BOSTON — The individual basketball ability of a particular player isn’t too hard to quantify. But what about a player’s ability to complement other players?
Some people at the Sloan Sports Analytics Conference this weekend tried to determine the best way to put a team together. Two research papers regarding NBA chemistry were presented on Friday, each using a different method to figure out what types of players mesh well.
Robert Ayer, an MIT MBA, first presented a paper (click for pdf) called “Big 2’s and Big 3’s: Analyzing How a Team’s Best Players Complement Each Other.” Using data going back to 1977, Ayer grouped players in 14 different categories like “Limited, role-playing centers” and “Wing 3-point shooters.”
Then he determined the best two (or three) players from each team and measured if that team overachieved or underachieved according to their overall talent level, while also taking coaching into account. By doing that, he could figure out which Big 2 and Big 3 combinations helped their teams overachieve most.
Ayer’s conclusion was that the best Big 2 combination was from Category 8 (“Multi-faceted, high scoring wings, with high assists for their position and are great 3 point shooters”) and Category 12 (“High scoring post players, high rebounds, high block”). Current examples of those players would be Joe Johnson and Dwight Howard.
His best Big 3 combination was those two same players plus a player from Category 7 (“High scoring, high assist, high steals, high turnover point guards, who don’t shoot 3”). And a recent example of the best Big 3 combination would be Jameer Nelson, Hedo Turkoglu and Howard in 2009 (players can change categories from year to year).
Category 8 is the type of player that best complements any other type of player, according to Ayer. In addition to Johnson, Paul Pierce and Danilo Gallinari. That makes some sense, because every star that can draw a double-team needs somebody to knock down shots, but one-dimensional shooters (Kyle Korver) will hurt you in other areas.
Ayer also concluded that you can succeed with two players from Category 2 (“High scoring, dynamic guards”), like Kobe Bryant and Dwyane Wade, but it helped if you played at a fast pace.
Are Carmelo Anthony and Amar’e Stoudemire a good fit? Yes, but the stats Ayer used don’t account for defense beyond steals and blocks.
Later on Friday, group from NYU-Polytechnic Institute presented a paper (click for pdf) called “NBA Chemistry: Positive and Negative Synergies in Basketball.” They gave every player “skill” ratings for ball-handling, scoring and rebounding, and then figured out how well the skills (including duplicative skills, like two good offensive rebounders) worked together.
We measured synergies by how many additional points a combination of two skills create. For example, Chris Paul‘s offensive ballhandling is worth 4.8 points, while Reggie Evans‘ offensive rebounding is worth 3.1 points. We calculate that a team with Chris Paul’s offensive ballhandling and Reggie Evans’ defensive rebounding will have a 8.1 point advantage. Therefore we calculate synergies as worth 0.2 points (8.1-4.8-3.1).
The group then used their findings to propose mutually-beneficial trades. For example, when Paul was with the Hornets and Deron Williams was with the Jazz, the group believes both teams would have benefited from a straight-up swap of point guards. The Jazz would have been better with Paul’s ability to create turnovers and the Hornets would have been better with Williams’ scoring.
Both papers attempted to take a step toward measuring chemistry in basketball, an elusive factor in analytics to date. Each method has its flaws (Ayer’s 14 categories produce 364 different 3-man combinations, and the NYU group’s categories are rather broad), but you could see how, if refined, they might help a team put a roster together.