APBRmetrics

[Scott S]

Does anyone know what Dean Oliver did to come up with his "Offensive Skill Curves" in his book Basketball on Paper? He lists several of them and they seem reasonable, but he doesn't seem to say how they were determined. I tried plotting by games in the season and that didn't seem to work (I wouldn't have really expected it to - if a player is shooting poorly, you might expect him to limit his workload, which would interfere with this measurement.) There are some things I can think of to estimate this pattern, but they seem fairly intensive and it is tough to obtain meaningful data to extract this type of pattern so cleanly.

I have seen surprisingly little on this topic, but it seems very important for many aspects of analysis. Does anyone have anything else they would like to discuss on offensive skill curves? For those of you who have not read "Basketball on Paper" by Dean Oliver, offensive skill curves are defined as "plots of how efficient a player is ... and what percentage of a team's possessions he can use to still maintain that efficiency..." They plot Offensive Rating at 5% of the teams possessions used, 10%, 15%, etc. In the cases Oliver shows percentage of possessions and offensive rating have a negative correlation, that is, as the percentage of possessions goes up, offensive rating goes down.

[Ryan J. Parker]

I would like to understand this as well. I assume they would be determined on a per game basis. I'll run the formula through my box scores and see what I come up with.

Who would we like to look at? Kobe? LeBron?
_________________
I am a basketball geek.

[Neil Paine]

We've had a lot of discussion on the general topic of usage vs. efficiency:

*The Great Usage vs. Efficiency Debate
*How well would this team perform?
*Usage vs. Efficiency
*2006 All-NBA teams
*Next in Basketball Analysis
*Wins Produced - Wages of Wins (Berri, Schmidt, and Brook)

But I don't think we've ever been able to figure out exactly how Dean arrived at those graphs, much less how you'd go about constructing better skill curves for specific players. Eli W. probably came the closest to finding the average player's trade-off, but he's property of an NBA team now (as is DeanO, obviously), so they're probably not sharing whatever new things they find. I once referred to the usage-efficiency trade-off/skill curves as "the fundamental debate of APBRmetrics," but I'm not sure there's a way to definitively determine a player's trade-off at each point of usage... Unfortunately, the NBA won't let us hold controlled experiments with their players.

[Ryan J. Parker]

Interesting. I'm going to use the formulas in BoP and see if I can replicate something similar to what Dean's got. Never realized no one has been able to get the same results.
_________________
I am a basketball geek.

[Ryan J. Parker]

Here's my attempt:

http://www.basketballgeek.com/2009/02/17/basketball-on-papers-skill-curves/
_________________
I am a basketball geek.

[HoopStudies]

When writing Basketball on Paper, I had to give away some of what I did just to get an opportunity in the league (and to have a book worth publishing). But I held back on many things. Skill curves were a compromise. I thought they were important enough to introduce and complex enough to allow a peek.
_________________
Dean Oliver
Author, Basketball on Paper
The postings are my own & don't necess represent positions, strategies or opinions of employers.

[Ryan J. Parker]

Neil pointed out that Dean meant to have possession% on the x-axis and off rating on the y-axis. It definitely makes the graphs easier to read.

Here are the updated images:




_________________
I am a basketball geek.

[kjb]

If you're trying to reconstruct DeanO's skills curves, start with higher-usage guys first. Their usage tends to be more consistent game to game than lower-usage guys. In other word's a role player is better able to minimize his usage when he's having an off day -- a star often has to carry the offensive burden no matter what kind of day he's having.