2019 playoffs

Home for all your discussion of basketball statistical analysis.
shadow
Posts: 217
Joined: Fri May 29, 2015 3:38 am

Re: 2019 playoffs

Post by shadow » Tue Jun 11, 2019 4:25 am

You're welcome :)

Mike G
Posts: 4415
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2019 playoffs

Post by Mike G » Tue Jun 11, 2019 11:51 am

Shadow, can you put minutes or possessions alongside player rapm?
My understanding is that players with few minutes are shifted toward zero. And the upshot is that the top and bottom ends are unpopulated by such players. But it would be nice to just see how players rank relative to others with 100+ minutes or 1000 min. etc

Personally, I'd rather know that a guy played 5 total minutes with apm = -10 and an error spread of 25, than just see -0.1 without context. And the standard error is kind of estimable from the minutes.

shadow
Posts: 217
Joined: Fri May 29, 2015 3:38 am

Re: 2019 playoffs

Post by shadow » Tue Jun 11, 2019 2:10 pm

Sure, either one shouldn't be too difficult to add. Will let you know when it's been updated.

Update: Minutes played has been added to the sheet.

Mike G
Posts: 4415
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2019 playoffs

Post by Mike G » Wed Jun 12, 2019 5:20 am

Far out. Now we can estimate total points above or below norm.

Code: Select all

pts+   1998-          Min   RAPM
1604  LeBron James  10049   7.66
919   Tim Duncan     9370   4.71
902   Manu Ginobili  6075   7.13
691   Draymond Green 4332   7.66
532   Kevin Durant   5598   4.56

490   Kevin Garnett  5158   4.56
436   Kobe Bryant    8508   2.46
425 Shaquille O'Neal 6382   3.20
403   Dwyane Wade    6697   2.89
398   Kawhi Leonard  3806   5.02

pts+   1998-          Min   RAPM
383   Stephen Curry  4193   4.38
378   Ray Allen      6064   2.99
368   James Harden   4074   4.34
364   Robert Horry   4243   4.12
353 Chauncey Billups 5321   3.18

332   Ben Wallace    4524   3.52
313   Derek Fisher   6822   2.20
310   Jason Kidd     5881   2.53
301   Dirk Nowitzki  5895   2.45
289  Tayshaun Prince 4977   2.79
That's mostly perennial allstars and dynastic dudes, plus Horry and Fisher.

eminence
Posts: 137
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2019 playoffs

Post by eminence » Wed Jun 12, 2019 3:49 pm

I'd probably count Fisher/Horry as dynastic dudes, Horry just bounced around a bit.

Mike G
Posts: 4415
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2019 playoffs

Post by Mike G » Thu Jun 13, 2019 1:21 am

Trying to re-create RAPM by looking up BPM on b-r.com and RPM on espn.com, and assuming RPM is the avg of BPM and RAPM. If so (and I assume this isn't precisely correct), then rapm = RPM*2 - BPM
Shadow's RAPM doesn't exactly concur with some of these reconstituted rapm, nor does it have any consistent relation with BPM and RPM.
Steph Curry's RS BPM was 6.3 this year, and his RPM was 6.9 -- indicating that his RAPM was higher than 6.9. For his RPM to be the avg of BPM and RAPM, the latter would be about 7.5

Code: Select all

GSW         BPM   RPM   rapm?  shad
Curry       6.3   6.91   7.5   5.80
Durant      4.3   4.92   5.5   5.49
Green       2.0   3.54   5.1   2.18
Looney      2.1   3.43   4.8   2.45
Iguodala    1.9   2.53   3.2   2.33
Cousins     3.6   2.50   1.4  -0.33
Thompson   -0.8  -0.58  -0.4  -0.43
Jerebko     0.0  -0.19  -0.4  -0.44
Livingston -1.5  -2.51  -3.5   0.40
McKinnie   -2.5  -3.07  -3.6  -0.75
Jones       3.0  -0.50  -4.0  -0.02
Cook       -2.9  -3.77  -4.6   3.17
Bell       -0.4  -3.45  -6.5  -2.02
Column 2 is the avg of 1 and 3.
What is the flaw in my logic here? I realize that RPM may not be exactly midway between BPM and RAPM; but it shouldn't be possible that Draymond Green's RPM could be higher than both his RAPM and BPM.
Same for Looney, Curry, ... And what's with Cook?

eminence
Posts: 137
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2019 playoffs

Post by eminence » Thu Jun 13, 2019 1:58 am

Is RPM still updating in the playoffs?

