Among a sea of statistics geeks, sports fans, reporters, and students, President Barack Obama walks on stage to Marvin Gaye’s 1971 megahit ‘What’s Goin’ On’. The crowd goes wild. The 2018 MIT Sloan Sports Analytics Conference, the annual gathering of some of the brightest minds in sports, is already a resounding success. And its keynote speaker doesn’t even work in sports.
“Data is a tool, but doesn’t tell you what’s important,” says the 44th President. “It doesn’t tell you what your priorities should be.”
Deemphasizing the role of data is not exactly what you would expect to hear from the keynote speaker at a conference whose theme is “Talk Data to Me”. Seriously, everyone at this conference is obsessed with data, and for good reason. In basketball and the sports world as a whole, three-dimensional player capture has completely revolutionized the kinds of data teams are able to employ when evaluating talent. Advanced statistics have power to give people insights into things even the smartest basketball minds wouldn’t have dreamed of being able to quantify decades ago. And teams are making good use of it.
“Data gets used in a lot of different places,” said Clippers Owner Steve Ballmer. Teams utilize data in tracking player health, preparing gameplans, analyzing the business side of the organization, optimizing arenas operations, salary cap planning, and so on. The things people are able to construct and uncover with large amounts of data are monumental.
Over the course of the two days of the conference, speakers discussed all possible kinds of data, whether it pertains to basketball operations, business activities, or fan engagement. There were presentations about using data to retain season ticket holders, predict defensive movement, and analyzing fans’ responses to certain game actions.
However, the validity of certain data is still up in the air. Daryl Morey, founder of the Sloan Conference and GM of the Houston Rockets, discussed the legitimacy of ESPN’s Real Plus Minus (RPM) stat. Morey said that although a team performs better when the players with a high RPM are on the floor, it doesn’t necessarily mean that the player with a high RPM is contributing anything groundbreaking to their teams’ success. As Morey pointed out, “that player could be very replaceable by multiple players with that same skill set.” Regardless, teams are striving to find out as much as they can about as many things as they can, collecting mountains of data trying to dissect the most minute of details.
So what’s the biggest threat to data? Bias.
Biases distort our vision of reality and make evaluating talent objectively impossible. Among the specific types of biases targeted include winning bias, which implies that teams who won must inherently be doing something right solely because of the outcome, and confirmation bias, which is the tendency to view new information as supporting your preexisting beliefs. It’s not just front office executives that are working hard to eliminate bias. Steve Nash emphasized how important it is to “look back on our preconceived notions” and evaluate how we can see the game from a neutral lens. Nash noted that he always had the preconceived notion that point guards weren’t supposed to take the bulk of shots on offense. This held him back, as he was overly hesitant to shoot the ball throughout the early stages of his career.
One way to eliminate bias, as the President pointed out, is to account for a wide diversity of opinions when making decisions. If you can have the “confidence to want as many viewpoints as possible around the table,” Obama said, “you will have better outcomes.” This can be applied to anything from organizational decision making to player evaluation.
So how do you evaluate a player? Lots and lots and lots of data.
Not only do medical, psychological, and background evaluations factor into a team’s decision of who to draft, but the team situation a player is uniquely in is always taken into account, especially at the college level. Austin Ainge, Director of Player Personnel of the Celtics, noted that when Jayson Tatum was at Duke, he was an isolation midrange scorer who shot a low percentage from three. But Coach K wasn’t concerned about showing off all of Tatum’s talents, he was concerned with winning, as all coaches are. After hitting a high percentage of three pointers in a closed workout, and by evaluating his performance data independent from his team situation, they arrived at the conclusion to trade down from the #1 pick in order to acquire more assets, and then draft him at #3.
“We’re trying to take him out of the context he’s been in and try to apply him to ours and see how it will work” said Ainge. So far it has worked pretty well, as Tatum is in the conversation for Rookie of the Year and the Celtics are second in the Eastern Conference. His style of play is drastically different at the professional level, as he’s found a lot more confidence driving to the basket.
The most important thing to do when evaluating talent is to have perspective. You view things on what Steve Ballmer called a “wide time horizon” and see the game as constantly evolving.
Jonathan Givony, founder of DraftExpress, noted that when Jahlil Okafor entered the national recruiting scene as a high school freshman, his skillset of being able to post up and command the ball on the block was viewed as a crucial asset. Five years later, In 2015, The 76ers pounced on him, drafting him third overall, but by then the game had completely changed. The league no longer needed big men to be able to score in the post, and that trend has only continued, as many centers today take more three pointers than they do low post hooks.
The game is evolving on both sides of the ball. Defenses are evolving to foster more switchability (the ability to swap defensive assignments on screens) as a means of combating deadly three point shooting teams that frequently utilize high pick and rolls. Just years ago, as Rockets’ General Counsel Rafael Stone noted, PJ Tucker was out of the league because he “didn’t have a position”. Now, Tucker’s ability to guard multiple positions is what makes him so valuable, and is a pretty large reason why the Houston Rockets signed him to a contract worth more than $31 Million after being out of the NBA for four years. Stone also says he believes we’ll see a “resurgence of defensive coaches [and] defensive minds” as a result of the offensive explosion the league has seen in recent years.
PJ Tucker (6’5″) guarding Marc Gasol (7’1″). Image courtesy of the Houston Chronicle.
Despite this massive gathering showing such support for analytical approaches to team building, or maybe because of it, I still got the impression that the teams that utilize these practices are a little scared. After all, as their approach has become more popular, the competitive advantage has worn off. There is a big difference between there being ten dumb teams in the league versus there being two dumb teams in the league, and that seemed to be the biggest underlying point of concern among general managers from traditionally analytics-oriented teams. However, as the game continues to evolve, executives are confident that the NBA is “winning the war for talent,” as Daryl Morey put it. Top players in the league now include guys from Cameroon, Latvia, and Greece, as the league has become increasingly international in recent years and the sport of basketball has nearly secured its spot as the second most globally ubiquitous sport. This talent explosion couldn’t be happening at a better time for the league, as cutting edge technology has completely revolutionized the game, and it’s about to get a lot more intense.
