Daryl Morey loves good data, and lots of it. As general manager of the Houston Rockets, the Northwestern graduate has made a name for himself with his devotion to using data analytics to make team decisions—everything from where to shoot from on the floor to whom to acquire in a mid-season trade. Morey talks with Kellogg Insight about the importance of assembling a staff that understands analytics, how to ensure you are using the data wisely, and the need to always keep your eye on the prize when crunching the numbers.
To hear Morey discuss how the Rockets have incorporated data analytics into their organization, check out this month’s Insight In Person podcast.
(Editor’s Note: This interview has been edited for length and clarity. Special thanks to Kellogg School faculty members Thomas Hubbard, Keith Murnighan, and Ned Smith for their assistance with the interview.)
Kellogg Insight: When collecting data, do you know the questions you want to ask beforehand, or do the questions arise from the data you’re able to collect?
Daryl Morey: For us the questions are very simple. Everything is judged on the probability of winning a championship over a three to five year time horizon. If the data we gather or a decision we make can affect that, we’re going to do it. For us, the success function is pretty easy to figure out. It’s unfortunately a very daunting equation because your odds are pretty terrible in a league of 30 where only one wins every year.
KI: At Kellogg, we have a new program on data analytics. We’ve adopted the perspective that data analytics is a leadership problem, not a statistical problem. What are the key challenges business leaders face when integrating data analytics into the rest of the organization?
DM: At the Rockets, we have an owner who is a visionary guy, and has been his whole career. He absolutely believes in the value of data analysis to help drive decision making. He’s seen it work in his other businesses and he was the pioneer in basketball to say, “Hey, I’m going to go make a full commitment here.”
At other teams, I hear a lot of frustration. The decision makers will go in other directions, and often in ways that don’t work, because they are not versed in using information.
“Everything is judged on the probability of winning a championship over a three to five year time horizon.”
KI: How do you create that comfort in an organization?
DM: I think most of it really comes down to what you hire for and what you reward. We want to make sure people understand the value of information. You don’t always have to use data to help drive a decision, but you do always have to see if you can do that. I live that, embody that, and we hire for that. The people who move forward are the ones who make the best decisions.
KI: With respect to data analytics, what insights from other industries cross over into your purview?
DM: Sports are really a late adopter of using data to drive decision making. If you look at Wall Street, or you look at consumer or credit card companies, or you look at Procter & Gamble, all of these are actually quite a bit ahead in terms of using data to drive their decisions. Sports are late to the party. We have a bunch of contacts working at quant funds. They are dealing with very similar data sets to ours, in that they have data that changes pretty rapidly through time. We’re trying to forecast players; they are trying to forecast companies.
KI: How can something as intangible as style of play be captured in analytics?
DM: Well it turns out it is not very intangible. You can see it on the floor. The most advanced data that’s out there—25-frames-per-second positional data of all the players and the referees on the floor—that data is very granular but very rich. If you want to look at style of play, i.e., how quickly the players move, how much they get up the floor, how often they are spaced, how often they are clumped—that kind of data can really glean the different styles of play in the NBA, both at the team level and at the individual level. If you want to look at all the derivatives of movement in terms of position, velocity, acceleration, jerk—you can get all those things with the data that we now have.
“Everyone’s using data even if you don’t know it. It’s the information from the world that you’re processing in your own experience and the mental models you form around that.”
KI: Obviously in basketball there are certain players who work well together. How do you use the analytics to determine which types of players are like that?
DM: That’s one of the tricky things. Just like anything else, we have some questions that are fairly easy to answer, like style of play, and questions that are very hard to answer.
Baseball is a little bit easier. There is much less interaction. It’s pretty much a solitary pitcher-batter interaction, with some impact of the umpires, the catcher, and the fielding positions.
Whereas in basketball, just take one jump shot: Does the jump shot go in because the guy can shoot well, because the pick was set well, because the defense was bad, because the pass was good? There are a hundred things that go into just one shot, and that’s one of the simple things to look at. To try to completely break down all the interactions between players is extraordinarily difficult and one of the things that we constantly try to improve. If we had games where the outcome didn’t matter to my livelihood and others’, and you could say, “Hey, this game we’re going to try this with different mixes of people,” that would probably be the only way to really isolate some of those effects.
KI: Coaches have long watched film, and showed the players what they were doing and what they could do better. Does the quantitative aspect of this data help the coaching process?
DM: Yes, absolutely. I like that you point out that video information and coaching experience are just another form of data. Everyone’s using data even if you don’t know it. It’s the information from the world that you’re processing in your own experience and the mental models you form around that.
So how do you separate the signal from the noise? How can you know if you’re dealing with something spurious or a real trend that you need to deal with? A guy is hot in the first half—is he hot because he’s picking really good shots? Do we need to close him down? Or did we actually follow the game plan well and we were just not lucky in the first half and we should just stick to our game plan? Those kinds of decisions, that’s where data can help and does help in the coaching ranks.
KI: Do you try to use data in real time during the game?
DM: Yes. If you’ve ever been to a game, there’s an army of people helping. A smart coach like Coach [Kevin] McHale will use all the information available. If a guy is killing us, it’s definitely discussed. Usually the head coach will make a decision on how we might adjust things at half time.
KI: In your career, who has been the hardest audience to sell analytics to?
DM: In baseball, the analyses showed that people were doing things incorrectly for many, many years. In basketball, we were more fortunate, but certain players and coaches did believe that some time-honored things were more correct than some of the easy data showed.
Trying to convince Coach [Jeff] Van Gundy, who is an analytically smart guy, that the “2 for 1”—where you take two quick shots at the end of a quarter [instead of going] for one good shot—was better was a bit of a challenge. Coaches historically had wanted to go for one good shot. It turns out that two quick bad shots are definitely better than one good shot, and the smarter coaches have moved to that.
Coach Van Gundy, over time, became convinced that it was the right thing to do, even though he didn’t always implement it, because he felt like the difference in winning or losing wasn’t big enough. I think he’s right about that.
KI: Did “Moneyball” make your life any easier?
DM: Yes. I think our owner probably would have come to hiring me or someone like me without that book, but I do think that book helped spur that. It’s helped the success of some of the ideas, especially in baseball, and then now more recently in basketball, and over time probably in football and hockey.
KI: What’s something that you realize now because of data analytics that you and others in your position didn’t know 10 years ago?
DM: The shot selection of a team has fairly steadily, fairly dramatically shifted over time into these zones: shooting near the basket, and shooting at the three-point line. [The benefit of this] is one thing that I think was known but maybe not recognized. People certainly didn’t make very major investment decisions against it. There is a big difference between “I believe something” and “I believe something; I’m willing to put a lot of money behind it, and a lot of investment behind it, and my future career behind it.”
How Diverse Should Your Team Be?
NBA data point to a sweet spot in balancing diversity and similarity of skills.