2017 Sloan Sports Conference: 5 questions with Aaron Barzilai

The 2017 MIT Sloan Sports Analytics Conference is a gathering of some of the smartest and most engaging basketball minds on the planet. While we’re here Nylon Calculus is going to try and pull a few of these brilliant minds aside for some short Q&As. Also, check out our running blog for Day 1 and Day 2 of the conference.

Aaron Barzilai (@basketballvalue) is the Chief Data Officer for WinnersView and the former Director of Basketball Analytics for the Philadelphia 76ers.

Nylon Calculus: What have you enjoyed at the conference so far?

Barzilai: I wanted to make sure this year that I got more of the research papers. I thought there were some really interesting ones. Definitely R&D, doing innovative stuff but you still need to think about what that’s going to turn into as a product, or in this case, the actionable instructions that are going to filter down to the one or two things a player gets told ahead of the game. But it’s just really cool to see the cutting edge stuff.

The possession sketches stuff was interesting. That group, Luke Bornn, Andy Miller, those guys from XY Research always do really good stuff. I was also intrigued by the 3-D Body Shots presentation. I thought that was really interesting, just computer vision in general was amazing, and that from a 2-D picture they’re able to figure out 3-D orientation of the limbs and draw stick figures that, I think, had like 17 elements to the skeleton. I noticed that in the first video they showed, not only were they showing you the clip of the play, you could even see the skeletons of many of the fans and the people on the bench, and I was joking with a friend of mine that now we can’t start to do analytics on who’s the best 12th-man — who jumps the highest and up-and-down off the bench the most during a game?

I thought that was all really interesting. Again, I wonder how far it is from something that’s actionable.

Nylon Calculus: So you’re one of the few who has been to every single Sloan Conference, do you feel like there is a thread to the research? Do you see an arc to the kind of research that has been done over the history of the conference?

Barzilai: It’s almost more like the classic player trajectories. There’s different waves of people. At first, many people did not have jobs with teams. Many teams did not have people working for them yet, and so you kind of had the usual suspects in the first wave there. Then you saw people like Kirk Goldsberry and Rajiv and Yu-Han from Second Spectrum were presenting and then they’re kind of inside now. And so now you’re seeing a lot of great work coming out of Luke Bornn and his team, but also Patrick Lucey who did the Body Shots work. It will be interesting to see how public they remain.

In a lot of ways it’s kind of like the next draft. There’s probably someone in college, an undergrad right now that’s going to do their first research talk next year and be really impressive two years from now. And then four years from now they’ll no longer be presenting because they’ll be in-house somewhere.

Nylon Calculus: Seeing the evolution of the conference, it’s grown so much, are there any areas that you feel are underserved? Is there anything specific you’d like to see more of?

Barzilai: As it became huge, now we’re having multiple things on the NBA. I think this conference has always been NBA-centric, largely due to Daryl Morey’s benevolent interest. But it’s always had a football panel. I think it’s always had a hockey panel. You know baseball, I wasn’t able to go, but on Twitter it looked like the Statcast presentation was really good. So I feel like the sports have always been pretty well represented. I mean, this year I saw a poster on MMA analysis. I went to a DFS panel, there’s an eSports things, which I think people of my generation might not understand but is clearly the future.

The conference is big enough that it covers a lot of things. What’s good and bad is that it has grown so much you can’t attend it all. The very first conference I want to say was on the order of 100 people. It was actually in classrooms at MIT. I feel like we spent more time with everybody in the same room and then we might break out and come back together. Here there’s five, six, seven things going on all at once. So you can’t possibly go to the research presentations and go see everything.

But still I always enjoy the conference, every year I’ve been. I still think you’ll hear different strains of complaints. You know, some people will say it’s not researchy enough, even though it as the research presentations. You just need to understand what the conference is. It’s a lot of people coming together. The content is great and it’s an opportunity to talk to people you haven’t seen in awhile. If you think of the people working for teams in analytics, which turns out to be a very small segment of the conference now, many of them don’t travel with the team. So you might be on Team X and not really talk much at all with your counterparts. It’s kind of a nice chance to connect on common experiences.

Nylon Calculus: Can you talk a little bit about WinnersView and what you’re doing with them?

Barzilai: WinnersView is this fun project I’ve been working on, a sports media start-up focused on analytics. We’re focused on the NBA and the NHL for starters. It’s a joint venture that’s funding it, one is a company that Ted Leonsis owns, who of course the Washington Wizards and the Washington Capitals. Tegna is the other partner. They are the largest local television station group in the country.

It’s really exciting. We’re trying to on sharing analytical insight. Like the session we were just in, Data Driven Storytelling, I think it’s more and more how people want to understand the game. So, we’re trying to do things that are unique. One of them is focusing on short videos. I think 99 percent of analytics content that we all consume is all written. More and more people like to consume media via videos, so we’re trying to videos that are easy to understand and appeal to a much more mass audience than the thousands of people here at this audience.

We’re focused in a few markets now and hoping to expand in the future. We’ve done stories recently on how Otto Porter was leading the league in points per possession on cuts off the ball. Or looking at how CJ McCollum has turned himself into basically the most efficient mid-range shooter in the league. It’s been really a lot of fun. I’m thinking about the same kinds of things I was thinking about in the front office, but doing it in a more public way.

Next: 2017 Sloan Sports Conference: 5 questions with Chris Herring

Nylon Calculus: What are some of the challenges of adapting this kind of analytic insight into video?

Barzilai: It’s almost like trying to boil down a whole article, whether it’s a research paper here or just an insight, into essentially the equivalent of 140 characters. I mean it’s not quite that, but it’s also not that far off. I don’t have a thousand words, I have a hundred or something like that. You really need to boil it down to its essence. What’s the one-sentence summary of this piece? Like if you watch this piece and then you’re telling someone about it at the bar tonight, what are you going to say?

The other trick is, you just can’t go into the gory details of your method. The art is, if your cousin watches it we just want him to think, “hey, that’s cool about Otto Porter.” But we also want to use just enough of the language to signal to people like the folks at the Sloan Conference, that there was a thorough analysis behind it. It’s really, really hard. It’s a different kind of challenge. I mean we do tremendous amounts of work and then you just have to boil it down.

Like we did a piece back in December about who was going to make the All-Star Game, based on what their stats were through then. So we did a ton of modeling, all this cross-validation and stuff, and we spent a ton of time doing the research on that piece and then it comes down to — John Wall has an 80 percent chance of making the All-Star Game.We’re not trying to tell a chronological tale of how we came to this insight. It’s important to remember that it’s about the story, it’s not about the process.