NBA HD: Are standing reach/wingspan overvalued?

Perhaps players with larger wing spans are less coordinated when they stretch out to their full extension. The extra length doesn't really help them then.

If we talk about floor position, then I wonder how the team's systems would affect their rebounding rates,

For example, a center such as bonner who would usually be situated outside would probably not be in a position to go for O-rebounds a lot, in the same way that his defensive counterpart, would probably be out of position to go for the boards.

Wouldn't this be more useful if you restricted the study to, say, only centers? It seems like that would help sort out the issue that Ben noted.

McGee is only 22 years old - calling him an average rebounder at this stage is like calling Greg Oden a bust ... too early to know. Give McGee a couple more years, then we can compare apples with apples.

If you have strength data (for example, bench press data from the NBA combine), could you try that as a predictor in place of weight?

Do you have replication materials for this? (Dataset and code.)

Rebounding rate is very likely correlated to the floor position
Lets assume that where a player takes his shots is a good predictor of where he plays. So a multifactor regression using the %age of shots from 3pg, %age at rim, %age within 10 feet (whatever) and %age out to the 3pt line may be a good idea to try.

Wingspan is looked at as a defensive tool, not necesarilly as a rebounding tool. Longer arms make you a bigger defender, and make it harder for someone to dribble/pass around you. Your goal was to determine if wingspan is over rated, but I think you judged it in the wrong way.

Also, the only time wingspan really comes up is with guards, for the reasons I mentioned above, not with centers.

It'a because McGee is the laziest/dumbest player in the league. I once saw him in "warmup" before a game where he didn't take any shots...he just pretended to dunk an invisible ball and swat away his teammates shots

tell that to a 6'7" 228 lb cross dressing 7 time nba rebounding champ

I think you are looking at the wrong stats for effectiveness of wing-span. Rebounding is more about position than reach. Think of Charles Barkley - top rebounder at 6' 5" (2m).
Look at blocks and steals (but mostly blocks). A player with long arms can sneak up on a shooter.

Ben makes a very valid point - it's like factor analysis but the inverse - you know the factors but do not know the underlying 'core' variables.

Also wingspan, height & weight are probably correlated like hell so there is no surprise in wingspan not adding much (or any) power to the height/weight model

For one thing, weight correlates (probably strongly) with height.

There are lots of different ways to run these regressions. I mean, it's a good finding that standing reach is less strongly correlated than height, but it sounds like your wingspan regression was conducted with height and wingspan, rather than wingspan replacing height. Does that even matter? I don't know.

What about the difference between height and wingspan? Just a thought.

I really hope you are using REB/48 or REB/36. But of course a wide body is going to be a good rebounder; other guys can't occupy that space.

Weight as the best predictor makes sense, as it should correlate strongly with height too. I understand that body types can vary a lot, but in general weight should vary roughly with the cube of height or something like that, and the outliers on the heavy side will have an advantage in boxing out, etc. So really it's not surprising that weight is the best proxy.

To put it simply: quantify effort and you will have your missing variable.

I remember seeing Kobe in a game they lost against the Cavs being out on the wing guarded by Lebron during free throws towards the end of the game, then Kobe dove to the hoop anticipating a miss (which occurred) and grabbed the rebound where 4 other players had better position and where all taller and heavier. Luckily it didn't matter he chucked up some jumper and they lost but that sequence was impressive.

My guess is that your variables are acting as proxies for factors that you are not measuring. Specifically I would guess that height acts as a proxy for where a player spends most time on the court, ie proximity to the basket when rebounds come off the rim. If this is a large component of the predictive value of height then we may be over-estimating the beta truly associated with height/length based measures of rebounding predictors and underestimating floor position as a variable (obviously, since floor position is not included in the model).