2017-18 RAPM/RPM/etc. Thread

[Blackmill]

Very happy to see improvements being made to the traditional RAPM model. I've felt for a while that a Bayesian model was needed (but why aren't there any uncertain estimates???). Unfortunately, almost all of the important details are missing in the author's write up, and this ambiguity is obscuring what could be major weaknesses or strengths of the model as a whole. So, if you find it confusing, that's because it is.

Anyways, I hope the author is willing to explain his model in more detail. I'll wait until then to post my thoughts, except for this, the author's model ignores temporal and spatial context within the possession. So, with better data and a better prior, but without modeling the game's dynamics, we get a reported 60% improvement over RAPM. That paints a pretty optimistic view for what we could do with models that use temporal and spatial information. Within the decade, I think traditional RAPM will look as crude as PER does now.

[laika]

Read on Twitter

I've merely skimmed this, but it's interesting. Cool to see Rod Strickland on the 97-98 list as he seems to be written off as a talented yet lower impact player. You also have Antoine Walker and Marbury on there who were considered negative impact guys (not that I disagree). Although I wonder if it being early in their careers has anything to do with it.

Looks like another completely worthless model. We have lots of evidence that Curry was far better than Westbrook last year. Yet Westbrook is twice as good? Sorry. Back to the drawing board.

[Blackmill]

[pending edits]

[Bad Gatorade]

For those interested, Daniel Myers (the creator of BPM) is in the process of revising BPM and alleviating some of its issues.

http://www.apbr.org/metrics/viewtopic.php?f=2&t=9689

There is a Tableau graphic that has some 5 season RAPM splits (I love the logic behind this - 5 years should be enough to eliminate much of the collinearity, but also not so large that important details for each player will be lost amongst aging curves and what not). Don't think the data can be downloaded, but you can just peruse the graph and most of the interesting names will pop up.

Top 10 in each group of seasons (unless I missed somebody):

97-01
-------
Michael Jordan: 9.9
Tim Duncan: 8.6
Vince Carter: 7.8
Shaquille O'Neal: 7.7
Kevin Garnett: 7.2
John Stockton: 7.2
Rasheed Wallace: 6.7
Grant Hill: 6.7
Gary Payton: 6.3
David Robinson: 6.2

02-06
-------
Kevin Garnett: 10.9
Dirk Nowitzki: 8.9
Tim Duncan: 8.8
Andrei Kirilenko: 6.9
Baron Davis: 6.7
Manu Ginobili: 6.6
Metta World Peace: 6.6
Paul Pierce: 6.6
Tracy McGrady: 6.6
Ben Wallace: 6.4

07-11
-------
LeBron James: 11.2
Steve Nash: 8.7
Dirk Nowitzki: 8.6
Kevin Garnett: 8.2
Dwight Howard: 7.3
Kobe Bryant: 6.7
Manu Ginobili: 6.7
Chris Paul: 6.5
Dwyane Wade: 6.2
Luol Deng: 6.0

12-16
-------
LeBron James: 11.5
Chris Paul: 9.8
Stephen Curry: 9.8
Kevin Durant: 8.7
Kyle Korver: 7.8 (yes, Kyle Korver)
Kyle Lowry: 7.6
Russell Westbrook: 7.5
Draymond Green: 7.4
Andre Iguodala: 6.6
Paul George: 5.8

[Colbinii]

For those interested, Daniel Myers (the creator of BPM) is in the process of revising BPM and alleviating some of its issues.

http://www.apbr.org/metrics/viewtopic.php?f=2&t=9689

There is a Tableau graphic that has some 5 season RAPM splits (I love the logic behind this - 5 years should be enough to eliminate much of the collinearity, but also not so large that important details for each player will be lost amongst aging curves and what not). Don't think the data can be downloaded, but you can just peruse the graph and most of the interesting names will pop up.

