Ed. note – I’m back, and hopefully for a good long while.
This year I found myself with 5 major conventions I was interested in going on within a 2 week span, covering my varied interests in baseball, stats, gaming, and religion. Alas, I could only afford to go to one, so I took advantage of the fact that the Joint Statistical Meetings were being held in Chicago, thus eliminating my need to pay for airfare.
JSM is the largest single gathering of statisticians in the world. The American Statistical Association, the prime organizer among the half dozen statistical societies that co-sponsor the event, always books the host city’s signature convention center in order to hold 6,000 attendees over the course of 6 days. This was only my second time attend, having previously attended the 2008 conference in Denver.
The conference has many different subjects being analyzed, from finance and risk to data modeling and visualization. But, with this being a blog that focuses on baseball, I’ll share 4 things I learned (and 1 thing I already knew) from the sports sessions.
1) The gap between academics and practitioners in statistics is narrowing
The ASA has a well-deserved reputation as an organization that favors academic pursuits in statistics. Its membership is predominantly employed in academia, and membership growth has not kept pace with the growth of the profession. Yet, it seems the explosion of data has led to more cross-over. One paper that was presented listed a FiveThirtyEight writer among its authors.
2) There are more ways to get into baseball than by studying it directly.
One of the presenters I was able to interact with has spent a lot of time looking at SportVu, the NBA’s player tracking data. This person is now slated to start working for a baseball team in the fall because of that work. You can probably guess that the position will focus on understanding Statcast data.
3) Sports teams aren’t looking for subject matter experts with stats knowledge; they want stats experts with subject matter knowledge.
At a panel discussion about stats in sports, both panelists that are currently employed by major sports teams noted that front offices are loaded with SMEs. They want people with stats backgrounds who can analyze the data well. The odds of anyone getting into a front office today in the same manner as Bill James is highly unlikely.
4) It pays to think about analogs from other fields.
Dan Cervone presented a model trying to value court space in the NBA. He built his model analogous to how real estate valuations are made. It’s the type of thinking that leads me to pay attention to JSM, and it also is, in my opinion, a requirement to getting the most out of the conference.
5) Ideas at JSM are starting points, not end points.
The fact that JSM does tend to emphasis statistical methods over results means many papers aren’t necessarily providing new insights into well studied issues, but new ways to analyzing the questions at the heart of the issue. Take this presentation on predicting outcomes of plate appearances. It uses a type of regression modeling designed to handle structured outcomes like in baseball. It may not provide any new insights into how baseball is played, but an idea like this could end up in the next great forecasting system.