When the Philadelphia 76ers called the Annenberg School for Communication’s Damon Centola for input into scouting practices, he recommended they look beyond the highest-ranking coaches to the periphery. Incorporating this additional knowledge increased the team’s ability to predict three-point shooting skills from 57% to 66%, a case study described in his new book, “Change: How to Make Big Things Happen,” and excerpted below.
“Change,” which uses cases studies like this to illustrate what drives behavior change, was published in January by Little, Brown Spark. The excerpt is reprinted with permission.
A few years ago, I received a phone call from the director of performance research and development at the Philadelphia 76ers. He had seen some of my research on social networks and wondered whether my work might be helpful for the NBA.
The problem, he explained, was scouting.
If you’ve read “Moneyball” by Michael Lewis (or seen the film starring Brad Pitt and Jonah Hill), you’ll immediately understand the challenge: Professional sports scouting has long been an old boys’ club. Most professional scouts are former players or managers. They have well-established biases regarding how to evaluate players. Longstanding norms in scouting can privilege certain kinds of players (who tend to fail) and ignore other kinds of players (who may succeed).
“Moneyball” describes how the Oakland Athletics baseball team threw out the venerated norms of scouting and devised an entirely new way of building a team roster. The new scouting strategy led the A’s to break the record for the longest winning streak in American League history.
The Philadelphia 76ers wanted to know, “Can we do that too?”
Scouting lore
In the NBA, there are some fairly famous stories of professional scouting gone horribly awry. In 2011, the very last pick in the NBA draft—literally, the last person to make it onto a team—was Isaiah Thomas. That name sounds familiar because he was named after the 1980s-era Detroit Pistons Hall of Famer, Isiah Thomas.
This Isaiah Thomas—the 5-foot-9, 2011 late-stage draft pick—was not a major college star like his eponym. In fact, he had been lucky to get a spot on the Sacramento Kings’ roster. Many thought he would disappear soon after. But he rose through the NBA ranks, ultimately becoming an all-star in both the 2016 and 2017 seasons and winning the prestigious All-NBA Team honor for the 2016/17 season.
By contrast, in 2013 the very first pick in the NBA draft—the highly coveted “number one overall” pick—was Anthony Bennett. At 6-foot-8, the UNLV power forward invited comparisons with basketball great Larry Bird. Bennett was poised to be an all-star. But four years later, in 2017, as Isaiah Thomas was making his second appearance on the NBA All-Star team, a series of disappointing seasons for Bennett led him to the minor leagues.
In 2017 Bennett played for the Maine Red Claws, and in 2018, was traded to the Agua Caliente Clippers of Ontario—teams most of us have never heard of.
When the Philadelphia 76ers called, the team had an NBA championship in mind. They wanted to know how to improve scouting procedures to help identify the unlikely Isaiah Thomases of the world and avoid the unfortunate Anthony Bennetts.
Even before the phone call ended, I knew what they had to do. I just didn’t know whether they would be willing to do it.
Untapped knowledge
At the time, the Sixers already had a large staff of data scientists analyzing everything about the players, from total number of seconds played to total distance traveled to data showing players’ postures and body language during games. With all of those data points at their disposal, they thought an algorithm to bring them success must be in reach—the needle in the haystack that would lead them to victory.
My approach was different.
Although data science is an essential part of the puzzle, there is also tacit human knowledge never included in the data analysis, mostly because it’s hard to know which bits of knowledge matter and which don’t. If the right bits are never recorded, they can never make it into the algorithms.
I was interested in the hidden insights that might lie within the human social networks among the Sixers’ staff. Was there untapped knowledge in the “outer rim” of coaches that could be used to improve their scouting?
The main challenge was that organizational networks in professional sports are highly centralized. Just like successful managers, politicians, and physicians, coaches work in a hierarchical world. Some coaching staff members are more powerful than others. Influence flows from the people at the center of the network (like the head coach or general manager) to everyone else. My goal was to see whether changing the pattern of these networks might lead to better predictions about player performance.
You may wonder whether this is really possible. For a professional sports team, hundreds of millions of dollars are at stake each year. The chain of command is difficult to disrupt. While the President of the United States can be intentional about bringing in diverse voices from the network periphery, what could a sociologist possibly do to make the Sixers’ network more egalitarian?
My idea was to turn the coaches’ scouting problem into a quiz game.
The quiz
My research team and I developed a simple application so that when the coaches logged in, using either their phones or laptops, they would be connected in a fishing-net pattern. They would then answer questions about the performance of draft prospects the Sixers were actively considering.
Scouting season was already underway and the Sixers had started to look over their top prospects for the upcoming draft. For the duration of this study, I was sworn to secrecy. Any leaks by me or my research team could result in the Sixers’ draft prospects receiving additional media attention that might draw another team’s interest during the draft.
