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

by Andy Glockner

  • Bob Bellotti was the inventor of the points-created metric, an offshoot of James’s runs-created formulation for baseball, which attempted to define a player’s contributions in a single calculation. He also wrote a series of books starting in the late 1980s, with a 1988 publication called Basketball’s Hidden Game earning him initial entrance into the NBA with the Milwaukee Bucks. He consulted for Milwaukee for nineteen years before switching to the Washington Wizards when Bucks general manager Ernie Grunfeld moved to that franchise.

  • Bob Chaikin created a computer analysis/simulation program called B-Ball, which used statistics to try to suss out player impact and successful lineup combinations. Beginning in 1992, Chaikin has worked as a statistical consultant, analyst, and/or scout at various times for the (then) New Jersey Nets, Miami Heat, and Portland Trail Blazers. Since 2008, he’s been back with the Heat, where he also does college and NBA Development League statistical evaluations.

  • Jeff Sagarin and Wayne Winston are former MIT undergraduate classmates who worked together to create WINVAL, which is credited as the first attempt at an adjusted plus-minus metric for players. Sagarin is best known for his college football (and other) rankings systems that get prominent play at USA Today. Winston is a statistics professor at Indiana University who once taught Dallas Mavericks owner Mark Cuban in college. The Mavericks adopted the WINVAL metric shortly after Cuban bought the team in 2000, and Cuban credits the system for part of the Mavericks’ subsequent success in that era.

  • Dean Oliver is considered basketball’s best proxy for Bill James. The Caltech-trained statistician was one of the forefathers of the APBR movement, first breaking new ground in his writing at his own site, the Journal of Basketball Studies, in the mid-1990s, and later as the author of the seminal basketball analytics book, Basketball on Paper, which he published in 2003. The book brought significant attention to what Oliver called the four factors of basketball success: effective field goal percentage, turnover percentage, offensive rebounding percentage, and free throw rate, as well as adjusting player and team analyses for possession-based game tempo, and assigning varying fractional credit for team wins and losses to individual players.

  Oliver also was a pioneer in bringing basketball analysis to the digital world, first as a leader on the Usenet group rec.sport.basketball in the text-heavy Internet era of the mid-1990s, and then on the APBR analysis group on Yahoo! groups in the early 2000s. Oliver posted the initial message in that forum, which saw a total of twenty-seven posts that February, and the first question he talked about exploring was whether “Hack-a-Shaq,” the strategy to intentionally foul poor free throw shooters, actually worked. (Fourteen years later, this discussion became a frenzy during the Western Conference semifinal series between the Los Angeles Clippers and Houston Rockets, with both teams using the strategy so much that a rule change was discussed, but not enacted, in the summer of 2015.)

  Oliver subsequently served front-office roles with the Seattle SuperSonics (where he was the first full-time NBA analytics hire), the Denver Nuggets and, after a stint as director of production analytics at ESPN, the Sacramento Kings.

  • John Hollinger is another of the APBRmetricians who moved on to high-profile roles. After first founding a basketball writing site called AlleyOop.com in 1996, Hollinger spent three years working at The Oregonian newspaper in Portland, Oregon. In 2002, he was hired by SI.com and also started publishing Pro Basketball Prospectus, which later changed its name to Pro Basketball Forecast. In the mid-2000s, Hollinger started writing for ESPN.com, eventually moving all of his annual print publication work to ESPN’s Insider premium online product.

  He is the creator of the player efficiency rating (PER) statistical metric that attempts to quantify everything a basketball player does in one composite number, and calibrates it against the rest of the league every season. It’s considered to include an incomplete valuation of a player’s defensive contributions, but remains one of the most widely known tools, thanks to Hollinger’s writing platforms and subsequent success. In late 2012, Hollinger was hired by the Memphis Grizzlies to be their vice president of basketball operations.

  • Roland Beech launched the impactful website 82Games.com during the 2002–03 NBA season, making advanced NBA stats easily available and consumable for the public. In 2009, Beech was hired by the Dallas Mavericks as their director of basketball analytics, with a pioneering role as part of the team’s coaching staff, not the front office. Beech traveled with the team on the road, sat behind the bench during games, and liaised with cerebral and stats-friendly head coach Rick Carlisle. Beech played a crucial role in the 2011 NBA Finals, where his lineup analysis allowed Carlisle and his staff to make crucial adjustments against the Miami Heat that helped the Mavericks win the title. In August 2015, Beech moved to the Sacramento Kings to lead their analytics department, replacing Oliver, who was pushed out in the aftermath of a team management shakeup.

  As renowned as they were (and some still are), these men were not the only ones playing around with advanced basketball statistics in the early 2000s. Another prominent member of the APBR community during that formative era was Kevin Pelton, who as of August 2015 was an NBA writer for ESPN Insider. He also was the lead writer on ESPN.com’s February 2015 ranking of all thirty NBA teams in terms of their analytics commitment and implementation.

