I had the opportunity to attend the Sports Analytics Innovation Summit (SAIS) at the Marriott Union Square in San Francisco, and want to share a little bit about what I learned over the course of the two days. About half the talks focused on the business of sports, while the other half dealt with analytics used in the sports themselves, from Olympic training to what helps NBA teams win in the playoffs. Following the jump is a list of the key takeaway points I gleaned from SAIS.
- You may have figured this out already, but sports are a huge and growing business these days. The representative of SAP, one of the sponsors of the conference, mentioned that his company is making a big push into the space. Alex Sugarman, VP of Strategy & Development for the Chicago Cubs, presented some slides on the market size of professional sports. I believe the figure was around $70B (that's a "B" for billion).
- Many of the talks focused on ticket pricing, especially dynamic ticket pricing. It seems that the secondary market (i.e. sites like StubHub) has forced teams to re-think and improve their pricing strategies. Nothing drove this point home to me more than the fact that the Cleveland Cavaliers hired a lady named Sezin Aksoy (who gave a talk), who previously worked for the airlines industry, to improve their pricing model. She told us that some of the same things we hate about the airline industry (such as overselling seats) would also make their way into sports ticketing. John Abbamondi from the San Diego Padres mentioned that they wanted to "train" their fans to buy tickets early before prices would go up during the season.
- Analytics plays a huge role in Olympic-level training. It's to the point where if a young skier in some small town in Colorado starts to post increasingly good times, an e-mail will be automatically sent to Troy Flanagan, the High Performance Director of the US Ski and Snowboard Association (who, by the way, was formerly an aerospace engineer) who gave a talk about the training those elite athletes receive nowadays. It's really incredible stuff.
- Roland Beech, Director of Basketball Analytics for the Dallas Mavericks, gave a talk that mostly focused on using Synergy statistics to predict playoff success in the NBA. His talk reminded me the most of my own analytics experience, although he was most likely holding back on most of the interesting things he has actually done. That brings me to the next theme...
- Secrecy. In many ways, SAIS was like the cute girl that flirts with you, but has no intention of letting you get anywhere with her. Teams have so much access to data and analytics, and are probably doing a ton of awesome stuff behind closed doors, but they absolutely have to treat those advances like trade secrets. So much so, that one of the speakers at the meeting was actually a lawyer whose talk focused on keeping trade secrets in baseball! Throughout the meeting, speakers would give us a taste of what they're doing, but then follow with, "I probably shouldn't say too much more about this." Argh!
- Kirk Lacob spoke about some of the analytics that the Warriors are doing, including use of the SportsVU player tracking data. In fact, they have contracted with a company that is taking the raw player tracking data and automating the process of translating it to Synergy-like play categories. And who knows what else, Kirk couldn't divulge much more. What was very clear from his talk, and should comfort you as a Warriors fan, is that the front office is seriously committed to doing everything it takes to win. Kirk specifically said they are not in this to make money. The goal is to win. Unfortunately, Kirk bolted from the conference immediately after his talk, so I didn't get a chance to catch up with him. (Not that he could've told me much more, of course.)
- Ben Alamar, Director of Analytics & Research for the Oklahoma City Thunder, talked about innovation in analytics. I was expecting him to talk about new stats, but what he focused on was even more intriguing. Basically, Alamar has developed tools to test a player's decision-making ability. The example he showed was a picture clearly depicting a 3-on-1 fastbreak opportunity, with three of the offensive players labeled A, B, and C. The player taking the "test" (which is more like a game to them) would have to decide within 5 second of seeing a very short video clip which of the three players scored on the play. It wasn't clear yet how much the test correlated with winning, but Alamar did tell us that so far, the players who have taken the test have performed much, much better than, say, the folks working in the front office who have taken the same tests (think of them as the "controls" in this case). Unfortunately, OKC is not allowed to bring the test to the draft combine yet. They can only test players at the Portsmouth Invitational, in addition to members of the Thunder. It wouldn't surprise me, however, if other teams around the league are starting to go down this road, if they haven't already.
- Kirk Goldsberry, a Visiting Scholar at Harvard, discussed more of his visualization work involving the SportVU player tracking data. Much of what he presented was the same data from the MIT Sloan Analytics conference in March, but he did show some new passing data. What was novel about this set of data was that it showed the trajectory of the pass, from beginning to end. Needless to say, a lot more can be done with those kinds of data than just drawing them on a plot. Unfortunately for us "regular" fans who don't have access to the data (yet), we'll just have to wait for Kirk to crunch the numbers himself and let us know what he finds.
- Steven Angel, Senior VP at the NBA league office, actually gave a very interesting look into how the league deals with issues like making rule changes and refereeing the refs. They do a lot more than I ever thought. For example, the league actually did a study to try to determine how much particular refs affect the outcome of games. If I understood correctly, they basically created a set of game predictions using Vegas spreads (sound familiar?), and then regressed those onto the refs, which were treated as independent variables. Think of it as an adjusted +/- for refs. Very cool stuff. Also, someone asked Angel about what the league intended to do about flopping. Basically, he said they didn't want to put that responsibility on the refs. Instead, it will probably be dealt with through fines.
- Elaine Allen, Professor of Statistics, and Julia Seaman, a graduate student, both at UCSF, gave a talk about the baseball DL. The most interesting stat I learned was that in the last 10 years no team that put a player on the DL in the month of September has gone on to win the World Series. Apparently, if this trend holds up, that would eliminate the Yankees and the A's. We'll see.
- Brian Hampton, Director of Football Administration for the 49ers, gave a talk about game strategy. One of the things I learned that surprised me is the anti-technology rule that the NFL implements. You know all those cards that you see Jim Harbaugh covering his face with during the game? They need those cards, because laptops and computers are not allowed to be used in the game, by anyone. As a funny aside, Hampton mentioned that he once saw Ken Whisenhunt holding a card that he had constructed for the 49ers! Although Brian couldn't share much in the way of analytics details, he was able to share with us the "kneel zone" table, because they were fairly certain that by now most of the league, if not all 30 teams, had pretty much the same table. For a given down and remaining number of timeouts (say 2nd down and 1 timeout left), the kneel zone table gives you a clock time that you are safe having the QB kneel.
I had a good time at the meeting. Maybe one of these days I'll be up there presenting my own stuff and telling people I probably shouldn't say any more about this. ;)