The F5 Q&A With Dean Oliver
Lessons learned from a life on the analytics frontier
The only good thing that’s happened to my timeline over the past few months is Dean Oliver’s return to the public sphere. Oliver is the proverbial godfather of NBA analytics and has been sharing his thoughts and analysis on Twitter since leaving his job with the Washington Wizards at the end of last season.
Now employed by ESPN, Oliver also has a new book out. It’s called Basketball Beyond Paper and it traces his unique path to and around the NBA as well as the lessons he’s learned from working directly with coaches and players. It’s a sequel to his first book, Basketball On Paper, which has become something of a Rosetta Stone for many people who are getting into NBA analytics.
For a book ostensibly about stats and numbers, Basketball Beyond Paper has a uniquely personal touch to it. The NBA is a business built on relationships and Oliver’s book reminds readers how much easier it is to get buy-in when you have colleagues that both like and trust you.
I emailed Oliver about his new book to discuss what he thinks would happen if every analytics department disappeared over night, and his NBA player comp.
This Q&A has been edited for length and clarity.
F5: Coaches can be unwilling to talk about analytics concepts with players. I think the reason behind it is to not overwhelm them, but you've had success (most notably with Kristaps Porzingis) sharing these ideas with players. Do you think there's more room for coaches to incorporate analytics concepts in game plans with players?
DO: I think players are actually pretty familiar with the most basic analytical things - shooting 3s over mid-range shots, effective field goal percentage. Players have grown up with those basics. Players always want to be better and if they have data on what it is that they can do better and how to do it better, they will be receptive.
One of the tough things with an 82-game schedule is keeping on-message without saying the same thing and sounding robotic. Game plans can get robotic. Analytics can get robotic. Sure, a different opponent has different players to prepare for, but you don't want to lose your own identity by over-preparing for who the other guys are. I think analytics can help with that.
I think there is a lot of room to bring more analytics than just shoot 3s, not mid-range. It requires coordination across the coaching staff and management. That can take some time, for sure.
I loved the chapter of your book on game planning and scouting an opponent. You had a section of that chapter discussing whether it was better to make LeBron James a passer or scorer when playing against him. What about Nikola Jokic?
I'm glad to hear that different people like different chapters of the book. That was a tough chapter to write. My editor trashed it in its original form. So I went back to the drawing board and it came out better.
To your question - in the first few years of Jokic's career, you did want Jokic to be a scorer, but he, as a lot of great ones do, adapted to the point that it is really balanced. There is no single key to stopping him. It's probably a mixed strategy to show him constantly different looks, but that requires all your defenders to be really tied together, which doesn't happen out of the gate.
In the chapter "Make 'Em Better" you talked about how much time teams spend practicing things that may not even work. For example, do shooting drills without defenders make players better shooters? I think the same could be said for when players hire personal trainers in the offseason. Have any personal trainers ever contacted you and asked for evidence-based ideas to improve their clients game?
We had players who brought their trainers in and I had conversations with them. I think they are intrigued and I think it would really help. Porzingis had his personal trainer who was a big fan of analytics. That helped him a lot, not just me.
I do believe that basketball is a game of quick decisions that players have to commit to. If you question yourself when you shoot, you'll miss. You have to be comfortable during games with your decisions, but you have to practice those decisions a lot to get good at them. I am working on a Sloan presentation with Ben Alamar to illustrate how long it takes for players to get "familiar" with each other so that they shoot better. Part of that is realistic drills.
A lot of time is spent thinking analytically about how players perform. Less time is devoted to thinking analytically about how coaches perform. What are some good ways to tell from the outside whether a coach is doing a good (or bad) job?
What coaches are really supposed to do is help their players be successful. That means that we should be able to see bumps in performance for players fairly broadly under a coaching staff. We don't know for sure that it's the head coach or an assistant, that's for sure. And with organizations that have good communication between management and the coaching staff, it could involve management, too.
That makes it particularly hard to tell from the outside whether a head coach is actually making a difference. I mention in the book that the process of identifying a coach isn't just applying a formula. The formulas won't capture coach performance without a lot of noise in that signal. I think that, in hiring, you can use analytical concepts to hire a coach. You can see how consistent their thinking is with what we know about the game and you can see how and how well they convey it to different people. A coach fundamentally has to interact - not just preach - with players and staff. Setting up that environment where everyone is comfortable talking openly about the game is big.
So, "from the outside", I think it's really tough to know whether they're doing a good or bad job. I hear people complaining about body language or facial expression of coaches, but people are bad at interpreting those. I hear people complain about coaches putting players in or taking them out - there is often a lot that goes into those decisions, so I don't think that's a great surrogate. I do think that a good coach does have a solid idea of who his best players are and he plays them generally more. But that's not universally true as some teams are, as we know, trying to lose.
Which analytics concepts have you found that coaches are least receptive to?
Open vs contested shots was a major source of conflict. What makes a shot open or contested led to lots of discussion, some productive, some not. Identifying what hand/arm position actually matters against a shooter is still tough, though the new data is there to do it better. Without that, the NBA had subjective tracking of contested shots that was at least reasonably consistent (it's hard to be unbiased). Second Spectrum had a truly unbiased algorithm for labeling contested shots, but it is only so accurate. So even unbiased data led to disagreements within the coaching ranks.
I once was talking to an assistant coach and he jokingly said that every NBA analytics department could disappear overnight and not a single coaching staff would care. He was exaggerating, but what do you think would be the biggest change we'd see on the court if teams didn't have analytics departments going forward?
That's a tough question, one I've never heard before or thought about. I think it really depends on the staff. One of the reasons we have analytics is to try to understand what makes the best coaches so good. When I worked with George Karl in Denver, he was instinctive, but his instinct was often consistent with analytics. He played Sam Perkins as a shooter well before shooting bigs came about. So if you took away analytics from him, he'd still be very good. If you took it away from less instinctive coaches, I suppose they would fail more. I suppose analytics can help level the playing field across coaches, but I'm not sure it's doing that, as it then comes down to how well coaches build their instinct from analytics.
The NBA season is long and intense for people on the team side. Did you have a favorite way to destress on a day off when things got overwhelming?
Day off? A full day? I don't remember that. I'd take the couple of hours off that I could get to go for a walk in a city on the road. Or pick up a scooter and go around the town. I had friends in a lot of NBA cities, so I would try to see them if there was time. I tried to bike to work and games when I was at home and that had a lot of power to de-stress.
I'm not sure how much pick-up you still play, but when you were most active, who was your NBA player comp?
My ankles have been broken, my adductor has been torn - I don't play anymore, despite how much I miss it. I couldn't play very well the last decade anyway as my skills have deteriorated. I was a defensive-oriented point guard who attacked the rim at my best. I have no natural depth-perception (I lost my sight in one eye when I was ten), so I never could shoot well. But I had a feel for where offensive and defensive players were. I wanted to be Magic Johnson. I ended up more like Kris Dunn, I guess. Poor-man's Kris Dunn.
What's one thing you can't live without during the NBA season?
Frankly, I always had my own custom box scores that incorporated a lot of what I cared about. I've had them for 20 years, putting new things in and modifying parts here and there. At ESPNAnalytics.com, we're actually rolling out a basic version of my box scores this coming week. It will have information on player value, how good the starters vs bench were, defensive information, the Four Factors, and so forth. We'll grow it from this point and make it better, but it's pretty good now. For me, that's my cheat sheet.