RAPM request thread

[J.E.]

People have made requests for various RAPM versions. This shall be a thread where people can make requests and I will post the results (within a reasonable amount of time). Be prepared that some of them I'll just post on pastebin

Also, please request variations that actually make sense. So, no "RAPM for all games played on a Monday"

Variations can differ in years used, leading/coach/age-adjustment, and other things

[DSMok1]

I'm going to ask a hard one first:

I would like to see an RAPM that is informed by a prior containing only age and MPG information.

EDIT: Perhaps Team Margin should also be included, but not age?

A regression I ran recently on BPM, since 2000:

Code: Select all

SUMMARY OUTPUT

Regression Statistics
Multiple R        0.748050082
R Square          0.559578925
Adjusted R Square 0.559226926
Standard Error    1.927022923
Observations             6262

ANOVA
                      df          SS           MS          F     Significance F
Regression                  5  29516.42523 5903.2850471589.717637           0
Residual                 6256  23231.13891 3.713417345
Total                    6261  52747.56414

                 CoefficientsStandard Erro   t Stat     P-value   Lower 95%   Upper 95%  Lower 95.0% Upper 95.0%
Intercept              -6.938        0.434     -15.990 1.940E-56       -7.789      -6.088      -7.789      -6.088
RegressedMPG            0.289        0.019      14.852 4.618E-49        0.251       0.328       0.251       0.328
Age                     0.041        0.016       2.588 9.677E-03        0.010       0.072       0.010       0.072
Team Mar                0.099        0.015       6.806 1.095E-11        0.071       0.128       0.071       0.128
ReMPG*Age              -0.002        0.001      -3.265 1.100E-03       -0.004      -0.001      -0.004      -0.001
ReMPG*Mar               0.005        0.001       8.056 9.352E-16        0.004       0.006       0.004       0.006

[colts18]

-15 year RAPM (Age and no age adjustment)

-Yearly NPI and PI RAPM from 2001-2015

-15 year xRAPM

-15 year 4 factor RAPM (Rebounding, Turnovers) and pace

[Crow]

There are a lot of ideas that have been mentioned over the years in past threads.

Here are some of what I remember (or just thought of) and believe to be worthwhile:

Multiseason playoff only RPM (or multi reg. season & playoffs run - just reg. season)
Of course pairs thru 5 man lineups
3pt focused RPM factor
ORPM on a players usage plays vs not
RPM against top ten teams, offenses, defenses (multiyr if sample size is concern)
League avg RPM for certain lineup types like 2PGs, 3 3pt shooters, smallball, etc.
RPM against a top 10 at position counterpart
Home vs away RPM
East vs west
Overall RPM - "Counterpart RPM" (plus minus of only activity where last act is made within that matchup) equals Indirect RPM
4th quarter RPM
RPM when with 3 or 4 plus starters vs less
ORPM with a very high TSA guy vs not
RPM in high combined usage lineups vs. avg or low.
RPM with a blended metric prior
Adjusted Player Tracking Factor Data RPM
PlayerRPM by perceived each position played (overall or offense and defense)
RPM splits for player high, avg, low usage by game

Others that are even more likely to be considered too much:
RPM in games with high, avg, low foul calls
Before / after allstar break
League avgs for Yr before, of, after "contract yr"
League avg. RPM at high elevation (SLC and Denver) by position or player type
Plays made in first 18 seconds of shot clock vs last 6.
When up or down by 10.
Any 2 or more of these criteria combined vs solo criteria present or none.

The test of whether these splits are worthwhile begins with seeing and thinking about the results. Don't know if worthwhile until you try. My feeling is that trying is likely to be more worthwhile than not. Something interesting is likely to be touched on even if the results related to the main objective are fuzzy.

[fpliii]

Very cool. Thanks J.E. for this. For now I'd love to see 2015 prior-informed vanilla RAPM.

[J.E.]

I want to limit this thread to (R)APM (not RPM) only and want to keep it relatively simple. I'm not going to create several different types of SPMs here, as that's an entirely different field and should probably warrant a different thread

[J.E.]

