Sunday, March 4, 2012

Sports Analytics: MIT Sloan Conference Impressions

The fascinating MIT Sloan Sports Analytics Conference which took place in Boston this weekend (and which I had the privilege to attend) is known affectionately by its attendees as "Dork-palooza" -- a celebration of the number crunchers and obsessive compulsive fans of sports statistics. Panelists, researchers, and attendees discussed a wide variety of topics with a very ambitious goal: to change the way we watch and measure team, individual, and business performance in sport.

The presence and influence of the great Bill James cast a warm glow over the proceedings. James' hugely influential work in baseball statistics, and the Society of American Baseball Research's "saber-metricians" remain the best case study of how the geeks can influence a sport. In panels, research papers and hallway conversations many conference participants talked to the broader project of replicating in other sports James' revolutionary success rethinking how baseball is measured and discussed.

A couple of themes stood out for me in the small subset of the conference I got to sample:

New Data Collection Methods. A couple of research papers on the NBA showed how promising the STATS SportVU 3D positional data will be for future basketball analysis. Data mining the 3D positional information of players and ball can provide some intriguing insights. Several panelists and companies talked about re-introducing context to measurement -- think, for example, about the problem of NFL receivers dropping passes when measuring a quarterback's completion percentage. A hockey panelist talked about "contextual +/-" in assessing the influence of a player's time on ice. And a promising company called StatDNA, which presented in the startup competition, uses trained experts to visually analyze soccer matches, coding events with remarkable fidelity and perhaps some interesting predictive power. I didn't attend any of the bio-metrics panels, but it seems equally interesting work is being done with acquiring granular medical data and using it to enhance athletic performance.

New Methods of Communicating. In a remarkable panel called "Box Score Rebooted," the participants talked about the need to focus on, for lack of a better word, poetics, the languages we use to communicate deep, data-driven concepts. For example, Bill James talked about the decision to name the arcane Blown Save Win concept in baseball "BS Wins" to preserve the double entendre. Muthu Alagappam tried to expand the concept of "positions" in basketball beyond the five traditional ones with data analysis and visualization, revealing 13 actual behavioral groupings that were much more evocative of observed play (I got to see an early preview of this research when his company, Ayasdi, presented to us at Benchmark a few months ago). Kirk Goldsberry developed a visual language of NBA players' shooting efficiency -- point production per shot attempt -- at 1,200 relevant positions on the offensive end; he produced a series of heat maps which allowed delightfully intuitive comparisons of the capabilities and tendencies of the league's stars.

Modernizing Old-School Methods. Obviously, the human mind is amazing at visual pattern recognition, and sports are filled with phenomena that are intuitively obvious to people both inside and outside the sports who pay close attention to games, but which would be hard to quantify through statistics or machine learning. Several panelists reminded the audience of the continued vitality of film analysis to players, coaches, and scouts. The ability to digitize and add metadata and search to video clips, as well as the ability to personalize the film session and distribute game film on iPads, may ultimately be more important than whiz bang  innovations like USC's Voronoi Tessellations (cool as they were).

"Fanalytics", Social Media and Business Intelligence. Sports is a business, and cash money rules. While a lot of analysts justified their work because it could theoretically add efficiency to drafting, trading, transferring, and managing salary for players, there was an equally compelling movement to bring data to bear on the customers -- sports fans -- to study their tendencies, behaviors, and spending patterns to better serve them (and get them to spend more, of course).  From the CEO of Ticketmaster, to Mark Cuban and other team owners, to Vegas bookmakers, there was a consensus that data mining and analysis could provide meaningful uplift, and that social media engagement was crucial. The growing importance of social media to the experience of sport cannot be overstated. In particular, the degree to which Twitter is connecting audiences, athletes, clubs, and even the analysts themselves, is astonishing. It would not be a stretch to call Twitter the deus ex machina of modern sport. [Disclaimer: Benchmark is a significant venture investor in Twitter].

In a broader context, the SSAC is further proof of the revolutionary and disruptive business impact of big data, business intelligence systems, and advanced statistical and marketing analytics. These tools have transformed entire market segments -- think about what data and analytics mean to the book selling business after Amazon, or to the video game business after Zynga, or the advertising business after Google, not to mention oil drilling, or inventory management, or Wall Street investing. Several Benchmark portfolio companies like Hortonworks (Hadoop), Domo (BI and visualization), New Relic (application performance monitoring), and others are helping customers solve data/analytics problems, while many, many portfolio companies are using data analytics fundamentally in their businesses to create competitive advantages and enhance customer experiences.

Analytics has become a key business practice across industries. But it's cool to see it flowering in sports, where it uniquely functions both as management tool and a component of the fan experience -- a unique form of content separate from, but enhancing, the core product.