BOSTON — I have no idea where to go, I have no idea what to write and I have virtually no idea what anyone is talking about.
It's nice to be back at the MIT Sloan Sports Analytics Conference.
This year's model is the seventh edition of a conference that began in 2006 as a small gathering of likeminded folks eager to share cutting-edge analysis that would sharpen their understanding of sports, and has since exploded into a major annual industry event. Sloan 2013 boasts more than 2,700 paid attendees, representatives from more than 90 pro teams across six different sports (including 29 of 30 NBA teams; the Los Angeles Lakers, alone, appear to be unrepresented), a slew of big-name panelists and big-money sponsorship from the likes of Under Armour, StubHub and primary underwriter ESPN, spread throughout the sprawling Boston Convention and Exhibition Center.
There are three simultaneous panel discussions on a variety of topics going on at any given time, plus presentations of research papers aiming to advance existing analysis, chats about potentially "revolutionary" concepts in the Evolution of Sport track — basically, five big-idea things happening every hour, with 10 to 20 minutes separating the infobursts. There's a show floor with more than 20 trade tables occupied by companies hawking stuff like data security solutions, social advanced box score programs and cloud-based clipboard applications enabling coaches to remotely store and share plays they've designed. There's a "technology room," which is to say, a game room, featuring Wii Golf, air hockey, a race-car-simulation video game and foosball tables, where you can blow off some steam when it all gets to be a bit much.
And it will get to be a bit much. With so much content, so many voices, so much stimuli and — to borrow a phrase that was very popular Friday, thanks in part to the presence of longtime stat rock star and New York Times best-selling author Nate Silver — so much "noise," how could it not? There's a massive amount of stuff here; it's hard to make sense of it all.
Which, now that you mention it, is pretty much exactly the problem facing the analysts, the NBA teams that employ them/for whom they want to work and the fans watching at home. We're being told more and more frequently that this stuff matters, that we should be paying attention to it and that we need to know about it ... but when it comes time to explain why, the message often seems to get muddled in translation.
Big Data is a big deal at this year's Sloan conference — while companies in other industries have been sifting through the loads and loads of unstructured information they produce for years in search of new insights and revelations, the sports world and the teams that inhabit it have been a bit slow to catch on. Now, though, not only has the league itself invested in data mining through the creation of its brand new stat site, powered by enterprise software giant SAP's HANA database technology, but more and more NBA teams are starting to deploy tools like STATS' SportVU system, which can unlock extraordinary new discoveries if teams can find the right key.
SportVU positions six special video cameras above the basketball court at different angles to capture, record and store tons of in-game information — player movement, referee movement, ball movement; where, how and how fast players are running; where, how and how fast passes are thrown; etc. Recording all that movement in high definition at 25 frames per second, every second, for an entire game makes for an awful lot of data points — 1 million individual records per game, in fact, according to STATS' Brian Kopp. That leaves the 15 teams that have purchased the cameras looking for an analytical edge — the New York Knicks, Toronto Raptors, Washington Wizards, Golden State Warriors, Houston Rockets, San Antonio Spurs, Boston Celtics, Milwaukee Bucks, Oklahoma City Thunder, Minnesota Timberwolves, Philadelphia 76ers, Phoenix Suns, Orlando Magic, Dallas Mavericks and Cleveland Cavaliers — holding onto more information than they know what to do with.
It sounds like a good problem to have, and it certainly seems preferable to not having access to a treasure trove of information. But it's still a problem, one recently articulated well by Washington Wizards vice president of basketball administration Tommy Sheppard to Slate's Jason Schwartz: "It's like saying you’re going to Wal-Mart or Ikea to get something. You better know what you want, or you're going to walk out with a ton of s***."
And even if you know what you're after — like how frequently individual players accelerate and decelerate on the court and where they most often do so, as NYU-Polytechnic Institute professor Dr. Philip Z. Maymin sought to learn in his research paper, or whether teams benefit more from crashing the offensive glass or from sending guys back on D as soon as a shot goes up, as explored by the MIT team of doctoral candidate Jenna Wiens, graduate students Guha Balakrishnan and Joel Brooks, and professor John Guttag — you've got to be able to explain what you walk out with. Along with Big Data, signal and noise, another key phrase on Friday was "actionable information," and its use was primarily associated with a question: How do you take all this analysis and present it to organizational decision-makers in a way that they can A) understand it and B) actually use it to make decisions?
It's not just about owners, general managers and vice presidents, though. Analytic revelations — especially the kinds that SportVU make possible by showcasing practical, on-court visuals and doing so nearly in real-time (Kopp said during a Friday panel that they can generate in-depth reports within 60 seconds) — need to be translated for players and coaches, too, if they're to be leveraged for actual in-game use. And while an increasing number of coaches are going along with the rising analytics movement — Rick Carlisle with the Mavs, Erik Spoelstra with the Miami Heat, Terry Stotts with the Portland Trail Blazers and others; heck, even George Karl's cool with having opponents' end-game tendencies loaded onto a laptop on his sideline — some are still reticent to lend too much credence to something they're not really understanding in part because it's not really being explained well.
"We need better people not at doing the stats, necessarily, but at communicating the stats — at building the bridges," said Kirk Goldsberry, who co-wrote a paper with behavioral analyst Eric Weiss that uses the SportVU data to examine interior defense, long one of the under-explored elements in analytical work, during a late Friday panel.
Goldsberry has made a name for himself over the past couple of years for being one of those bridge-builders, combining his academic training in spatial analysis and visual analytics (he's a visiting scholar at Harvard and cartographer, by trade) with his passion for basketball to introduce new ways of visualizing what happens on the basketball court. His work, both on his own CourtVision site and on ESPN's Grantland, relies heavily on clear, colorful shot charts and heat maps to quickly communicate information like shooting accuracy or defensive efficiency that's culled from the plotting of thousands of individual shot attempts. It's one of the most enticing parts about Goldsberry's work — yeah, the analysis on topics like who shoots well from where is interesting, but more importantly, it's easily digestible in a way that most of the rest of what you see here doesn't quite match.
During his Friday panel, Goldsberry inverted a common stat-head's criticism, arguing that when a crotchety coach snarls at the prospect of integrating advanced stats into his approach, the failure isn't the coach's, but rather the analyst's — that the responsibility for converting nonbelievers lies with the person doing the preaching. It seems like there's at least a tacit understanding of that in the community at large — the fact that Friday morning featured a panel specifically dedicated to how data visualization technology can be used to help communicate complicated stuff was a nice touch (even if it didn't quite come off, according to some) and you can see how smart, simple applications of visuals could totally work for players and coaches, so long as the primary goals of the project are clarity and communication.
And I'm thinking the same's true of people like me, too. If writers, bloggers, talking heads and others are going to use advanced stats to make points, support arguments or bolster claims, then the onus is on us to make sure we're doing so clearly and not just stacking up a bunch of arch numbers and opaque terminology in the hope that it makes us look smart. (I know I've been guilty of this in the past.) If we really believe these analyses can help us better understand the game and league we love, then we need to get better at teasing out the elements that appeal to human beings. Fans are well within their rights not to care about advanced stats, but if they turn away because it all seems impossible to grasp, then the failing's ours; that much, at least, seems clear.