Stat geekery: KG rules
|11.04.09 at 10:46 am ET|
Another day another stat geek model (stat geek is a term of endearment). This one comes courtesy of Joe Sill at Hoopnumbers, who describes himself as an analytics consultant from Chicago. Sill has a PhD in Computation and Neural Systems from Cal Tech and a BS in Applied Math from Yale, so he’s a pretty smart dude.
Sill works in the field of Adjusted +/-, which is a system that has drawn some attention due to its orginators Wayne Winston and Jeff Sagarin and their work with the Dallas Mavericks. In layman’s terms, Simple +/- ratings, which can be found in most online box scores, account for how many points a team is better or worse with an individual player on the floor. For example, Marquis Daniels was +31 against Philly Tuesday night.
Simple +/- doesn’t factor in context, which is where Adjusted +/- comes into play. (If you want to see the math, see Joe’s page explaining his concepts). Adjusted +/- is said to be “noisy,” which means you might get some odd results over a short period of time, but it’s an interesting concept and it’s getting more and more play in NBA front offices.
The takeway from Sill’s analysis over the last three years, which he calls Regularized Adjusted +/- (RAPM) is that Kevin Garnett grades out the highest of any NBA player, and by a fairly wide margin. The second-highest ranked Celtic is Rajon Rondo, followed by Paul Pierce and then Ray Allen. Interestingly, Allen graded out highest last season on the Celtics.
A good rule of thumb for advanced NBA metrics is that they are a useful way to help tell a story. John Hollinger’s PER rating, for example, is a number derived strictly from the box score and he will tell you that it doesn’t account for individual defense beyond getting numbers like steals, blocks and rebounds. It’s a piece of the puzzle, but not the whole puzzle.
As yet, no one has advanced a truly credible single number that completely explains performance, like VORP in baseball, and there’s a strong thought in the stat community that it may be impossible to find such a number in basketball analysis.
With that in mind, what Sill’s numbers might tell us is that Garnett is far and away the “most valuable” Celtic in ways that don’t necessarily show up in the box score and that Allen had something of a late-career renaissance last season. We saw that manifest itself on the floor last season, especially when Garnett was not on the floor, so the numbers make intuitive sense.
(Hat tip to basketball-reference’s layups blog for directing me to Sill’s site, Hoopnumbers)