J.E.
Posts: 817
Joined: Fri Apr 15, 2011 8:28 am

Re: 2019 playoffs

Post by J.E. » Thu Jun 13, 2019 4:28 pm

eminence wrote:
Thu Jun 13, 2019 1:58 am
Is RPM still updating in the playoffs?
Yes
What is the flaw in my logic here? I realize that RPM may not be exactly midway between BPM and RAPM;
"rapm = RPM*2 - BPM"
is just too much of an oversimplification. Probably mostly for two reasons

- Because of its inherent regression to 0, (single season) RAPM is unlikely to give the super elite players the high ratings they deserve. Playoff only RAPM has sample size issues that go beyond that
- the 'statistical plus minus' I feed into RPM is different from BPM

Mike G
Posts: 4415
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2019 playoffs

Post by Mike G » Thu Jun 13, 2019 4:50 pm

Thanks, JE
So it seems some players may have radically changed 2019 RPM due to playoff performance.
Are playoffs weighed more heavily than the RS? Are recent games more impactful than early season?

Was BPM ever a component of RPM?
Was this a misconception all along? Did anyone but myself have this impression?

Is there exactly one RAPM for a player, regardless of who calculates it? Are Shadow's numbers the same as yours?
And what is with Quinn Cook? Does he have RAPM = +3.2 and RPM = -3.8 ?

shadow
Posts: 217
Joined: Fri May 29, 2015 3:38 am

Re: 2019 playoffs

Post by shadow » Thu Jun 13, 2019 6:26 pm

Quinn Cook had a -3.2 RAPM in the regular season in my version:

https://docs.google.com/spreadsheets/d/ ... ingle=true

J.E.
Posts: 817
Joined: Fri Apr 15, 2011 8:28 am

Re: 2019 playoffs

Post by J.E. » Thu Jun 13, 2019 6:43 pm

Mike G wrote:
Thu Jun 13, 2019 4:50 pm
Are playoffs weighed more heavily than the RS?
No
Was BPM ever a component of RPM?
No
Was this a misconception all along?
It appears so
Did anyone but myself have this impression?
Daniel and I both built our own SPM (statistical plus minus) years ago around the same time. Daniel published his and it became BPM. I didn't publish mine, except when it was in the very early stages. Given that mine and his are in the same family of metrics (SPM) I guess people assumed I was using BPM when I said I was using an SPM prior
Is there exactly one RAPM for a player, regardless of who calculates it?
In a world where everyone works with the exact same data, yes. Chances are that's not the case, but you'd want to see at least a very high correlation

shadow
Posts: 217
Joined: Fri May 29, 2015 3:38 am

Re: 2019 playoffs

Post by shadow » Thu Jun 13, 2019 7:18 pm

For the 2017-18 regular season my RAPM had a correlation of 0.98 with rd11490's. For the 2016-17 season (reg+playoffs) it had a correlation of 0.97 with JE's.

Mike G
Posts: 4415
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2019 playoffs

Post by Mike G » Thu Jun 13, 2019 7:45 pm

shadow wrote:
Thu Jun 13, 2019 6:26 pm
Quinn Cook had a -3.2 RAPM in the regular season in my version:

https://docs.google.com/spreadsheets/d/ ... ingle=true
ooops, my bad


thanks

DSMok1
Posts: 904
Joined: Thu Apr 14, 2011 11:18 pm
Location: Maine
Contact:

Re: 2019 playoffs

Post by DSMok1 » Thu Jun 13, 2019 8:21 pm

To follow up:

The RAPM "part" of a combined RPM-style metric with a good SPM prior will not look at ALL like vanilla RAPM (which regresses towards 0).

Rather, it will look more like "what is showing up in the PbP Lineup data that is not showing up in the SPM already". For many (perhaps most) players, that value may not be large. It would take a lot of lineup data to shift a player far away from a good SPM prior.
Developer of Box Plus/Minus
APBRmetrics Forum Administrator
GodismyJudgeOK.com/DStats/
Twitter.com/DSMok1

Crow
Posts: 6148
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2019 playoffs

Post by Crow » Sat Jun 15, 2019 6:52 pm

Raptors win title. 5th oldest on minutes weighted average age (during regular season). Young by historical standards but still almost half a year older than the youngest in modern times. With Gasol's arrival, the effective age went up.

Which contenders get older to get into the range of titlewinners? Which stay the same or get younger and ignore the historical record? Among those 'too young" by historical standards: Bucks, Sixers, Celtics, Thunder...

Post Reply