APBRmetrics

Out of curiosity, why not just use Daniel's ASPM as the prior? Or would it make any difference?

Probably because it's tricky to cross-validate the informed RAPM, to figure weights for the prior. You can't just split your sample up randomly to cross validate like J.E. can with PbP based stats. Which is why, Neil, I asked you for the data splits you gave me for ASPM. Hopefully I will figure out how to use the odd/even data to cross validate ASPM as a prior for RAPM....

This won't be too easy, though.

Does this RAPM adjust for quarter? Ex: Starters get less credit for 2nd quarter since they play less in the 2nd quarter? How does it deal with games that are blowouts in the 4th? Does this also adjust for strong opponents?

Did you add a variable to adjust for height like for example 7 footers have higher Defensive RAPM than a 6 footer?

colts18 wrote:Does this RAPM adjust for quarter? Ex: Starters get less credit for 2nd quarter since they play less in the 2nd quarter? How does it deal with games that are blowouts in the 4th? Does this also adjust for strong opponents?

Did you add a variable to adjust for height like for example 7 footers have higher Defensive RAPM than a 6 footer?

The fake lineup data is set up in such a way that starter/non_starter distribution is similar to those of games after 2000, which had similar quarter outcomes.
All the other (R)APM rules still apply; so, yes, it adjusts for opponents
The 10year list has no height info; the year-by-year rankings get influenced by height, years of experience and BoxScore stats

J.E., is there a chance you will do the fake RAPM for the 80's? B-R has quarter by quarter and MP since 1986 so I would assume its possible to do it from 86-90.

colts18 wrote:J.E., is there a chance you will do the fake RAPM for the 80's? B-R has quarter by quarter and MP since 1986 so I would assume its possible to do it from 86-90.

Hey

It's possible but unlikely. Had I known that quarter by quarter score and MP was available only since 1986 I would have included those from the start. Are you an 80s Pistons fan?

J.E. wrote:

colts18 wrote:J.E., is there a chance you will do the fake RAPM for the 80's? B-R has quarter by quarter and MP since 1986 so I would assume its possible to do it from 86-90.

Hey

It's possible but unlikely. Had I known that quarter by quarter score and MP was available only since 1986 I would have included those from the start. Are you an 80s Pistons fan?

No, just a fan of the lowly Wizards I'm just interested in the numbers for Magic, Bird, Kareem, and MJ.

What's the R value between these fake RAPM's and the real RAPM?

colts18 wrote:What's the R value between these fake RAPM's and the real RAPM?

http://www.apbr.org/metrics/viewtopic.p ... 067#p13656

J.E. wrote:R^2 for real RAPM and fake RAPM is ~0.25. Fake RAPM has everyone even closer together than RAPM does, so it's really hard to get a good rating when being on a bad team etc.

I believe that's for 01-11

sideshowbob wrote:

colts18 wrote:What's the R value between these fake RAPM's and the real RAPM?

http://www.apbr.org/metrics/viewtopic.p ... 067#p13656

J.E. wrote:R^2 for real RAPM and fake RAPM is ~0.25. Fake RAPM has everyone even closer together than RAPM does, so it's really hard to get a good rating when being on a bad team etc.

I believe that's for 01-11

I think R value was from the first version of the stat that didn't adjust minutes played to the scoring margin entering the quarter.

J.E. wrote:10 year fake RAPM (91-00) give the following ranking

Jordan (+4.5)
Bird (+4)
Magic (+3.8)
Rodman
Shaq
Pippen
DRobinson
Wayne Cooper
Mourning
Cliff Levingston
Eddie Jones
Doc Rivers
Danny Ainge
Dan Majerle
Sam Perkins
Penny Hardaway (+2.6)

Rodman at #4 by this method and time period when first published seems worth noting given the past controversy about his relative ranking / greatness. The ranking is now 9th in the current data set.