1

Beauty at a Smaller Dimension

Photo by Sheila Tostes on Flickr

Photo by Sheila Tostes on Flickr

One of the simplest ways to understand analytics in basketball — meant broadly to include both advanced stats and the use of systems like SportVU or Catapult that track player movement or bio-feedback — is as a way to see more of what is already in front of us. At the MIT Sloan Sports Analytics Conference this year, there were presentations on new methods for doing just that, from dissecting the act of rebounding into smaller slices of positioning, crashing, opportunity and actual rebounding to developing a system that can harness the data produced by in-arena SportVU systems to more accurately recognize and compare ball-screens.

But Sloan is also a marketplace, with companies competing to sell their goods and/or services to teams and teams harnessing proprietary technology that stays behind closed doors while a trickle of new data leaks out to the public, with many clamoring to know “But what does this TELL us?”

Statistician and Yale professor emeritus of political science, statistics and computer science Edward Tufte’s presentation on the conference’s final day was a much-needed reminder that this data shouldn’t tell us stuff, but should instead show us things, and do that job as clearly and comprehensively as possible.

Anyone who even half-follows the developing world of analytics could be forgiven for occasionally wondering if this vivisection of the game of basketball isn’t in danger of turning into an autopsy. You might have especially been feeling this way if you’d seen the flat, flavorless presentation on ball-screens which seemed to present as revolutionary something anyone with two eyes or (maybe even one eye) could do: tell if a ball-screen had occurred on the court. (I generally do this by seeing if a player guarding the ballhandler ran into another player standing still and didn’t do it accidentally. And, in fairness, the paper itself does a much better job of explaining why this work is relevant and possibly important in basketball.)

But Tufte urged us not to see it in such pessimistic terms. “Sports analytics is about the sports activity itself,” he said. A simple statement but a radical one in a sports environment often saturated not only with commercialism, but painted in broad and stentorian strokes by commenters fixated on a moral dimension of sports, extracting meaning based on assumption, not observation. “Sports analytics on a good day is about the subtlety and inherent beauty of sports performance,” Tufte continued. “Don’t let anyone tell you that analytics is somehow reductionist, that it takes the joy, the passion out of sports. I think it mainly takes the stupidity out of sports.”

Referencing this image from the New York Times, which details the complicated flow of signals across a baseball diamond with refreshing clarity, Tufte said, “Now we watch the game just a little bit differently because you know this is going on. This also illustrates that people were doing [this] a long time ago. They were doing it by hand, though.” That is, a better understanding of the spatial interactions on the baseball diamond or basketball court didn’t begin with motion graphics, with advanced technology that can track the velocity of players 25 times a second. Consider the humble play diagram, and how much it conveys with just a few lines and numbers. (And incidentally, how subtle yet obvious is it to describe the action of dribbling the ball to a different spot on the court with a bouncing, squiggly line?)

In the rush to quantify every dimension of on-court action, we can also lose site of the fact that the study of sports is a social science, not a natural one with immutable, discoverable laws that apply across all of space and time. “Social science is not rocket science,” Tufte said. “It’s much harder than rocket science. We make discoveries about human beings and they learn about the discoveries and they flip them, they reverse them.

It’s important to keep this kind of thing in mind when talking about “solving” basketball by emphasizing corner 3-pointers and shots at the rim. To crib from game theory, basketball is not a game against nature — a contest in which a single player must make a strategic decision against a disinterested opponent like, say, a tornado. That is, the tornado doesn’t adapt to your decision to head for the storm shelter or hop on a bicycle.

But in the NBA, the landscape of strategy is constantly shifting, based not only on rules that make the 3-point line shorter in the corner or handchecking illegal on the perimeter, but on defensive strategies like strongside overloading or having the big drop down on pick and rolls. As teams learn things from analytics, they decide to do different things, which in turn means teams learn different things from analytics.

Beyond the flexible nature of the thing itself, there’s the thorny issue of how to present the research — truly Tufte’s area of expertise. A presentation earlier in the day entitled “Getting Information from a Haystack: Finding Value in New Data” seemed to point towards this, but it turned out to be more about how to package and sell services to clients, not how to work as clearly and intelligently as possible with the data. For Tufte, the beautiful expression of the information is an end in itself, and not tied to how profitably or seductively it can be presented. In fact, he sees that as a pitfall. “People oversee patterns,” he said. “We’re all probably better at believing than seeing. That’s not a good combination: over-detecting patterns, believing too quickly and under-detecting bullshit.”

Tufte sees good information design, good data visualization as a way to combat this problem. It comes down to seeing the field in its depth and breadth as clearly as possible and designing illustrations that reveal that complexity with clarity. Perhaps the most resonant chunk of Tufte’s presentation was this video clip of physicist Richard Feynman defending the battlements of rigorous analytical inquiry against those who would say it ruins simple beauty.

In discussing the beauty of a flower, Feynman says, “It’s not just beauty at this one dimension, one centimeter; there’s also beauty at a smaller dimension.” That is perhaps analytics greatest promise: that understanding itself is beautiful, that there is a joy inherent in discovery, and that there is beauty wedged in between the things we can currently see. Yes, analytics holds forth utility to teams and players, competitive advantages to be gained or overcome. But for spectators it is beautiful science. “Science and knowledge only adds to the excitement, the mystery and the awe of the flower,” says Feynman. “It only adds.”

 

Steve McPherson

Steve McPherson is an editor for Hardwood Paroxysm and his writing has appeared at Grantland, Rolling Stone, A Wolf Among Wolves, The Cauldron, TrueHoop, Complex, Narratively, Polygon and elsewhere. His Twitter handle is @steventurous.