Adjusted Four Factors and A4PM

EvanZ · Post #1 » by **EvanZ** » Tue Feb 21, 2012 2:43 pm

If you read my other post here, you know I'm working on adjusted versions of various metrics. Now I have done adjusted versions of the four factors (Joe Sill did this a few years ago, but all his work is gone now, since he got an NBA job), and also regressed the FF onto RAPM, to calculated a new version of +/- called A4PM (adjusted four factor +/-).

http://thecity2.com/2012/02/21/new-play ... ctor-a4pm/

Post #2 » by **Doctor MJ** » Wed Feb 22, 2012 3:47 am

Interesting stuff. Good to see someone else finally pick this up. I feel like I saw some info and can't find it again. Is this from '06 to '09?

I find Shaq's effect on eFG% really astounding. We see a few heavy hitters (Nash, Wade, Garnett) breaking +1 there, and yet Shaq breaks +2 THAT late in his career. Is that as out there as it looks?

EvanZ · Post #3 » by **EvanZ** » Wed Feb 22, 2012 11:59 am

Doctor MJ wrote:Interesting stuff. Good to see someone else finally pick this up. I feel like I saw some info and can't find it again. Is this from '06 to '09?

I find Shaq's effect on eFG% really astounding. We see a few heavy hitters (Nash, Wade, Garnett) breaking +1 there, and yet Shaq breaks +2 THAT late in his career. Is that as out there as it looks?

Thanks. This is from 2010-now, so last ~2.5 seasons. Joe Sill did adjusted four factors a few years ago, but all that stuff has been taken down. As far as I know, regressing the adjusted FF on RAPM is my "invention". Now the question is really whether it's more or less predictive than RAPM itself.

Post #4 » by **Doctor MJ** » Fri Feb 24, 2012 2:42 am

EvanZ wrote:
Doctor MJ wrote:Interesting stuff. Good to see someone else finally pick this up. I feel like I saw some info and can't find it again. Is this from '06 to '09?

I find Shaq's effect on eFG% really astounding. We see a few heavy hitters (Nash, Wade, Garnett) breaking +1 there, and yet Shaq breaks +2 THAT late in his career. Is that as out there as it looks?

Thanks. This is from 2010-now, so last ~2.5 seasons. Joe Sill did adjusted four factors a few years ago, but all that stuff has been taken down. As far as I know, regressing the adjusted FF on RAPM is my "invention". Now the question is really whether it's more or less predictive than RAPM itself.

So you just mislabeled Shaq as "Miami" (since he wasn't there the past couple years)? Okay, if it's that timespan, I won't take Shaq with his super-limited minutes that seriously, but for the rest of the players I will.

I have to confess that I don't immediately grasp why making an A4PM would be an improvement over RAPM. It'd be nice to see you expound on that.

I love seeing component APM analysis like offensive and defensive, and now again with your isolated 4 factor analysis, because it's always good to see all the components when getting a truer multi-dimensional understanding of a player's potential for team impact, but I'm not seeing why re-combining it gives us an edge over the holistic RAPM approach.

EvanZ · Post #5 » by **EvanZ** » Fri Feb 24, 2012 11:01 am

Doctor MJ wrote:I love seeing component APM analysis like offensive and defensive, and now again with your isolated 4 factor analysis, because it's always good to see all the components when getting a truer multi-dimensional understanding of a player's potential for team impact, but I'm not seeing why re-combining it gives us an edge over the holistic RAPM approach.

So, I thought I was pretty clear about this:

One thing, of course, is to figure out whether it’s more or less predictive than straight-up RAPM. I haven’t done that yet, so until then, I certainly wouldn’t recommend using this as a replacement for other forms of APM or RAPM. Notably, until proven otherwise, I believe Jeremias Engelmann’s RAPM ratings are the gold standard right now. For now, what A4PM really gives is a neat way to see (at least, roughly) how each player actually gets their rating from each of the four factors.

Having said that, I'm not ruling out the possibility that A4PM could be more predictive. I don't really have theoretical grounds to be sure either way. And one very good stathead that I have communicated with about this seems to be convinced it will be more predictive. When this season is over, I'll make sure to do some retrodiction using both metrics so I can determine, at least, empirically, which system is more predictive.

Post #6 » by **Doctor MJ** » Sat Feb 25, 2012 1:20 am

EvanZ wrote:
Doctor MJ wrote:I love seeing component APM analysis like offensive and defensive, and now again with your isolated 4 factor analysis, because it's always good to see all the components when getting a truer multi-dimensional understanding of a player's potential for team impact, but I'm not seeing why re-combining it gives us an edge over the holistic RAPM approach.

So, I thought I was pretty clear about this:

One thing, of course, is to figure out whether it’s more or less predictive than straight-up RAPM. I haven’t done that yet, so until then, I certainly wouldn’t recommend using this as a replacement for other forms of APM or RAPM. Notably, until proven otherwise, I believe Jeremias Engelmann’s RAPM ratings are the gold standard right now. For now, what A4PM really gives is a neat way to see (at least, roughly) how each player actually gets their rating from each of the four factors.

Having said that, I'm not ruling out the possibility that A4PM could be more predictive. I don't really have theoretical grounds to be sure either way. And one very good stathead that I have communicated with about this seems to be convinced it will be more predictive. When this season is over, I'll make sure to do some retrodiction using both metrics so I can determine, at least, empirically, which system is more predictive.

Okay, well I did read that but your last post left me thinking that you had a hypothesis in play. Regardless, I love seeing someone do this work.