Very cool. Thanks J.E. for this. For now I'd love to see 2015 prior-informed vanilla RAPM.

https://docs.google.com/spreadsheets/d/ ... sp=sharing

Note: This has no aging adjustment whatsoever - it's just straight up RAPM - so this is not what you'd optimally use for prediction

[J.E.]

-15 year RAPM (no age adjustment)

https://docs.google.com/spreadsheets/d/ ... sp=sharing

This one's got a couple of interesting names here and there.

Draymond Green is ranked 8th, Amir Johnson 11th, Baron Davis 12th, Jeff Foster 29th

Zach LaVine and Anthony Bennett in the bottom 20 out of 1467. Jeff Green as well, who was discussed recently.

Not many in the bottom 20 are still in the league, besides those already mentioned: Tony Wroten, JJ Hickson, Norris Cole and Tim Hardaway Jr.

[J.E.]

-15 year RAPM (Age adjustment)

https://docs.google.com/spreadsheets/d/ ... sp=sharing

All those who played many minutes at very young/old age get a bump compared to the file in my last post.

KG moves into #1, Stockton into 4th etc.

Also removes LaVine and Bennett from the bottom 20. Jeff Green, Norris Cole, Tim Hardaway Jr and JJ Hickson still there

Interestingly it ranks Green ahead of Curry. Other players that are relatively young and in the top 102 are: Aldridge, Harden, Amir Johnson, Griffin, Rubio, Paul George, Westbrook, Conley, MKG, Kawhi, Klay Thompson, Anthony Davis, Dwight, Danny Green, Jrue Holiday, Lowry, Love, Hibbert, Lawson, Ibaka, Bledsoe, Thad Young, Bosh, Reggie Jackson, Durant, Marc Gasol, Lillard, Asik, Gallinari, Middleton, Exum(!), Crowder(!), Butler

[permaximum]

2-year NPI-RAPM without any adjustments (Playoffs included, equal weighting for different seasons and playoffs). It would be very useful to develop a player tracking metric.

[permaximum]

BTW I wonder, why is Jordan placed 36th in 15-year age adjusted RAPM while he was placed 268th in 14-year age adjusted RAPM?

[colts18]

Very cool. Thanks J.E. for this. For now I'd love to see 2015 prior-informed vanilla RAPM.

https://docs.google.com/spreadsheets/d/ ... sp=sharing

Note: This has no aging adjustment whatsoever - it's just straight up RAPM - so this is not what you'd optimally use for prediction

Awesome work J.E. Is that RAPM with prior years (How many prior years) or a multi year RAPM where 2015 is weighted higher than previous seasons?


Is it possible to run a 15 year playoff RAPM using 15 year RS+PS RAPM as the opponent and teammate adjustment?

[Mike G]

-15 year RAPM (Age adjustment)

https://docs.google.com/spreadsheets/d/ ... sp=sharing
...
Interestingly it ranks Green ahead of Curry...

Draymond, Jeff, and Danny have been mentioned very recently.
You probably don't mean Gerald, Willie, JaMychal, Erick, or others in the last 15 yrs.

[sideshowbob]

Wow, really appreciate this - this seems like a lot of extra work. Kudos sir!

Very cool. Thanks J.E. for this. For now I'd love to see 2015 prior-informed vanilla RAPM.

https://docs.google.com/spreadsheets/d/ ... sp=sharing

Note: This has no aging adjustment whatsoever - it's just straight up RAPM - so this is not what you'd optimally use for prediction

Just a quick question. This kind of set (single-year PI) was never made public by yourself for 2014, correct? How comparable would it be to the gotbuckets/talkingpractice set for previous years (specifically 2014), which use a base year of 06 I believe. If I recall, your old 01-12 sets used a 01 base year, is this just a continuation of that or is this 15 informed by 14 NPI?

[fpliii]

Very cool. Thanks J.E. for this. For now I'd love to see 2015 prior-informed vanilla RAPM.

https://docs.google.com/spreadsheets/d/ ... sp=sharing

Note: This has no aging adjustment whatsoever - it's just straight up RAPM - so this is not what you'd optimally use for prediction

Thank you for sharing!