APBRmetrics

John Hollinger's PER Diem article for Dec. 4: "You got paid, so what happened?"

(Can be found at ESPN insider, or other ways if one chooses.)


Nearly all on teams with some form of advanced analytics staff.

Not getting listened to or making mistakes or just the player's and /or coach's fault?

Jeff Green in bottom 4% in the league on xRAPM estimate. Was bottom 20% in 2010-11. Only was worse his rookie season. Cut his estimated negative impact greatly as a sophmore but has slid backwards every season since. Guess RAPM over time was not given much weight in that decision. On what metric / database findings was that decision justified? Whether it was per game boxscore stats or the look of his basketball skills or something else, it does not look good for that call right now.

Crow wrote:http://bbs.hupu.com/4720496.html

Nearly all on teams with some form of advanced analytics staff.

Not getting listened to or making mistakes or just the player's and /or coach's fault?

I liked this article a lot. Not sure if that should be linked though, only because it is paid content (ESPN Insider) and looks to just be C+P elsewhere...

I disagree with some of the findings, though. Roy Hibbert has regressed offensively, but he (and perhaps he-alone) is anchoring the #2 Pacers defense which doesn't get a lot of help from its wings. Pacers are actually running bad slightly; opponents, for example, are shooting 80% at the line vs them this season. Perhaps not worth a max at his current production, but hardly a total flop.

Ty Lawson, as well, has historically has some clear pre ASB/post splits. I wouldn't dismiss his start as indicative of the entire season to come.

Green is #7 in the Celtics' rotation. Only Pierce and Garnett have scored more per minute, among the top 6. He seems to back Pierce, and he's shot better than Bass this year.

I can't see how his impact is negative with this team. Sullinger could get more minutes, but I don't see who else will, in the frontcourt. Pierce is 35 and playing the same minutes he did 4 years ago.
I'd say they'd be in a world of hurt without Green.

Crow wrote:Jeff Green in bottom 4% in the league on xRAPM estimate. Was bottom 20% in 2010-11. Only was worse his rookie season. Cut his estimated negative impact greatly as a sophmore but has slid backwards every season since. Guess RAPM over time was not given much weight in that decision. On what metric / database findings was that decision justified? Whether it was per game boxscore stats or the look of his basketball skills or something else produced by rigorous programming efforts, it does not look good for that call right now.

Jeff Green is teeeeeeeeeeeeeeeeeeerrrrrrrrrrrrrrrrrrrrrrriiiiiiiiiiiiiiiibbbbbbbbbbbllllllllllllllllllllllleeeeeeeeeeeee.

No one I have spoken with over the past year has any clue as far as the justifications for that extension. Occam's razor would suggest he has incriminating photos of Danny Ainge.

Green is 5th highest on team on usage rate and 9th on eFG%, TS% and offensive rating.

8th on TRB% and Assist %. 9th on PER and ws/48 and it is a drop off the cliff from the 8th placed guy on ws/48.

4th place on salary.

Tied for 3rd worst team win % while on the court for the main rotation.

I will generally trust these stats and RAPM estimates for helping to partially explain why the Celtics are barely over .500.

Totally depending on Green for backup SF (if they are not ready yet to try Joseph much) while having an open roster spot is a choice that does not make sense to me and shouldn't last.

I disagree with some of the findings, though. Roy Hibbert has regressed offensively, but he (and perhaps he-alone) is anchoring the #2 Pacers defense which doesn't get a lot of help from its wings.

they combine great shot defense (best in the league) with great defensive rebounding (also best in the league). but what's amazing is how good their overall defense is despite the fact that they are dead last in the league in forcing turnovers - that's not easy to do...

bchaikin wrote:I disagree with some of the findings, though. Roy Hibbert has regressed offensively, but he (and perhaps he-alone) is anchoring the #2 Pacers defense which doesn't get a lot of help from its wings.

they combine great shot defense (best in the league) with great defensive rebounding (also best in the league). but what's amazing is how good their overall defense is despite the fact that they are dead last in the league in forcing turnovers - that's not easy to do...

exactly my point/full agreement. I'd say Hibbert is the main reason for leading the league in shot defense/rebounding...