Among the neat gadgets unveiled during the conference was a high-resolution shot capture system that showed not only whether a shot is made or missed, but precisely where the ball crossed the plane of the basket or where it missed. Using this information allows individuals to tell where players habitually misses shots to one side of the rim and allows teams to lay the groundwork for correcting their form.
Player tracking data is the most coveted and the most versatile type of data, and is likely where most of the next advancements in the field of basketball analytics will emanate from. This innovation has been huge, as stats emanating from player tracking data become more commonly used and accepted. Former President of the Los Angeles Lakers Mitch Kupchak stated that the introduction of GPS data has “changed the way our sport looks at analytics,” but teams are looking to do even more with this information.
During a presentation dubbed ‘Bhostgusters’, my mind was promptly blown when a group of PHD students and research associates introduced a system of synthesized NBA defenses paired with a real-time sketching system. Bhostgusters uses 30,000 possessions from the 2016-17 NBA season to train what they call ‘ghosts’. Ghosts basically act as the computer-generated defense in NBA 2K, predicting where players will go and how they will react to offensive action. What makes it special though is the staggering amount of factors that are taken into account to determine a defender’s next action. Everything from those 30,000 NBA possessions is recorded; the amount of time a player’s been on the court, the game clock, the shot clock, and the game situation. Then, by using a deep-learning system that I am not even going to pretend to be able to understand, the algorithm is able to predict defensive responses to offensive actions. The user interface system allows a coach to draw up a play (on an iPad) in a specific game situation and see how the defense is likely to react based on data accumulated over the course of the season and over the course of the game. Instead of viewing the play as a one dimensional image of basketball action, coaches can view an animated simulation featuring ghost defenses in their custom play, in order to see how the defense will handle the offensive actions. Even if the model has never seen the play before, it can still come up with “context-dependent behaviors” to predict defensive movement. This incredible presentation was only one of dozens of projects from graduate students outlining technological advances and statistic models.
While the ‘Bhostgusters’ system itself may or may not ever find its way onto an NBA sideline, deep imitation learning, the foundation of the predictive model, is already changing the way teams utilize data. The days of basic linear regression are almost over, as organizations now strive to use deep learning on spatial data in order to gain a competitive advantage. The holy grail is a system that is able to take real-time data, apply an algorithm, and spit out data in real-time to help teams manage games. Some of the stuff seems unusual and niche right now, but in the words of the great Sam Hinkie: “What’s novel today is a building block for what you build tomorrow.”
Tomorrow certainly seems exciting from an NBA fan’s perspective. Multiple panelists, including Wizards owner Ted Leonsis, discussed the very real prospect of in-game betting making its way into professional sports in the next decade. Additionally, the league continues to work hard on eliminating dead time from timeouts, free throws, and replay reviews, which currently amount to around 20 minutes per game.
When it comes to the structure of the game itself, there could be far bigger changes than the elimination of conference playoff seeding (which was a big topic of discussion this weekend, but I’m going to save it for another post). Among the ideas floated around was the pitch to change the free throw system to be all-or-nothing; if a player gets fouled on a two or three pointer he just takes one shot worth two or three points. The idea of switching the draft and free agency seemed to gain a lot of support from front office guys, as this would allow teams more flexibility to make exciting moves on draft night without worrying as much about salary cap ramifications. However, there will always be resistance to change no matter how incremental, and what NBA Senior Vice President of Strategy and Analytics Evan Wasch called the “momentum of the status quo” will always play a factor in keeping the game the way it is.
The Sloan Conference shows both sides of the basketball analytics coin. The massive turnout, fanfare, and enthusiasm around the field demonstrates that the science that was once ridiculed has gained legitimacy among the sport’s most influential figures. However, at the same time, the experts in their respective fields understand that the competitive advantage that arose from the advent of analytics is silently dwindling. Every new general manager who takes an analytical approach, while undoubtedly being inspired by analytics trailblazers like Hinkie and Morey, decreases the competitive advantage by elevating competition and replacing a regime that did not utilize data most effectively.
“The most important thing we do in sports is to mimic people who have been successful,” said former Cavaliers GM David Griffin. That’s just human nature. Everyone at the conference knows that data leads to good decisions, the challenge for the next 10 years will be evaluating which methods work and which ones don’t.
Former Cavaliers GM David Griffin. Image courtesy of Associated Press.
Perhaps the most strangely uplifting moment of the conference was when President Obama discussed the disparity in information consumed in today’s society.
“Essentially we now have entirely different realities that are being created with not just different opinions but now with different facts.”
This is inarguable, as in 2018 it can quite often feel like we’re living in separate worlds, dealing with different sets of truths. Obama said it would be “very difficult” for society to function over the long run with that disparity in reality, and it’s hard to disagree with him. Hearing hundreds of extremely intelligent people all talking about using facts and tangible information to make decisions should not have been as reassuring as it was. But our society too often deemphasizes objective accuracy in favor of a number of biases just like the ones NBA front offices are working hard to combat. Maybe the NBA front offices’ commitment to finding out what is objectively true while eliminating our preconceived notions that cloud our judgement is exactly what the world needs, and maybe it’s where the world is going. After all, Obama did say:
“What’s true in sports is true more broadly”
I sure hope he’s right.
Jacob Mooallem is a student manager for the Indiana University Men’s Basketball team. He spends too much of his time on Twitter.