View previous topic :: View next topic |
Author |
Message |
BadgerCane
Joined: 05 Sep 2007 Posts: 5 Location: University of Miami-Florida
|
Posted: Tue Sep 18, 2007 12:58 pm Post subject: Normalizing Ortg |
|
|
Instead of using the multipliers for Win Shares, would there be a way to normalize Ortg to Usage? For instance, adjust everyone's Ortg up or down for if they had a Usage Rate of 18. |
|
Back to top |
|
|
94by50
Joined: 01 Jan 2006 Posts: 403 Location: Phoenix
|
Posted: Tue Sep 18, 2007 8:45 pm Post subject: |
|
|
I would use a standard of 20% of team possessions, according to the BoP formulas. The question is where to go from there - how much does one normalize for excess or short usage? |
|
Back to top |
|
|
BadgerCane
Joined: 05 Sep 2007 Posts: 5 Location: University of Miami-Florida
|
Posted: Tue Oct 09, 2007 8:33 am Post subject: |
|
|
Makes sense. It's just clearly not a linear number. 10 points at 20% does not become 20 at 40%. It becomes closer to 16. I'd just love to see a formula for getting production at one percentage to accurately project the numbers at the different percentage. I'd also imagine assists, points, and ts% change up and down at different rates. |
|
Back to top |
|
|
deepak_e
Joined: 26 Apr 2006 Posts: 325
|
Posted: Tue Oct 09, 2007 10:38 am Post subject: |
|
|
Didn't Dean talk about adjusting ORtg based on Poss% in BoP? I thought he did. |
|
Back to top |
|
|
kjb
Joined: 03 Jan 2005 Posts: 665 Location: Washington, DC
|
Posted: Tue Oct 09, 2007 11:06 am Post subject: |
|
|
Someplace on these boards, Dean posted a general rule about how much ortg changes based on usage. I can't recall what the number was, and I can never get anything out of the search function (or remember how to do it through google). _________________ My blog |
|
Back to top |
|
|
BadgerCane
Joined: 05 Sep 2007 Posts: 5 Location: University of Miami-Florida
|
Posted: Tue Oct 09, 2007 4:34 pm Post subject: |
|
|
http://sonicscentral.com/apbrmetrics/viewtopic.php?t=333&highlight=ortg+usage
Found it. He writes on the third page of that thread that for each percentage point raise in usage, Ortg goes down 6. However, he also says that for low usage players, this effect can be up to twice as high. Iknow this is kiddie math compared to a lot of what goes on on this board, but it seems like a formula for Ortg at 25% Usage Percent would = Ortg-.6*(25-Usg)...How would anyone suggest adjusting for the lower percentage players? |
|
Back to top |
|
|
davis21wylie2121
Joined: 13 Oct 2005 Posts: 374 Location: Atlanta, GA
|
Posted: Tue Oct 09, 2007 4:53 pm Post subject: |
|
|
Funny you should ask...
In my team projections, I used this formula for "ORtg penalty/reward" ("USG" = initial usage):
Code: | =(-0.0000005711205723074*(USG^4))+(0.000094227235900401*(USG^3))+(-0.00400384128781*(USG^2))+(0.01818781961676*USG)+1.2044305949478 |
This ends up giving players with an initial usage of 20% a penalty of about -0.63 for a 1% increase in usage (or a reward of +0.63 for a 1% decrease). If your usage is 30%, the penalty/reward is 0.23, which why Allen Iverson couldn't get his ORtg up into the stratosphere even if he didn't take so many shots. At the other end of the spectrum, a usage of 10% will get you a penalty/reward of 1.07 pts/100 poss for every 1% increase in usage -- meaning that the proverbial "team of Hoibergs" would have some problems if forced to create their own shots.
Last edited by davis21wylie2121 on Tue Oct 09, 2007 4:55 pm; edited 1 time in total |
|
Back to top |
|
|
Ben F.
Joined: 07 Mar 2005 Posts: 290 Location: MD
|
|
Back to top |
|
|
davis21wylie2121
Joined: 13 Oct 2005 Posts: 374 Location: Atlanta, GA
|
Posted: Tue Oct 09, 2007 5:01 pm Post subject: |
|
|
I took Dean's statement about it being 0.6 for 20% players, 0.8 for <18%, and 0.4 for >23%, and ran a regression such that the lowest-usage players (at about 10%) would experience a penalty/reward of +/- 1.0, the highest (36%) would see +/- 0.1, and everyone else would fall somewhere in between (with the slope increasing the further toward the extremes you went). It's a primitive, highly theoretical model, but I don't see any other way of modeling teammate interaction (unlike Berri, I don't believe that ignoring it and hopping for the best is an option). |
|
Back to top |
|
|
Mountain
Joined: 13 Mar 2007 Posts: 216
|
Posted: Wed Oct 10, 2007 5:23 pm Post subject: |
|
|
fyi, somewhat related to this topic-
study by Ken Pomeroy of year to year usage change in college
http://www.basketballprospectus.com/article.php?articleid=9
Wonder if rate of change in offensive efficiency for usage changes in college is similar to pros. Maybe that will come out later at basketballprospectus from Ken or Kevin.
Wonder if year to year change in usage rates is similar in NBA, or less? |
|
Back to top |
|
|
|