I saw that SABR is providing highlights of the meeting so it doesn’t make much sense to provide my own summary here. But I thought I’d provide a few of my own take-aways from the second day of the SABR Analytics conference.
The GM Panel
I thought the panel discussion with Dick Williams (gm of the Reds) and Billy Eppler (gm of the Angels) was very interesting. It is interesting that the two men come from different backgrounds. Williams has more the business background, while Eppler comes from a background as a scout for the Yankees. So I think the perspectives of the two GMs are different. It seems that Eppler is more interested in what happens on the field (he watches very pitch of every game), while Eppler has a broader perspective and wants to oversee the minor league teams — he won’t be watching every pitch of every Reds game.
With respect to analytics, it seems that the Angels have more going on currently than the Reds. Williams was talking about building the analytics foundation — before this is done, one can’t get involved in more sophisticated analytics work. (For example, the Reds are less apt to use the new Statcast information currently.)
Williams was asked what skills are needed in the front office. He said, sql, R, data visualization, and machine learning. They want people are good in identifying a problem and working on the problem from beginning to a final conclusion. (This is the first time I hear R mentioned at the meeting.)
Two of the afternoon sessions were devoted to technology. One session brought together people who are involved in companies deeply involved in new types of measurements. For example, is now possible to make measurements of brain activity to better understand how batters make decisions about the type of pitch and whether to swing. It is now possible to put sensors on players’ arms so we understand better pitching motion and batter swings. This is really exciting stuff. Aaron Boone talked about all of the cliches that coaches use to try to improve performance. Now we’ll have data to really explain what it means to “stay on top of the ball.”
The second session included several people who work for Statcast. The type of data being generated currently is really fascinating. For example, I’d love to obtain measurements for each batted ball on exit velocity and launch angle. Some aggregate and extreme values of these variables seem to be available, but I’d like to data on each individual batted ball.
I was happy to see that two of the presented papers on this 2nd day talked about multilevel models — this is what I do. There is a challenge on how you present a talk on this type of modeling to a SABR Analytics crowd. Jonathan Judge from Baseball Prospectus gave an interesting talk, but he was careful not to use any formula — the problem was that I did not know much about what multilevel modeling he actually did. (I’ll be learning more about his work soon.) Scott Powers and Eli Shayer from Stanford took the other extreme approach– they essentially gave the same talk that they might give at a statistics conference — I think they may have lost a few with all of the formula. I’m giving a talk on similar modeling — I think it is safe to say that the technical level of my presentation lies between the two earlier talks. I strongly believe that multilevel modeling could become the default way of doing regression.
Anyway, it has been a fun meeting. I’ve had an opportunity to talk with my Baseball Analysis with R coauthor (we rarely have met face to face) and has been enjoyable to connect with a number of people who love baseball and appreciate the added enjoyment by the use of analytics.