For several weeks, several times a week, the Sixers flew in top prospects to the training center. They would run the players through a series of drills, including brief two-on-two and three-on-three games, free-throw shooting, sprints, three-point shooting, and so forth. There was intense interest in identifying the best “shooters.”
Each day, either I or one of my graduate students would arrive at the Sixers’ training facility in Camden, New Jersey, typically in late morning. The training sessions and drills would already be underway. Once we had set up our materials, our team contact would ping the coaches to let them know it was time to join the study. At that point, just one additional drill—the three-point shooting drill—remained for the day.
We ran a total of five of these studies on five different days. Each one worked the same way. Following an alert, each coach would log on to the site and see the profiles of that day’s prospective recruits. The quiz asked them to enter their predictions, based on everything they had seen so far that day, for each player’s three-point shooting percentage in the upcoming drill.
After the initial forecasts, the coaches could see the anonymous predictions made by the other coaches to whom they were connected in the network. They could either ignore that information and stick with their first instinct or use their colleagues’ opinions to revise their guess. They would then submit their final response.
That was it.
Each quiz took about 10 minutes. Then the coaches would get back to work. A few hours later, the players would complete the three-point shooting drill and we could test the coaches’ predictions. During the drills, coaches watched several players at a time. They could readily evaluate shooting form but were unaware of accuracy. Just like everyone else, the coaches had to wait for the study results.
A sense of empowerment
The first week’s study didn’t go over well.
Most of the coaches were indifferent. Some were genuinely annoyed; their comments were pretty much exactly what you would expect. Lots of jokes.
But after the first week, attitudes improved dramatically. In fact, the coaches wanted to participate. A few things had happened that prompted this change of heart.
First, the coaches realized that the quiz was kind of fun. Second, coaches are naturally competitive. Once they understood the idea of the quiz, and that they could do better or worse than their peers, they were more motivated.
But the main reason that the coaches were more engaged was an unexpected byproduct of the fishing-net pattern: The coaches realized their voices were being heard.
The coaches had initially assumed that the higher-ranked coaching staff would dominate the interactions in the quiz game. They didn’t know about the egalitarian networks I was using to connect them. After the first session, some of the lower-ranked coaches (from the “outer rim”) saw they could exert genuine influence over the group. They felt a sense of empowerment.
I hadn’t noticed this idea of empowerment in my previous studies, probably because I had never been able to talk face to face with the participants before. I also did not expect it because empowerment seems like an odd concern for a group of 6-foot-5 ex-athletes. But evidently some of the coaches had been feeling that their voices were not always heard.
Talking with them afterward, a few mentioned to me that the quiz was satisfying because they could see their own influence affecting the group’s decision, moving it toward a better answer. But most notable, everyone I talked to was glad that the group opinion was not dominated by the same senior individuals who typically influenced meetings. This fact, more than anything, helped to create buy-in among the coaches during the remaining weeks of the study.
Seeing the results
Once we tallied up the data from all five sessions, the results were striking. In only 10 minutes, the coaches’ ability to accurately predict a player’s three-point shooting significantly improved, from 57% to 66% accuracy.
The coaches thought the findings were interesting, but it was the Sixers’ management who really sat up and paid attention. The experiment gave them new insights into how the network periphery among coaches and support staff might improve not just scouting decisions but judgments about how much playing time athletes should get, as well as decisions about how long practices should last, and how much recovery time athletes should have between workouts. The network periphery holds a good deal of tacit knowledge, and an egalitarian network offered a new way to collect and use it.
Biases are strange things. They make us more likely to choose answers that are familiar rather than correct, even when those mistakes are costly. Centralized networks tend to reinforce these bad habits of thought. Once biases are established, ideas that resonate with them become simple contagions. They are easy to understand and easy to spread.
The real problem is that our biases and the networks that reinforce them can prevent us from finding new ways of solving hard problems. They can even keep us from seeing clearly the information that is right in front of us.
Thankfully, the network periphery can, and does, support real social change. In 2001, the Oakland A’s were the second poorest team in Major League Baseball, using an oddball strategy to try to gain an edge. Nobody thought it would work. Today, this oddball idea has been adopted by every Major League franchise. There has been a sea change in the social norms of Major League scouting. And it spread from the periphery.
Damon Centola is a professor of communication, sociology, and engineering at the Annenberg School for Communication at the University of Pennsylvania, where he is director of the Network Dynamics Group. He is also a senior fellow at Penn’s Leonard Davis Institute of Health Economics. The text above is excerpted from “Change: How to Make Big Things Happen” (Little, Brown Spark, 2021). ©Damon Centola. Reprinted with permission.