  Pelton traces his own statistical leanings back to his childhood reading of Rick Barry’s Pro Basketball Bible, and has spent much of his writing career trying to make advanced statistics more accessible to mainstream readership. At one point he was a moderator of the APBR forum, he wrote for 82Games.com, and he worked as a consultant for the Indiana Pacers. He also created the WARP (wins above replacement player) quantitative model.

  Before all of that, though, he was just trying to find some people like him who wanted to talk about basketball in new, mathematical ways.

  “I went looking [around] and found Dean’s stuff on the Internet—obviously, far and away the most advanced and most prominent work at that point—and then I think sometime around 2001, [I] stumbled into the APBR analysis discussion group,” Pelton said of his initiation into the world of advanced basketball thinking. “At that point, [that] is where Dean was posting. Hollinger, I think, showed up at some point that year, in 2001–02, because that’s when he started back writing about the NBA after he was at The Oregonian. Roland Beech was in the next couple of years, Justin Kubatko (founder of Basketball-Reference.com), people like that. That’s sort of when the stuff started to take its modern shape around that point.

  “That was the only place where you could have these kinds of discussions, with people who were on the same page, that had the same level of interest. Now that’s pretty easy on Twitter to find that kind of community because there’s enough people, but back then, it really was a small group. A lot of it was kinda trying to figure stuff out as you went. Around that time is when you had the first version of adjusted plus-minus from Wayne Winston and Jeff Sagarin. And what does this mean? Does this actually tell us something useful? There was no data or background at that point. We just had to use our intuition and discuss it amongst several of us and figure it out as we went.”

  Digging through the old posts on the Yahoo! ABPR group, it’s striking how many concepts being discussed back in 2001 are topics that are still being explored today. Just in Oliver’s posts alone, he broaches the concepts of individual player defensive value, defense effects on shooting percentages, why the Charlotte Hornets tended to play better without star forward Derrick Coleman, how to translate college statistics to the pros, and whether teams can actually win championships based on specific personnel strategies—in this case, a “twin towers” approach with two centers.

  Today, thanks to modern technology and a couple of decades of thinking about these types of questions, we’re moving closer to legitimate answers. In 2001, though, things really boiled down to a handful of really smart people bouncing numbers and ideas and theories off each
other; given the much more limited data available at the time, there was no real way to test accuracy. In one of Oliver’s posts, he wrote about tracking a specific game for defensive stats analysis purposes, and “hoping TNT doesn’t cut away” from the game so he wouldn’t lose any of his data.

  So, whenever a new stat or approach came along, there was significant exploration—and disagreement—about its meaning, especially when it came to outlier cases.

  “One of the memorable discussions that I’ve talked about in the past was the Timberwolves PR crew—the person who does their game notes was a friend of mine—they started including Kevin Garnett’s net plus-minus in their game notes,” Pelton said. “This was around his MVP season, and it was preposterous. It was like a plus-20 differential when he was on the court versus off the court. And people were like ‘No, this is preposterous. There’s no way one player can be that valuable,’ so it was kind of difficult to wrap your mind around that at the time because there was no context for it. There was no history to compare it to. Obviously, there is now, and [it] quickly became [available] when Roland Beech started putting it on the 82games.com site. So that’s sort of [an example] of breakthroughs and difficulties that these different concepts created.

  “I think [discussions] could certainly be a bit contentious. There are a lot of opinionated people in this field. That certainly hasn’t changed. A fundamental level of respect [existed, though]—especially for those who had contributed over an extended period of time and we’d seen the way their thought process worked.”

  While in today’s professional sports, teams are more frequently hiring independent researchers and/or writers to help with their analytics evaluations, many of these formative discussions were happening prior to the Red Sox hiring James and well before Oliver was hired by the SuperSonics. The idea of moving these types of analyses into the mainstream and then getting hired to implement them for a sports franchise wasn’t very reasonable at the time.

  “I’m sure it was for some people, but it seemed too big of a goal,” Pelton said of the sentiments of those participating. “The idea of getting people that understand the concept of true shooting percentage in there would be a big thing. And obviously, Moneyball kind of accelerated things and showed the possibilities in another sport, and that kind of raised the bar in a way.

  “There was a discussion in the intro [of the first Basketball Prospectus we did for the 2009–10 season]—I wrote my own intro—and I think there were eight teams at that point who had someone working for them in the front office or as a consultant to the organization. Now it’s probably twenty-eight or something like that, in a really short period of time.”

  Despite all of the brainpower working to unlock basketball’s deepest secrets, implementation hasn’t been the smoothest path. There are so many hurdles to overcome beyond whether someone can actually collect and analyze the data properly to come up with new, innovative solutions to a basketball problem. Many of those challenges involve communication across nonmath specialists and, at the end of the day, impacting an extremely dynamic sport played by human actors who are not capable of flawless implementation of strategy, even if the strategy itself somehow is flawless.

  Still, Pelton thinks back to the formative days on the board, where everything was new and exciting and still really foreign in a lot of ways, and appreciates how far things have already come.

  “It’s kind of easy sometimes in the process to just see incremental changes and not notice them that much,” he said, “and also see the setbacks and get discouraged by them, but when you actually take those opportunities and can step back and look at the bigger picture and the context of everything, it’s really crazy how quickly this has happened.