Top 10 in each group of seasons (unless I missed somebody):

97-01
-------
Michael Jordan: 9.9
Tim Duncan: 8.6
Vince Carter: 7.8
Shaquille O'Neal: 7.7
Kevin Garnett: 7.2
John Stockton: 7.2
Rasheed Wallace: 6.7
Grant Hill: 6.7
Gary Payton: 6.3
David Robinson: 6.2

02-06
-------
Kevin Garnett: 10.9
Dirk Nowitzki: 8.9
Tim Duncan: 8.8
Andrei Kirilenko: 6.9
Baron Davis: 6.7
Manu Ginobili: 6.6
Metta World Peace: 6.6
Paul Pierce: 6.6
Tracy McGrady: 6.6
Ben Wallace: 6.4

07-11
-------
LeBron James: 11.2
Steve Nash: 8.7
Dirk Nowitzki: 8.6
Kevin Garnett: 8.2
Dwight Howard: 7.3
Kobe Bryant: 6.7
Manu Ginobili: 6.7
Chris Paul: 6.5
Dwyane Wade: 6.2
Luol Deng: 6.0

12-16
-------
LeBron James: 11.5
Chris Paul: 9.8
Stephen Curry: 9.8
Kevin Durant: 8.7
Kyle Korver: 7.8 (yes, Kyle Korver)
Kyle Lowry: 7.6
Russell Westbrook: 7.5
Draymond Green: 7.4
Andre Iguodala: 6.6
Paul George: 5.8

At first glance this stat looks good. Thanks for posting this Hatorade!

Sent from my SM-G960U using RealGM mobile app

[DatAsh]

For those interested, Daniel Myers (the creator of BPM) is in the process of revising BPM and alleviating some of its issues.

http://www.apbr.org/metrics/viewtopic.php?f=2&t=9689

There is a Tableau graphic that has some 5 season RAPM splits (I love the logic behind this - 5 years should be enough to eliminate much of the collinearity, but also not so large that important details for each player will be lost amongst aging curves and what not). Don't think the data can be downloaded, but you can just peruse the graph and most of the interesting names will pop up.

Top 10 in each group of seasons (unless I missed somebody):

97-01
-------
Michael Jordan: 9.9
Tim Duncan: 8.6
Vince Carter: 7.8
Shaquille O'Neal: 7.7
Kevin Garnett: 7.2
John Stockton: 7.2
Rasheed Wallace: 6.7
Grant Hill: 6.7
Gary Payton: 6.3
David Robinson: 6.2

02-06
-------
Kevin Garnett: 10.9
Dirk Nowitzki: 8.9
Tim Duncan: 8.8
Andrei Kirilenko: 6.9
Baron Davis: 6.7
Manu Ginobili: 6.6
Metta World Peace: 6.6
Paul Pierce: 6.6
Tracy McGrady: 6.6
Ben Wallace: 6.4

07-11
-------
LeBron James: 11.2
Steve Nash: 8.7
Dirk Nowitzki: 8.6
Kevin Garnett: 8.2
Dwight Howard: 7.3
Kobe Bryant: 6.7
Manu Ginobili: 6.7
Chris Paul: 6.5
Dwyane Wade: 6.2
Luol Deng: 6.0

12-16
-------
LeBron James: 11.5
Chris Paul: 9.8
Stephen Curry: 9.8
Kevin Durant: 8.7
Kyle Korver: 7.8 (yes, Kyle Korver)
Kyle Lowry: 7.6
Russell Westbrook: 7.5
Draymond Green: 7.4
Andre Iguodala: 6.6
Paul George: 5.8

At first glance this stat looks good. Thanks for posting this Hatorade!

Sent from my SM-G960U using RealGM mobile app

I feel like these 3-5 year studies are the most accurate way to truly judge a player's impact. 1 year RAPM just isn't enough data in my opinion; too much collinearity. I think the aging on these longer studies is less of a problem than the insufficient data problem on the 1 year studies.

*Note, I'm not talking about a player's goodness, which I consider a product of impact, role, portability, and scalability; I'm merely talking about impact.

A few surprising results for me:

Jordan is higher than I would have expected for 97 and 98. I would have expected something more or less in line with 07-11 Kobe, given that I've always seen that version of Jordan more or less equal with prime Kobe, just with a much more off ball focused role. Could have something to do with there only being 2 years of data for Jordan, but could also be that I underestimate 96-98 Jordan. I'd love to see what he looked like 88-93

Shaq is lower than I would have guessed. Especially since the 00-02 Lakers were so dominant and Kobe is so low too. Do I underestimate the impact of the non Kobe/Shaq Lakers?

Kyle Korver, my gosh. Just goes to show how valuable a player can be when placed in a role that really suits his strengths. That 17 Cavs squad is probably one of the most loaded we've ever seen, offensively - Lebron, Kyrie, Korver, Love. It's a shame that Lebron was basically that team's defense.

Unsurprised by Lebron and KG looking like monsters. Impact metrics are a big reason why I rate these two so high.

