Solving for W

This year’s Sloan Sports Analytics Conference featured a bevy of fascinating papers, panels and discussions, a good grip of them basketball-centric. Taken together, they paint a picture of an NBA stats landscape that had even Bill James, Godfather of Sabermetrics, suggesting the sport might soon rival baseball in its employment of advanced analytics.

TrueHoop at MIT Sloan Sports Analytics Conference

But amidst the Matrix-like reams of data flowing forth from the analytics ether, an elephant in the room remains: If you can’t correlate the data to actual success, then what’s the point?

The question loomed large over a number of otherwise thought-provoking presentations; the silent specter of a science simultaneously shedding light upon -- and wilting in -- the shadow of its own singularity.

In From 5 to 13: Redefining the Positions in Basketball, the winning "Evolution of Sport" talk at the conference, Stanford senior Muthu Alagappan employs topological data analysis (It’s OK, neither do I) by way of 20 seasons of NBA statistics. What he winds up with are 13 distinct groupings of players, each encompassing its own specific skill sets. Alagappan contends that such categorization paints a much clearer picture for how teams should approach roster construction, rotation implementations and draft priorities.

By just about every conventional account, Tyson Chandler is a center. According to Alagappan’s research, however, he’s a “Rim Protecting Big.” Think Mike Conley is a point guard? Try “Ball-handling Defender.”

On its face, this all makes perfect sense: If data suggests that such categories more accurately reflect a player’s abilities, skills and statistics, clearly the time is nigh for teams to begin constructing their rosters accordingly. Because, presumably, doing so will translate into more wins.

Here’s the problem: Alagappan stops well short of drawing that conclusion. His one example only hinted at such a correlation: Team A (last year’s Dallas Mavericks) boasted an especially dynamic roster with no less than 10 of the 13 positions covered, while Team B (last year’s 17-win Timberwolves) was duplicative in numerous positions.

But that’s one season. To the extent that teams whose players are very good statistically are more likely to achieve postseason success, then of course your findings -- which draw from those same statistics -- will suggest their own efficacy. But do such tautologies really get closer to defining how NBA teams win?

Why not look at the last 30 or 40 seasons, and see if those same truths hold? If teams that boast dynamic rosters -- say where 10 or 11 of Alagappan’s 13 positions are covered -- do, in fact, win, then you’re on to something. As it was, Alagappan’s research shows no such data.

Alagappan talks of “objectively defining a player’s true position” (italics mine), without suggesting how such definitions might translate into team success. Unless you can show that changing the categories -- and filling as many of them as possible -- better sets your team up for success, what does it matter?

Or, to put it another way, what useful purpose would such a positional reconfiguration achieve? Rob Mahoney, writing at the New York Times’ Off the Dribble blog, tackles this point beautifully:

At its core, position is merely a reflection of a player’s role within the concept of a team. In that way, position is not necessarily indicative of how well a player plays … but how a player plays. To put it another way: a 3-point shooter is a 3-point shooter by way of attempts (which reflect utilization and role), not makes. James Jones and DeShawn Stevenson play similar roles, and in terms of positionality, the fact that one shoots 42 percent from beyond the arc and the other 26 percent is absolutely inconsequential.

Perhaps a GM will pick up on Alagappan’s modeling and actively look to construct a roster broad and dynamic in its skill sets. Perhaps that team will win multiple titles. Then again, the notion that last year’s Dallas Mavericks were really, really well-constructed is one which even stats agnostics seem quick to endorse.

Maybe certain teams are well-constructed because they win, and not vice versa. Maybe it’s much easier to pinpoint success in hindsight than it is to try and duplicate that hindsight into an actual strategy for team construction. After all, if every team employed Alagappan’s methods, there would inevitably remain teams like the 2010-11 Timberwolves that boast far too many players in specific positional groupings. In this scenario, there would still be just as many losses as wins, just as many sweet-shooting teams as brick-laying ones, and just as many dynamic rosters as those mired in redundancy.

That doesn’t mean certain statistics aren’t more useful than others; the days of per-game metrics guiding front office decisions are clearly waning. As such, the ideas and perspectives offered up by Alagappan and others are nothing short of ground-breaking in terms of the sheer data they impart.

But there’s a difference between changing what we know about the game, and what -- and how -- we think about it. Using advanced analytics can show what we know, but it’s in how they’re used -- contextually, strategically, often in the heat of a split second -- that can make the difference between winning and losing, between trophies and lotteries.

For as much as modern analytics gives us in the form of fascinating raw data, we’re still very much scratching the surface of how that data translates into wins. Which, after all, is what it’s all about, isn’t it? Perhaps one day we really will find ourselves fully immersed in a brave new sports world of medical, mathematical and scientific analytics, where the human body itself functions more as cog than cognition.

In the meantime, what we’re left with is the image of a splitting atom, without much of an idea of how we get that image to power our homes. Through research presented in forums like Sloan, we’re flush with information -- lots of it -- but information without a real vehicle, much less a GPS-guided road map to wins and championship. And that’s OK. Because it’s in that lag time -- the gap between information and actionable results -- that the art, the music, the poetry, indeed the chaos of sports is allowed to breath.

Instead of seeing them as the paint which coaches, front offices and franchises will use to compose the future of sports, we should instead see stats as the strengthening canvas -- the increasingly sturdy base without which you wind up with nothing but a mess on the floor -- where the game is the paint, and the players are, and remain always, the artists.