- I very much agree that offensive rebounds should be part of an

offensive rating, but I think it is very tough to come up with the

right way to do it.

DeanO adds in an offensive rebounding part in both the numerator and

denominator of his offensive efficiency formula. Here are the

formulas.

Denominator Offensive Rebounding Part

= Individual Offensive Rebounds

* Team Offensive Rebounding Weight

* Team Play Percentage

Numerator Offensive Rebounding Part

= Denominator Offensive Rebounding Part

* Team Points per Scoring Play Adjustment

This Team Points per Scoring Play Adjustment roughly measures the

number of points per effective field goal made, so it is likely to

above two, implying that the offensive rebounding part going into

the numerator will be more than twice as large as the offensive

rebounding part in the denominator. Thus, a player who did nothing

but offensive rebound would get an offensive rating of greater than

200.

The problem with the offensive rebounding part is that unlike the

points part or assist part, there is no negative associated with

offensive rebounds. Players trying to score are at risk of missed

shots and turnovers, implying that increasing the number of points

possessions will not necessarily increase a player's offensive

rating. Players attempting to get assists do so at the risk of

turnovers, so trying to get more assists will not necessarily

increase a player's offensive rating.

But a player attempting to get more offensive rebounds takes no risk

(in his offensive rating), except for maybe an occasional turnover.

Thus, players who specialize in offensive rebounds are almost

guaranteed to have high offensive ratings. This is not so with

players who specialize in point production, but may be a bit true

with players who specialize in assist production (depending on the

relative weight of the assist part versus the turnover part).

Now the reason that this does prove to be a huge problem is because

the offensive rebounding part is a pretty small fraction of total

possessions even for most offensive rebounding specialists. Thus,

it is not that big of a deal.

My approach, without going into gory detail for now, is to split up

possessions into three parts - points, assists, and offensive

rebounds. Turnovers are apportioned to the points and assists part

according to their respective proportions of total possessions.

Then I calculate separate "offensive ratings" for each category

separately with offensive rebounds being measured as a percentage of

total team offensive rebounding opportunities (adjusting for minutes

played).

This gives me three separate offensive ratings with three very

different scales. I normalize all three scales to have the same

mean and standard deviation as the minutes-weighted distribution for

points per effective field goal attempt. Then I add them together

weighting them by their respective proportion of total possessions.

This methodology does tend to reward specialists, but overall I do

not think it gives any particular advantage to assist producers

versus points producers or offensive rebound producers versus assist

producers.

Hopefully, this is enough of a description for people to get an idea

of what I am doing.

I would like also to point out that IMO the possessions

used/offensive rating tradeoff is the single most important concept

in what we do. Many folks should get some credit for that concept,

but DeanO certainly is the guy who has done the most heavy lifting.

So plese do not see what I am doing here as some sort of agenda

against DeanO. I will forever be grateful for the pathbreaking work

that he has done.

Thanks,

Dan > No, I haven't gotten a chance to read the book yet,

I'm not sure whether you mean in DeanO's system or in general, so

> but I've been craving it for a few months now. Just

> haven't gotten around to picking it up. Can you

> summarize the book in 25 words or less? Kidding, of

> course.

>

> what about taking Charges? are those tracked?

I'll answer both questions.

In DeanO's system, they're counted as any other forced turnover, not

broken out (though certainly one *could* do that).

In real life, I think some teams track this (Denver?

http://www.nba.com/nuggets/news/notes.pdf ) but it's not official.

I'll take a look and see if they're in Harvey Pollack's book.

(As an aside, one of the neat things about being part of The Media is

the chance to see what things teams track in their game notes, like

the Wolves with what we now know as "Roland Ratio" last year -- when

there was a lot of general scoffing at the notion that a team could

be 25 points better with its star player on the court.)