Major League Baseball game. Football was an option, but the sport boasted too many variables. A single play could evolve in an endless number of directions, affected by myriad circumstances that would have to be considered ahead of time. As nice as it would be to gain a foothold in a market built upon an $8.5 billion business, the NFL was out.
But the NBA, then a $4 billion industry on the rise, presented a host of opportunities. The sport was fluid but defined. (A round ball? Five players on each team? Compared to other sports, that was analytical heaven.) The court size was also manageable from a data standpoint—only 4,700 square feet, compared to an NFL field that encompasses 57,600 square feet, end zones included. And there were some teams already in line to come aboard as early adopters. As the 2010–11 season approached, the task fell to Brian Kopp, Stats’ vice president of strategy and development, to get at least a handful of teams to commit. To do that, he’d have to first convince them of the data’s utility. “You don’t want to come in and say, ‘Look at all this cool technology. Look at all this data it can spit out,’” Kopp told me. “People think that’s cool, but teams had one of two reactions. One is, ‘That’s great, but I don’t understand what it means.’ The other was, ‘What am I going to do with that?’ ”
Four teams initially signed on for the 2010–11 season as SportVU guinea pigs: the Dallas Mavericks, Houston Rockets, Oklahoma City Thunder, and San Antonio Spurs. For a new tech-centric movement to hit the NBA, this was as perfect a starting foursome as you could orchestrate. San Antonio had quietly maintained a modern-day dynasty by eschewing the usual dictums of modern NBA strategy and roster development; this seemed the next logical step. Dallas was owned by Mark Cuban, a dot-com billionaire who was an active investor in analytics companies. Houston was led by general manager Daryl Morey, the MIT grad and Sloan conference cofounder. And Oklahoma City was being molded into a perennial contender by general manager Sam Presti, who started out as a $250-a-month intern with San Antonio before getting hired by Seattle in 2007—at the ripe age of 30—as the second-youngest GM in league history. These were the first four teams to embrace a style of tech they hoped could take their organizations to new heights.
The Warriors, with nothing to lose, were the fifth.
SportVU first came on their radar in mid-October, with opening night for the 2010–11 season still 11 days away. Pat Sund, Golden State’s basketball operations coordinator and son of then–Atlanta Hawks general manager Rick Sund, was at the (deep breath now) Northern California Symposium on Statistics and Operations Research in Sports, held at Menlo College in the affluent Silicon Valley suburb of Atherton.
While soaking up all the sports geekery one could consume, Sund bumped into Dean Oliver, a former director of quantitative analysis for the Denver Nuggets. Oliver, who, in 2002, literally wrote the book on basketball analytics (Basketball on Paper) and who would later become ESPN’s first-ever director of production analytics, talked up the SportVU system to Sund, who then took that info back to Schlenk and Kirk Lacob. Within weeks, the Warriors’ staff agreed to move forward on the tech and contacted Stats. A technician was flown in from Chicago to assess Oracle Arena for a potential fitting. It would be SportVU’s first placement on the West Coast.
By the time the SportVU cameras were fully operational in mid-January and ready to feed game data back to the Warriors’ front office, the team was already 15-23 and scuffing mightily under the direction of new coach Keith Smart. After starting the season with six wins over their first eight games, the Warriors cratered, losing 16 of their following 19 games.
But on January 14, 2011, when the Warriors smacked the Clippers, 122–112, thanks to a barrage of 14 threes on 26 attempts, it was with six SportVU cameras fully operational and preparing to feed in-game data to the front office in Oakland as soon as possible. For the rest of the season, with every home game captured by SportVU’s eyes in the sky, the Warriors won 20 games and lost 23. Even Curry’s shooting percentage went up a not-insignificant three percentage points. Schlenk told me then that Smart and his coaching staff weren’t doing a whole lot with the data coming in. at was to be expected when you combine a first-year head coach and a technology still in its infancy. But the Warriors were finally building a foundation of data that could be used to their advantage down the line.
And SportVU was capturing the Warriors at the right time, just as they started playing their best basketball in years.