Analyzing who / which type player affects the diminishing returns and what direction would seem to a necessary step to get to the point of helping or evaluating team construction choices.

Across the league for the cases where returns vs projected vary by more than some amount (positive or negative) are any type of players present or absent more often than others or than usual for that type? I’m thinking of criteria like younng vs vet, especially tall or short, presence of big “hops” or “great strength”, the great box out guy vs out of area rebounder (does he rob his teammates more often?). the effects probably will vary on offensive and defensive glass. Just some possibilities to continue to push further.

(By the way, that was my next-to-last post on that site before Berri started deleting my posts complimenting Eli’s work. Obiviously, I’ll never post there again.)

Great work, Eli. When Berri referred to a comment about “putting the nail in the coffin” regarding his (rebounding) work, that was one of my (deleted) comments citing your excellent work.

Eli: that r of .77 is at the player level, right? As best I can tell, there is little or no correlation at the team level (.05 last season). Which is itself an interesting issue: how can the two be correlated at the player level, but not at the team level? The only answer I can see is that the player stats have a very weak relationship with team totals, i.e. diminishing returns.

Rashad: I have looked at shooting efficiency a bit, and see no sign of diminishing returns there. In fact, there appears to be a “multiplier effect,” in which greater efficiency from position X is associated with greater efficiency at the teams’ other 4 positions. However, this could reflect team construction as opposed to an actual player interaction effect.

http://www.countthebasket.com/blog/2007/12/17/does-good-pitching-beat-good-hitting-in-basketball/

I didn’t at all mean to imply that there wasn’t a connection between a team’s ORB% and its DRB%. In a previous post I discussed that relationship and how both share some common factors, finding a stong correlation of 0.77 between the two:

http://www.countthebasket.com/blog/2007/11/30/rebounding-and-height/

Obviously many of the skills involved in being a good or bad offensive rebounder are the same skills that make one a good or bad defensive rebounder. Though what I think is most interesting are the differences that make some players (and teams) better at one than the other.

“The key is that for each team, their ORB% is independent of their DRB% - they happen at different ends of the court. It’s true that a team’s ORB% is identical to (one minus) their opponent’s DRB%, but that’s just like how a team’s FT% is identical to their opponent’s FT% allowed.”

This is simply not true, and frankly displays a pretty glaring lack of insight into how two variables might co-relate, unless you believe that grabbing defensive rebounds is a skill-less endeavor.

That is, a team that boxes out well will rebound well defensively, and thus prevent offensive rebounds. Therefore, a teams DRB% directly influences their opponents’ ORB% (unless you believe that DRB% isn’t a consequence of boxing out well) — how much and how to quantify this is a difficult question to answer, but if you truly believe that rebounding well defensively has no impact on offensive rebounding, then I suggest that you haven’t really thought this true. The reverse is true as well, of course, teams that crash the boards a lot will lower their opponents DRB%. BUT, this effect should be smaller when playing a good defensive rebounding team and larger when playing a poor defensive rebounding team

With ft% allowed, there is absolutely no (legal) way for any given team to influence the other teams FT%, other than perhaps some minor effects, like being very smart about not fouling Chauncey Billups, or by passing out “thundersticks” that may (or may not) distract opposing FT shooters.

The fact that more rebounding talent has diminishing marginal returns (which I think even Berri would agree with, especially when presented with this data) doesn’t address the question of how valuable that marginal rebound is.

Echoing a previous commenter I would be curious to see similar work for offensive efficiency metrics. I would imagine that you would see similar diminishing marginal returns. But again that doesn’t address of the value of adding at the margins. And the answer is, probably a lot.