Mandatory Credit: Kirby Lee-USA TODAY Sports
It has been remarked a couple of times this off season that the Celtics are going into the season with an upside down roster, their guards can’t shoot and their bigs can’t play defense. Certainly with Rajon Rondo, Phil Pressey and Marcus Smart on the roster, all with shaky three point shots, there is some truth to that. The Celtics front court also scored relatively poorly looking at the ability to protect the rim, using the new SportVu data.
But before we get too far into the specifics of the Celtics roster, as currently constructed, I want to question the idea that it would inherently be bad if a team did have an ‘upside down’ roster.
James Brocato, over at his website Shut Up and Jam, did an interesting study that calls that idea into question. He looked at long term stabilized adjusted plus minus data, known as RAPM, comparing players by position on offense and defense. There are issues with RAPM to be sure, but in the realm of orthogonal data on basketball defense, the one eyed man is king. In any case, in the study, Brocato found that, yes, centers tended to score higher on defense on average than guards, and guards on average, scored better on offense.
So, centers are more ‘important’ to defense, right?
Well, you may have noticed I used scare quotes there, because that’s not whole story. It’s true that the data supports the idea that a team benefits defensively having a center on the court. But the interesting thing Brocato found, was that the range of values between the top centers and bottom centers on defense was largely the same as the range between the top point guards and bottom point guards on defense.
Here’s Brocato’s image from his study:
I ran a similar study with both RAPM and ESPN’s Real Plus Minus (RPM), a version that combines box score stats and plus/minus data, and in both cases got the same result as Brocato, guard rated better on offense and centers better on defense, but with a spread that is largely similar. The implication is that hiring an above average back court defender, defined as say top third at position is as advantageous as hiring a top third front court defender. And the same thing applies to a top 5% defender at either spot.
The chart below demonstrates the idea, by taking a line up with all average defenders and replacing them with a top 5% guard or center and then a bottom 5% defender at each position:
In both studies, the relative difference between a top 5% defensive point guard and top 5% center being inserted into the line up was small. Using the 2014 RPM data as an example, Andrew Bogut ranked at the top 5% mark among centers and Chris Paul was at the top 5% mark for point guards. The RPM data indicates, that to say nothing of his offense, replacing an average point guard with Chris Paul gives his team almost as much of a defensive boost as replacing an average center with Bogut.
Conversely, Aaron Brooks ranked at approximately the bottom 5% mark for point guards and Atlanta rookie Mike Muscala ranked as approximately the bottom 5% mark for centers. The difference between their defense and the position average is similarly close.
To be clear, both this year’s data and Brocato’s longer term study only include players actual NBA front offices and coaching staffs considered plausible as an NBA center, it doesn’t imply that one could trot Shane Larkin out as the back line of a defense and not suffer a disproportionate fall off in defense. It is also too broad a brush study, to comment on the efficacy of twin tower line ups or three guard line ups and the like.
Even though the difference between the spread in the studies is not statistically significant, my prior is that there is at least some small increase in the dispersion in talent for centers if only because of the small pool among 6-11 human beings. There may also be more ways to ‘hide’ a back court player on defense than a center. But, what this data does indicate, I think, is that the difference between guards’ defense matters, perhaps more than we often give credit.