The Elephant In The Room
This weekend is the annual MIT Sloan Sports Analytics Conference and several Nylon Calculus contributors are in attendance. Here we’ll be assembling some of our thoughts, reactions and experiences from the conference. We’ll be continuing to update throughout the day and into tomorrow so check back for more.
Are You An Analytic? — 11:18 AM, Saturday
Ask and you shall receive.
An hour after finishing my last update, wishing for a conversation about the definition of analytics, Kirk Goldsberry delivered just that in his presentation—Analytics Illustrated. The bulk of Goldsberry’s talk was about how visualization can illuminate and that combining visualization with other narrative elements, like traditional reporting, can make a story exponentially more compelling. But he also touched on defining analytics, asking whether the fact that Chris Bosh is left-handed should count as analytics. Handedness is (mostly) a binary situation and converting it to a numeric statistic adds no value. But it is still information and, in the broad and most useful definition, analytics is information.
Goldsberry shared his preferred definition of analytics — “reasoning artifacts, products of analysis that enable us to understand something we didn’t understand, or characterize something that needed to be characterized.” Adding, “if we can’t agree on what analytics are and aren’t, then we can’t talk about what needs to get better.”
Well said.
— Ian Levy
Defining Basketball Analytics — 9:05 AM, Saturday
Yesterday’s Basketball Analytics Panel—featuring Mike Zarren, Sue Bird, Shane Battier and Mike D’Antoni, moderated by Pablo Torre—had plenty of overlap with iterations from previous years. They group worked through the standard array of basketball analytics hot buttons—shot selection, communication, distillation into actionable information, pace, draft lottery incentives, exploring biometric analysis and injury prevention.
However, at the outset, Zarren laid out a really interesting series of questions that I think would have made for a much more powerful discussion. He started by asking if people in the audience would consider him sending a scout to a game to be analytics. What if he asked that scout to chart something very particular? What if he then had the scout take that hand-charted data and put it into a spreadsheet? What if he then used that spreadsheet to look for correlations between the data this scout had collected and some other previously existing data set? Clearly, that line of questioning ends with something that fits the common conception of analytics. But at what point in that series of hypotheticals did the process turn from “eye test” (or something similarly pejorative) into analytics? Zarren added one more question, pushing in a slightly different direction, “If I know how well a player slept last night, is that analytics?”
The implication is that analytics, at least as they were beginning to be discussed in this panel, were not adequately defined—a problem that I would argue is pervasive.
It has been pointed out by several people that Charles Barkley is only selectively anti-analytics. Points and rebounds are basketball analytics—albeit very simplistic ones—and he does not appear to find them morally objectionable. He appears to be angry and dismissive only of a certain type of analytics.
Analytics are easy to recognize and define at the center—Real Plus-Minus, SportVU data tables, five-man unit statistics—but as you move away from those things in any direction, the line between “this” and “that” becomes fuzzy. I would argue that most people, regardless of the interest level, would characterize sports analytics as being about additional information. I would also guess that most people think of sports analytics as a very specific sort of information that is captured, stored and used in a very specific and technical sort of way. The reality is that it’s all information. Complex statistical player projections to total points, from how many hours of sleep a player got to how a skilled observer would evaluate the post moves of player, it all serves the same purpose. Drawing arbitrary lines to differentiate between those different types of information really just creates the opportunity for people to stand on either side of the line and argue.
Perhaps the solution to the perpetual “communication” problem in analytics is to stop separating it out from everything else that teams do to try and win games.
— Ian Levy