Chris Paul always comes out looking near GOAT level on impact metrics, and he's done so for a long time. What is a fair all time ranking for Chris Paul, if we are to believe these impact metrics?

Warriors have 4 guys in the top 10 for 12-16

[Colbinii]

For those interested, Daniel Myers (the creator of BPM) is in the process of revising BPM and alleviating some of its issues.

http://www.apbr.org/metrics/viewtopic.php?f=2&t=9689

There is a Tableau graphic that has some 5 season RAPM splits (I love the logic behind this - 5 years should be enough to eliminate much of the collinearity, but also not so large that important details for each player will be lost amongst aging curves and what not). Don't think the data can be downloaded, but you can just peruse the graph and most of the interesting names will pop up.

Top 10 in each group of seasons (unless I missed somebody):

97-01
-------
Michael Jordan: 9.9
Tim Duncan: 8.6
Vince Carter: 7.8
Shaquille O'Neal: 7.7
Kevin Garnett: 7.2
John Stockton: 7.2
Rasheed Wallace: 6.7
Grant Hill: 6.7
Gary Payton: 6.3
David Robinson: 6.2

02-06
-------
Kevin Garnett: 10.9
Dirk Nowitzki: 8.9
Tim Duncan: 8.8
Andrei Kirilenko: 6.9
Baron Davis: 6.7
Manu Ginobili: 6.6
Metta World Peace: 6.6
Paul Pierce: 6.6
Tracy McGrady: 6.6
Ben Wallace: 6.4

07-11
-------
LeBron James: 11.2
Steve Nash: 8.7
Dirk Nowitzki: 8.6
Kevin Garnett: 8.2
Dwight Howard: 7.3
Kobe Bryant: 6.7
Manu Ginobili: 6.7
Chris Paul: 6.5
Dwyane Wade: 6.2
Luol Deng: 6.0

12-16
-------
LeBron James: 11.5
Chris Paul: 9.8
Stephen Curry: 9.8
Kevin Durant: 8.7
Kyle Korver: 7.8 (yes, Kyle Korver)
Kyle Lowry: 7.6
Russell Westbrook: 7.5
Draymond Green: 7.4
Andre Iguodala: 6.6
Paul George: 5.8

At first glance this stat looks good. Thanks for posting this Hatorade!

Sent from my SM-G960U using RealGM mobile app

I feel like these 3-5 year studies are the most accurate way to truly judge a player's impact. 1 year RAPM just isn't enough data in my opinion; too much collinearity. I think the aging on these longer studies is less of a problem than the insufficient data problem on the 1 year studies.

*Note, I'm not talking about a player's goodness, which I consider a product of impact, role, portability, and scalability; I'm merely talking about impact.

A few surprising results for me:

Jordan is higher than I would have expected for 97 and 98. I would have expected something more or less in line with 07-11 Kobe, given that I've always seen that version of Jordan more or less equal with prime Kobe, just with a much more off ball focused role. Could have something to do with there only being 2 years of data for Jordan, but could also be that I underestimate 96-98 Jordan. I'd love to see what he looked like 88-93

Shaq is lower than I would have guessed. Especially since the 00-02 Lakers were so dominant and Kobe is so low too. Do I underestimate the impact of the non Kobe/Shaq Lakers?

Kyle Korver, my gosh. Just goes to show how valuable a player can be when placed in a role that really suits his strengths. That 17 Cavs squad is probably one of the most loaded we've ever seen, offensively - Lebron, Kyrie, Korver, Love. It's a shame that Lebron was basically that team's defense.

Unsurprised by Lebron and KG looking like monsters. Impact metrics are a big reason why I rate these two so high.

Chris Paul always comes out looking near GOAT level on impact metrics, and he's done so for a long time. What is a fair all time ranking for Chris Paul, if we are to believe these impact metrics?

Warriors have 4 guys in the top 10 for 12-16

RE: CP3

I dont think there is a "fair" ranking of him. His career is unique [as are most great NBA players] and he is easily dismissed for reasons I dont agree with. It isnt out of right field to put him next to or even slightly ahead of Wade nor is it unrealistic to have him closed to 40 than 20.

I think he is a fantastic player [and I know Gatorade views him in a similar light as myself] with few flaws that get brought up either too often or get magnified to degrees other players dont. Unfortunately that happens to all players who don't win championships [fairly or not], even if they perform as well as their peers [in CP3's case better than his peers] in the post-season.

Sent from my SM-G960U using RealGM mobile app