Hoop Math from the Sideline with Dean Oliver
By Tim Chartier, Davidson College
Tim Chartier
Dean Oliver is a pioneer in sports analytics. In 2002, Oliver wrote Basketball on Paper which became the handbook for how to do basketball analytics, and he became the first analytics consultant in the NBA. Oliver later joined ESPN and guided the sports analytics group in developing football and basketball metrics. Oliver received a B.S. in Engineering and Applied Science from Caltech where he played basketball for two years and then joined the coaching staff. He received a doctorate from the University of North Carolina at Chapel Hill in Environmental Science and Engineering. Oliver is now an assistant coach in the NBA for the Washington Wizards.
Tim Chartier: Your book, Basketball on Paper, started the basketball analytics movement. How did you develop your groundbreaking basketball analytics? In a sense, where did it all start?
Dean Oliver
Dean Oliver: Basketball on Paper wasn't the first basketball analytics book, but it did lay out a structure for doing the work more than the books that came before it. It started with being able to count possessions, then it filled in around that by looking at basketball questions. My inspiration was Bill James, who stated that he looked at baseball questions, issues that came up in games, big decisions that fans and commentators disagreed about. After trying in vain to just bring his methods over to basketball, I took that philosophical approach to basketball and left the methods. So I tried to answer questions and still do. If you watch a game and you hear people saying, "He should have fouled instead of letting him shoot the layup" —how do they know that? If they say, "that was his man to cover in that situation" (which came up with Rudy Gobert and Bojan Bogdanovic when they lost their series)—how do they know that and how can you check it? Taking a logical approach to those questions, even if not all the data is available, is what I encourage people to do and is how it started for me. In Basketball on Paper, I took a large series of questions and answered them in a logical step-by-step approach that led me and readers down what I think is a good path.
Tim Chartier: You continue to look, explore, and research the game. What types of analytics and research questions have developed beyond those in Basketball on Paper?
Dean Oliver: The research questions have definitely evolved since 2002 when I wrote Basketball on Paper. At that time, there were boxscores and the beginning of fairly accessible play-by-play information. If a boxscore shows us results, the play-by-play gets at some of the process towards those results, then all the additional data we've collected since then—manually or computer-aided tracking of details—is getting at processes within the processes. None of this changes the possession-based efficiency or the Four Factors. (Dean Oliver identified what he called the "Four Factors of Basketball Success," which are effective field goal percentage, turnover percentage, offensive rebounding percentage, and free throw rate. Oliver also assigned a weight to approximate each factor’s contribution to success.) It just fills in some of the how and why. How much of it is transition vs halfcourt? How much is pick/roll impacting our turnovers and shooting percentage? How much does a player's shooting decline when he's tired? The new data can help answer some of these questions.
Tim Chartier: You sit on the bench for the Washington Wizards and are much more than a data analyst. You are an NBA basketball coach. How does analytics play a role in your coaching?
Dean Oliver: The process of incorporating analytics into coaching is definitely challenging. What I like to say is that analytics is great at identifying symptoms of problems, but often needs to be incorporated with traditional coaching wisdom to actually identify the cause and the cure. Two players who don't work well in pick/roll—the data can say that and it can say that it's maybe just high pick/roll, and maybe it's just turnovers from the ball handler. The data can hone in on causes, but sometimes the cause is that a player has a hard time reading the defense going to his left—something that is harder to see in data but can be picked up by being around the player. There are similar challenges with player development, where metrics can show that Player X needs to get better at defending on the perimeter, for instance, but that can affect their ability to get defensive rebounds—so there are trade-offs where coaching has emphasized one, but not the other. Data has a hard time capturing what coaches have guided players to do (both in the past and the present).
Tim Chartier: You are a statistician and a coach, how do those inform how you watch a game?
Dean Oliver: How I watch a game has been dramatically changed by analytical approaches. I know generally what the analytics say accurately and what they don't. I watch a game for things that they don't see well (or don't see easily)—positioning in defense, especially hands and arms, screen effectiveness and angles, for example. It's not that there isn't data for these, but either the data is less trustworthy or it's just data, not the interpretation of that data into something we fully understand. Zone tagging and pick/roll coverages are examples of less reliable data. The location of players is great, but you have to do work with it to understand whether those positions are where we coached the players to be. For me, I constantly go back and forth between watching the game and studying the numbers to try to derive more information out of each of them.
Tim Chartier: If someone wants to develop and supply analytics for a coach in basketball or another sport, what advice do you have?
Dean Oliver: Know the game like a coach as well as you can. Understand where they have to resort to gut feelings in many situations, but understand where numbers can help in those situations. Recognize that personnel decisions—who should be in the game—are not the only part of what they do. A coach's job is to make every player better and help them play better together—each of those can have analytical input.
Beyond this, appreciate the job that coaches do using instinct. Be able to talk to them using traditional coaching language. If you don't have to, don't make them adapt to the more mathematical language. You should be able to make an analytical case with basketball language. And, really, be a friend in it. Coaching is tough, so they want to feel like you're with them. You may have to say things that go against what they think, but make sure they know you care about them.
Tim Chartier: If someone wants to get into sports analytics, what’s your advice?
Dean Oliver: It's more than knowing data, stats, and programming. You do have to know the sport and the questions where data can help. It's a very hard job because you want to be able to do all of those things, but who has time to know databases, stats/machine learning/AI, programming, and all the nuances of what a coach thinks? On top of that, you need to know what has been done in sports analytics for that sport—which is a lot in baseball, a lot in basketball, a pretty good amount in football and soccer, but perhaps less in other sports. So that means a lot of multidisciplinary studying.
If you can do all of that or even just a subset of that, plan on doing projects with whatever data you can get your hands on. Sometimes, you may have to collect your own data, and I recommend doing so. It's painstaking at times, but it can help you see some of what is involved in having high or low quality data, what is involved in making data more comprehensive or more focused.
Tim Chartier: Data analytics is a huge field. Many colleges and universities are developing or further developing programs. Many students are integrating it into their studies, with some majoring in data analytics. Do you have any advice regarding education in data analytics?
Dean Oliver: Data analytics is a great hammer, but, as they say: when all you have is a hammer, everything looks like a nail. Analytics is more powerful if you know other subjects. I always feel that my engineering background helps. It showed me that there are physical laws for how the world works—thermodynamics, chemistry, biology, physics. If you know those laws, then some of the kinks in the data sort themselves out. Sports does have some basic "laws" as governed by the rules and structure of the game—that's how the Four Factors came about, for example. Business analytics has "laws" as governed by different areas within economics (and probably much more). So I'm saying what I've said above—know the subject matter as well as you can. It will help you understand the data beyond simple numbers, help suggest other data to use, and help sort out correlation from causation better.
Tim Chartier is the 2021-22 Distinguished Visiting Professor for the Public Dissemination of Mathematics for the National Museum of Mathematics. Dr. Chartier also does consulting for the NBA League Office on issues related to game integrity.