Saturday was the second edition of the Northern California Symposium on Statistics and Operations Research in Sports, or NCSSORS for short. NCSSORS is one of two offsetting fall conferences (the other held at Harvard in odd-numbered years) that serve as the academic counterbalances to the much larger annual MIT Sloan Sports Analytics Conference.
The NCSSORS schedule was built around eight presentations by researchers from around the globe (three continents in all, with professors traveling from Australia and Japan to share their work). Some, like a discussion of how to measure the best lineups in baseball and a comparison of Europe's top goal scorers last season, were too specific to have any practical basketball application.
Only one of the presentations specifically focused on basketball. Steven Shmanske of California State University, East Bay studied whether a relationship could be found between an NBA player's effort and his level of rest and past effort (a presentation I subtitled "Both Teams Played Hard"). The difficulty there was that effort had to be quantified in terms of performance. Because most of the statistics we track on a game-by-game basis are offensive in nature, it is difficult to tell whether players are affected by fatigue at the defensive end of the floor, where the relationship between effort and achievement is much stronger by conventional wisdom.
That said, Schmanske's conclusion that rest and travel have no statistically significant effect on player performance is an interesting one. Houston Rockets analyst Ed Küpfer pointed out during the Q&A that he has found a tie between rest and team performance that satisfies the requirements for statistical significance. That might suggest the disconnect between player effects and team effects indeed lies at the defensive end of the floor.
Other presentations had some implications that could be generalized to basketball. Richard Ryall gave an entertaining presentation on how home-field advantage makes a difference throughout an AFL Australian Rules Football game. Separating by whether the home team was favored, he found that the home-field effect was strongest when the game was in the balance--a home underdog taking a lead or a home favorite trailing entering the fourth quarter, most notably.
Two afternoon presenters--Michael Schuckers and Tim Swartz--considered drafts. Schuckers offered evidence that the NFL's "draft chart"--consulted by front offices throughout the league when trading their picks--overvalues selections at the top of the draft. Swartz looked at how various salary distributions, including the NBA's scale contracts for first-round picks, compared to an idealized distribution that favors elite talent. My past research has suggested that the rookie scale does a good job of matching how players perform during their first NBA contracts, although the difference between the pay for the first and second overall picks in particular should probably be steeper.
I can't say I've ever seen a draft chart for the NBA like the one that is so common in the NFL. In part, pick-for-picks trades are rare in the NBA draft, with players or cash usually included on at least one end of deals. Also, top picks simply don't move that frequently, except as distant conditional future selections. The last trade involving a top-four pick came in 2006. My suspicion, however, is that high lottery picks are correctly valued in the NBA, where talent tends to drop off more quickly than it does in the seven-round NFL draft.
Besides the formal presentations, other researchers had a chance to share studies via poster presentations at lunch. The most interesting of these related to the NBA was by Nima Shaahinfar, who considered how the distribution of various statistics on NBA teams related to their overall offensive performance. I had seen that kind of study with scoring--the research suggests, rather clearly, that offenses are usually balanced because of a lack of elite talent, not because of a wealth of scorers--but not with other statistics. A broader distribution--that is, less balance--proved beneficial in several statistics (3Pas, FTAs, Asts), but teams did better when they shared their turnovers and their offensive rebounding more evenly. The implication of those results deserves more thought.
While academia was the focus of the NCSSORS program, insiders were represented as well. The morning and afternoon sessions both wrapped up with featured speakers who work for teams--St. Louis Cardinals senior quantitative analyst Sig Mejdal and Roland Beech of the Dallas Mavericks. Beech was an ideal choice as a featured speaker because he just wrapped up his first year as part of the Mavericks' coaching staff, a role that sees him travel with the team and sit behind the bench during games. Just a handful of other team analysts work physically in their organization's front office, let alone have the kind of connection that Beech does with the coaches.
Beech recalled realizing how much he had to learn before his first training camp, when the coaches walked through "Floppy"--a common action on the floor--and he had no idea where to go. In general, Beech focused on the respect his position has given him for the importance of coaching in the NBA. (For a similar perspective on coaching, see Dean Oliver's interview with SlamOnline.com that went live the same day.) That in turn has important implications for player evaluation, since we often tend to assume that players' statistics represent their true level of skill, which is fairly constant aside from typical aging patterns.
In part, I'm not sure how to put that information to use. Beech pointed to the example of Channing Frye, who developed into one of the league's best three-point shooters seemingly overnight last season in Phoenix. At the same time, there are plenty of examples of other big men who were never able to translate their skill shooting from 15-18 feet back behind the three-point line. Assuming other players can make the same leap might be dangerous.
To me, the more practical takeaway is the reminder that individual player ratings are a reflection of many things beside talent. Roles, coaching, motivation and noise all complicate the relationship between ability and performance. As a result, we have to assume a degree of variability in player ratings. Basically, player A has to rate a fair amount better than player B for us to be confident he is truly the superior player--to the extent such a judgment really exists.
Since NCSSORS organizer Ben Alamar works for the Oklahoma City Thunder and has strong ties within the NBA community, members of nine organizations were in attendance. As usual, some of the most interesting conversations happened over lunch or other social settings. Long after the conference, I listened in (while watching Washington knock off Oregon State in double-OT) to an entertaining conversation between Küpfer, Oliver and Mejdal comparing and contrasting their experiences and how quantitative analysis is received differently in baseball than in basketball. Alas, those insights will have to stay off the record.
The 2010-11 Pro Basketball Prospectus is now available in paperback form on Amazon.com. For sample chapters and more information, see our book page.
Kevin Pelton is an author of Basketball Prospectus.
You can contact Kevin by clicking here or click here to see Kevin's other articles.