APBRmetrics

Well I'm using that for my player's as well. I saw the Individual Offensive Rating in Dean Oliver's book. Should I be using Dean Oliver's formula from "Basketball on Paper" or can I keep using (PTS/POSS)*100 for individuals. Is their a big difference between the two?

Just did Dean Oliver's rating for Individual Total Possesions. Big difference in some players.

DSMok1 wrote:EvanZ, I would beg to differ. It's semantics, but I think we need to have clear definitions for the terms.

I define efficiency as the scoring rate. The points per shot, or offensive rating, or any other "rate" stat.

On the other hand, production should account for the usage.

I'm not sure I like using the word "efficiency", that's all, because of the connotation. I'm not sure "scoring rate" is the best term either, since some players may shoot higher %, but take shots much more infrequently. I think we can simply talk about the stat, itself (TS%, eFG%, FT%, etc), and the connotation of that stat separately.

I've done Individual Total Possession formula from "Basketball on Paper". When I add up all the individual possessions I get 566 possessions total. When I do (FGA+(.475*FTA)-ORB+TO)*.96 for the team I get 616 possessions. Which # of possessions is the right one.

Isn't it 0.44*FTA? That's the coefficient I normally see people using.

This is for college basketball.

EvanZ wrote:

xkonk wrote:. You asked who is scoring more efficiently; usage tells you nothing about efficiency.

Sure it does (in response to the bolded part). If there were no relation between usage and efficiency, then Tyson Chandler's usage would (should) be near 35%.

I was simply saying that usage the measure says nothing about efficiency. All it contains is shot attempts and turnovers. In theory, a player could have a 35% usage solely from missing shots and giving the other team the ball while another player has a 35% usage based only on made shots. Of course that doesn't happen, but there is no reason it couldn't from a numbers perspective.

Your concern is a more practical one that assumes that more efficient players should (and perhaps do) get more possessions to use, or that players are given as much usage as their efficiency can handle. I haven't seen any especially convincing data on that front.

At any rate, it sounds like zsap is more interested in overall production than efficiency, and in that case including some measure of possessions used or usage would be necessary.

I have a problem. When I add up all the individual possessions (using the formula from "Basketball on Paper") I get 637 total possessions. But when I do 0.5 * ((Tm FGA + 0.4 * Tm FTA - 1.07 * (Tm ORB / (Tm ORB + Opp DRB)) * (Tm FGA - Tm FG) + Tm TOV) + (Opp FGA + 0.4 * Opp FTA - 1.07 * (Opp ORB / (Opp ORB + Tm DRB)) * (Opp FGA - Opp FG) + Opp TOV)) I get 591 Possessions total. I assume that 637 is right, but why am I getting a different number for the formula I just mentioned.

Jumpin in on this party a bit late, but here's how I feel:

The best correlation I could find between Offensive Rating, Usage, and Offensive Regularized Plus-Minus was the following

Use
1) ORTG x Usg%
and
2) Usg%


Where #1 has a positive correlation, and #2 has a negative correlation.
The simple one-year equation was Offensive Rapm=0.5xORTG*(Usage/100) - .46xUsage - 1.5

While these numbers are deflated by the fact that it was RAPM and 1-year RAPM at that, it would make more sense to estimate overall offensive contributions by 107.2 + estimated offensive RAPM from the above equation (to give a "usage-free" ORTG similar to Kevin Pelton's), in my opinion.

According to this model, your team will be hurt significantly by a low-usage player who has an average (107) offensive rating.

This re-iterates Eli Witus' study that explains: As a player uses possessions more, his teammates efficiency increases per possession used.

Looks like there might be a missing parenthesis to cover the one at the very end. (or an extra one?)

Could you double check it and walk thru a calculation of a negative value or at least specify the input values and give the outcome?

Sorry, it was just an extra parenthesis. And it's not a very long study so there's lots of noise, to be sure.

Ok, thanks for the clarification. At first check, that seems like a handy formula.

bbstats wrote: The simple one-year equation was Offensive Rapm=0.5xORTG*(Usage/100) - .46xUsage - 1.5

While these numbers are deflated by the fact that it was RAPM and 1-year RAPM at that, it would make more sense to estimate overall offensive contributions by 107.2 + estimated offensive RAPM from the above equation (to give a "usage-free" ORTG similar to Kevin Pelton's), in my opinion.

According to this model, your team will be hurt significantly by a low-usage player who has an average (107) offensive rating.

This re-iterates Eli Witus' study that explains: As a player uses possessions more, his teammates efficiency increases per possession used.

If I understood your equation correctly, a player with an 107 offensive rating can range in usage from 1% to 34% and only move from -1.4 to +1 in RAPM. Similarly, a player with a 20% usage can move in offensive rating from 88 to 121 and only move from -1.9 to 1.4. Those don't seem like big ranges, and I'm curious what the error is like.

I also don't see how this involves a player's teammates. Aren't these measures all player-specific?

Xkonk, like I mentioned in my post, the 2011 RAPM numbers are deflated. Further, a regression equation with an R^2 of .45 (more importantly that it's low-sloping)is going to deflate even further in predicting.

If all you want to do is predict offensive output via ORTG & usage, such an equation would be helpful. At the least, it gives an OK "offensive ranking" via various ORTGs and usages.