  “I mean, the similar revolution in baseball took a couple of decades, at least, and this took about half the time, in part because [baseball] helped pave the way.”

  An October 2005 article in Sports Illustrated by Chris Ballard helped announce the arrival of what we now accept as the modern analytics-enhanced basketball world, and the piece doubles as a genealogy of many of today’s top NBA coaching and management thinkers.

  Beyond discussing some of Oliver’s work with Seattle, the article mentions twenty-nine-year-old Sam Presti, who was an assistant general manager with the San Antonio Spurs at the time and now is the general manager with the Oklahoma City Thunder. He’s widely cited as one of the sharpest front office people in the league, has an extremely strong draft track record, and also was at the center of perhaps the most impactful trade decision of this decade. It also introduces twenty-seven-year-old Sam Hinkie, who then was a special assistant to Houston Rockets owner Carroll Alexander, and now is the general manager of the Philadelphia 76ers, where he has been the chief protagonist in one of professional sports’ most debated team-rebuild strategies.

  Additionally, it mentions San Antonio Spurs head coach Gregg Popovich. Popovich, of course, is in the discussion as the greatest head coach in NBA history, having won five NBA championships in San Antonio while churning out annual fifty- and sixty-win seasons. He’s also responsible for the NBA’s newfound appreciation of resting players and more balanced minute distributions across his roster, and is the figurehead for what is the most secretive analytics-friendly franchise in the league.

  It also goes into some depth about thirty-one-year-old Boston Celtics senior vice president of operations Daryl Morey, who has since become the public face of NBA analytics. He helped found the wildly popular and influential Sloan Sports Analytics Conference (Morey has an MBA from MIT’s Sloan School of Management) while implementing increasingly sophisticated playing and personnel strategies as the general manager of the Houston Rockets.

  In the subsequent decade or so since that article ran, the entire landscape has continued to evolve and morph into what we see today (with much of it continuing to be proprietary and unseen by the public). There’s way more money involved in the NBA today than even ten years ago, and teams have to work harder and harder to find and maintain competitive edges. How they’re doing so varies wildly from team to team, and heavily involves state-of-the-art technology to try to move ever closer to solving an impossibly complex and nuanced sport.

  CHAPTER 2

  The Basketball Technology Revolution

  The power of having good algorithms is that it’s like you have a million pairs of eyes watching every single game. . . . It’s as if you have someone who has watched every single second of your opponent’s game and you can get very complete scouting, or every single second of a particular player that you’re trying to scout.

  —Rajiv Maheswaran, CEO and cofounder, Second Spectrum

  One of the major epicenters of basketball’s ongoing technology movement spent some of its formative months in 2014 about a mile or so from Staples Center, in a hybrid work/live high rise on Los Angeles’s tony Wilshire Boulevard. While the building’s marble-adorned lobby with the make-your-own-espresso machine, as well as the rooftop pool and Jacuzzi, suggested hints of decadence, the crowded second-floor office of Second Spectrum didn’t match that anticipated standard.

  The hall the office was located on was more traditionally apartment building-esque, with rows of closed doors on either side, and the sounds of hip hop pulsating from a room a few doors down. Inside, the one-room headquarters already was straining to handle the company’s growth. It was shaped like a flat-bottomed U, with a reception desk (sans receptionist on this occasion) somewhat awkwardly lodged near the doorway in front of the right prong. Rows of minimalist desks lined the walls on both sides of that prong. In the back right corner was tucked Rajiv Maheswaran, the company’s CEO and cofounder, who with his team is starting to fundamentally change the way many think about professional basketball, from teams and players to fans watching the games.

  Maheswaran and Yu-Han Chang were both computer science faculty at University of Southern California when they brainstormed an idea to track NBA player data, which turned into a research paper submission fo
r the 2012 Sloan Sports Analytics Conference. The paper, which used player- and shot-tracking data along with machine-learning techniques, redefined what was then known about rebounding. It broke down the process of collecting missed shots into functions of initial on-court positioning, the hustle to pursue the ball once it came off the rim, and the conversion of opportunities to secure a rebound. The work won the conference’s best paper honors, and helped earn the duo the attention of some NBA teams. They and a third colleague, Jeff Su, formally launched Second Spectrum in 2013.

  Feeding off the NBA’s new initiative with SportVU motion capture cameras that capture every movement on the court, Second Spectrum set out to interpret mountains of raw data, then layer video and/or graphics over it to serve a variety of constituents. By August 2015, the company, which has a staff literally composed of rocket scientists, was employed by nine of the NBA’s thirty teams, and was starting to marry data, technology, and presentation in landmark and visually unique ways.

  Teams were using them as a third-party provider to help enhance on-court strategy, and Second Spectrum also was translating in-game plays into statistic-heavy motion-enhanced graphics, so fans attending games could see things like the projected point values of the various pass and shot options as the ball moved around to different players and spots on the court. In-arena work started with the hometown Los Angeles Clippers, and the company also has had its graphics product (called DataFX) used on ESPN’s SportsCenter, NBA TV, Fox Sports, and other national outlets.