NBA roster building is driven by a hybrid of financial concerns, talent evaluation and the almighty "fit." Each team weighs those factors according to its own equation when adding, subtracting and trading players. Looking around the league, you’ll see an incredible number of variations.
Although the combined talent of LeBron James and Dwyane Wade was celebrated last summer, there were plenty of questions about the apparent duplication of skill. On the other end of the spectrum, you’ll find the NBA draft lottery, littered with the remnants of teams that were handicapped by overpaying for less talented players who appeared to fit what was already in place.
TrueHoop at MIT Sloan Sports Analytics Conference
Conventional wisdom has a powerful hold on the idea of player fit. You don’t want two players with the same focused skill set on the floor. Duplication is destruction, unless the talent is enough to overwhelm it. Powerful low-post scorers should always be surrounded by strong outside shooters. These ideas, repeated until they reach immutability, form some of the most basic foundations of team construction.
Today at the MIT Sloan Sports Analytics Conference, two different research papers took a swing at turning the idea of fit and complementary skills into measurable data.
Robert Ayers presented a research paper, Big 2’s and Big 3’s: Analyzing How A Team’s Best Players Complement Each Other. Using statistical profiles, he classified players into categories such as high-scoring, dynamic guards; high-scoring, high-rebounding centers; versatile, 3-point shooting wings, etc. He then looked at the effect different two- and three-player combinations of those categories had on a team’s wins.
Ayers found the two-player combination that had the greatest positive effect on wins was a versatile, 3-point shooting wing with a high-scoring, high-rebounding center. Throw in a high-scoring, high-usage point guard and you have the most effective three-player combination.
In all, four combinations of three different categories had statistically significant positive impacts on a team’s performance. Of those four combinations, three featured a versatile, 3-point shooting wing -- think Paul Pierce. Now take a moment and try to count how many players in the league fit that mold. For this group, we’d be talking about players clustered around a per-game stat line of roughly 16-4-4, shooting about 37 percent from the field. Only nine players in the league this season are approaching that stat line, and only four -- James Harden, LeBron James, Kevin Durant, Luol Deng -- are wings. That particular player type fits very well with many other different combinations of players, but finding one that’s suitably talented to help lead a team is exceedingly rare. Scarcity is often the limiting factor in achieving fit.
A second paper, NBA Chemistry: Positive and Negative Synergies in Basketball, by Philip Maymin, Allan Maymin and Eugene Shen, tackled the same issue with a slightly different statistical strategy. Instead of characterizing players and looking at past results to see how they fit, their method creates skill ratings and projects how players with those different skills would work together. This paper has been circulating for a while, and Kevin Arnovitz wrote about it a few months ago.
The knockout punch in their work is the idea that trading Chris Paul for Deron Williams at the end of 2009-10 would have made both the Hornets and Jazz better teams. Their data points to Paul’s ability to create more turnovers as a boon for the Jazz. In New Orleans, Williams would have had to share the ball less and would have been able to exploit his individual scoring skills. His tenure in New Jersey puts that last assertion in doubt, but the idea is intriguing nonetheless.
The predictive value of their system allowed them to identify some 220 trades around the league that would have been mutually beneficial, essentially identifying how inefficiently player skills are distributed among the 30 NBA teams.
Both statistical models give a soft shove, pushing "fit" toward a home for both objective and subjective analysis. The steady march of basketball analytics continues -- from which players are successful, to which combinations of players are successful, and now to how and why those combinations succeed.