I have Jeff Green's current true-talent estimate at -4.12 (with Mendoza Line of -3.3), and personally would much rather see Rivers give more run to Sullinger, who at least provides some potential upside.

Maybe Ainge was fooled by Green's production over the 600 MP with the Celtics in 2011, and ignored the 10,000 MP of prior evidence that he is indeed atrocious.

talkingpractice wrote:I have Jeff Green's current true-talent estimate at -4.12 (with Mendoza Line of -3.3), and personally would much rather see Rivers give more run to Sullinger, who at least provides some potential upside.

Maybe Ainge was fooled by Green's production over the 600 MP with the Celtics in 2011, and ignored the 10,000 MP of prior evidence that he is indeed atrocious.

I have Green as -3.8 this year, -0.61 in his time with Boston in 2011, -1.3 with OKC in 2011, +0.3 with OKC in 2010, and -1.1 with OKC in 2009.

My preseason projection for him this year was -0.80. My valuation for him was about $3 million a year (preseason).

In 2010, Jeff Green was #51 in Win Shares.
His Celtics part-season the next year was even better, in WS/48.

For both seasons combined, he ranks #65 in WS. This is still equivalent to the #3 player on a pretty good team.
Over those 2 years, he shot .534 TS%, and had Reb% 9.0, Ast% 7.2, Stl and Blk% 1.5
Just 10 players had as good or better rates in those 5 areas: LeBron, Durant, Dwight, Wade, Nene, Gerald Wallace, Garnett, Josh Smith, Millsap, and Brand.
Green's TO% was lower than those guys'.

DSMok1 wrote:

talkingpractice wrote:I have Jeff Green's current true-talent estimate at -4.12 (with Mendoza Line of -3.3), and personally would much rather see Rivers give more run to Sullinger, who at least provides some potential upside.

Maybe Ainge was fooled by Green's production over the 600 MP with the Celtics in 2011, and ignored the 10,000 MP of prior evidence that he is indeed atrocious.

I have Green as -3.8 this year, -0.61 in his time with Boston in 2011, -1.3 with OKC in 2011, +0.3 with OKC in 2010, and -1.1 with OKC in 2009.

My preseason projection for him this year was -0.80. My valuation for him was about $3 million a year (preseason).

Yeah, ASPM has valued Green much higher than we have over these past seasons. Our spm model (which we use as a prior for our metric) also doesnt dislike Green as much as our IPV does. But Green has just been such a disaster in our plus/minus data on the defensive end. I don't know much about ASPM-D, but I think your model can stomach his defense much more than ours can. He's habitually worse than -2 in our defensive numbers.

The cases where metrics disagree strongly can justify re-examination of the totality of the evidence, assumptions, prejudices, model design, etc.

talkingpractice wrote:

DSMok1 wrote:

talkingpractice wrote:I have Jeff Green's current true-talent estimate at -4.12 (with Mendoza Line of -3.3), and personally would much rather see Rivers give more run to Sullinger, who at least provides some potential upside.

Maybe Ainge was fooled by Green's production over the 600 MP with the Celtics in 2011, and ignored the 10,000 MP of prior evidence that he is indeed atrocious.

I have Green as -3.8 this year, -0.61 in his time with Boston in 2011, -1.3 with OKC in 2011, +0.3 with OKC in 2010, and -1.1 with OKC in 2009.

My preseason projection for him this year was -0.80. My valuation for him was about $3 million a year (preseason).

Yeah, ASPM has valued Green much higher than we have over these past seasons. Our spm model (which we use as a prior for our metric) also doesnt dislike Green as much as our IPV does. But Green has just been such a disaster in our plus/minus data on the defensive end. I don't know much about ASPM-D, but I think your model can stomach his defense much more than ours can. He's habitually worse than -2 in our defensive numbers.

Defensive ASPM is pretty rudimentary due to the complete lack of public defensive box-score stats--an R^2 of around 0.5 on reality. See all of the derivation here: http://godismyjudgeok.com/DStats/aspm-and-vorp/

Because of that, I'd certainly trust a properly validated & regressed APM derivative for defense over anything box-score based.

Just to be clear, do we think that RAPM (even eight years of it) is reality? I've had the feeling that some people implicitly think that, but it can't be true, right?