2012 SSAC Review: The rest of the conference (Part II)

Completing my review of the 2012 MIT Sloan Sports Analytics Conference, I give my take on the panel sessions that I attended, as well as the Research Paper competition and the Evolution of Sport.  I spent most of my time checking out the trade show and talking to people in the hallways, so don’t look at this as a full review of the conference.  There are other people who have written very good reviews, and you can find them in the retweets of the SSAC’s Twitter feed.

So here we go:

Baseball Analytics


  • Rocco Baldelli (Tampa Bay Rays)
  • Jeff Luhnow (Houston Astros)
  • Scott Boras (Boras Corporation)
  • Mark Shapiro (Cleveland Indians)
  • Bill James (Boston Red Sox)

I’m not a baseball person, but I found this session to be the most informative and illuminating of the panel discussions I attended.  Baseball analytics have reached a high level of maturity compared to other sports, so the members of the panel could discuss their role in sports operations as well as their limitations in an open way.  I thought that Bill James’ comments were especially good, as well as those from Jeff Luhnow.

Basketball Analytics


  • Jeff Van Gundy (ESPN)
  • John Hollinger (ESPN)
  • Dean Oliver (ESPN)
  • Mike Zarren (Boston Celtics)

Mark Cuban was listed in the program book as one of the panelists, but Dean Oliver replaced him at the last minute.  The panel had no lack of candor and memorable quotes thanks to Jeff Van Gundy.  As you might expect, Jeremy Lin was mentioned early and often in the session, and Van Gundy reacted very strongly against the commonly-held belief that Lin was an outlier “missed” by the league.  Nor did he feel that Lin’s situation was all that unique.  Oliver made a good point about how unreliable a statistic the assist is in basketball; it’s very dependent on the scorer at each arena.  (It’s even more unreliable in soccer.)  Zarren said that not all of the teams are on equal ground in statistical analysis, but each team will use analytics to varying degrees (if at all), and the NBA has yet to figure out how to get a uniform distribution of intelligence in the team offices.

Gambling Analytics


  • Matthew Holt (Cantor Gaming)
  • Chad Millman (ESPN)
  • Bob Stoll (Dr Bob Sports)

This panel was one of the most heavily attended relative to the capacity of the meeting room.  Every seat was taken and there were people crammed in the end aisles and the rear of the room.  I admit to feeling very conflicted about gambling in general and sports betting in particular, but I attended the session to understand how professionals in the sector think, and it turned out to be very illuminating.  The panel discussed whether injuries or personnel were more important when it came to making a bet or setting a betting line.  Stoll said that he handicaps based on knowledge of the lineups, which means that watches and/or records all preseason games so that he has knowledge of each team’s depth chart.  Holt made some interesting comments about the movement of the betting line with respect to time, and Millman talked about the challenges involved in bringing on credible betting experts on his podcast show.  Another example of a panel that could have used a panelist from outside North America (esp. Britain).

Research Paper Track

As it turned out I listened to just two research paper presentations (I spent more time reading the posters – a high-quality poster is very important!). Those presentations turned out to be the finalists of the competition.

CourtVision: New Visual and Spatial Analytics for the NBA (Kirk Goldsberry)

This was one of the research papers I read before the conference, and I spent some time conversing with Dr. Goldsberry at his poster display.  I told him that I felt that his skills in geographic information systems serve him well in the creation of new analytics in basketball, and he said that I was the first attendee to tell him that!  The work incorporates spatial coordinates associated with each shot on goal to determine a measure of a player’s shooting effectiveness at all sectors of the court — a player’s “spatial shooting signature”.  The images are very beautiful and convey a shooter’s effectiveness very well.

Deconstructing the Rebound with Optical Tracking Data (Rajiv Maheswaran)

This paper was the product of work performed by Dr. Maheswaran’s research group at the University of Southern California’s Computer Science department, using data from STATS’ SportsVu Optical Tracking system.  The paper had some interesting findings on the probability of an offensive rebound given the location of the shot (It peaks just below the basket, descends almost linearly to a minimum at around 19-20 feet, and increases beyond that point).  It also discussed the implications of the height at which rebounds are received and the optimal positioning of players to improve their chances of rebounding.  This was an example of the paper being better than the accompanying presentation; it started strongly but got bogged down in charts by the midway point and ended up being rather confusing.  Apparently the jury saw something I didn’t, because this paper won top prize.

Evolution of Sport

The Revolution in Advanced Sports Analytic Systems (Kevin Goodfellow)

I attended just one EOS presentation, and it was this one by Sports Data Hub founder Kevin Goodfellow.  He presented on the increasing use of Big Data methodologies in sports analytics as organizations struggle with issues of “width, depth, and speed”.  Regrettably, I didn’t take the best notes here, but I believe width refers to the number of data sources or streams, depth the volume of data in each stream, and speed the time requirement for obtaining actionable information.  (Kevin, you’re more than welcome to correct me!)  Goodfellow said that these factors would lead to further use of Hadoop databases, which will be used in Sports Data Hub’s newly-launched OpenSDH platform.  Such a database brings its own set of advantages, according to Goodfellow: price (it’s open-source), power (can deploy on any server hardware), and people (can find/encourage Big Data technologists among computer scientists, mathematicians, statisticians, etc.).  I understand that OpenSDH has attracted interest from teams in various sports leagues and is something worth keeping an eye on.