'17-'18 POY discussion

[Senior]

lmao if we ever do another RPOY project I can't wait to get to the prime KG years

[Jaivl]

lmao if we ever do another RPOY project I can't wait to get to the prime KG years

2008 is gonna be fun.

[Senior]

lmao if we ever do another RPOY project I can't wait to get to the prime KG years

2008 is gonna be fun.

06-07 is gonna be worse, at least 08 you can see Boston's historic defense but with 06-07 there's not many team results on KG's side. I'm gonna throw out the middle linebacker stuff for old time's sake too

like I think Kobe gets underrated here too but it's not like KG is Jamal Crawford

[eminence]

lmao if we ever do another RPOY project I can't wait to get to the prime KG years

2008 is gonna be fun.

If that does get redone one of these years I really think it should go forwards next time, progression feels so much more natural.

I can tell you this much, Dikembe is getting votes next go-round.

[Jaivl]

lmao if we ever do another RPOY project I can't wait to get to the prime KG years

2008 is gonna be fun.

06-07 is gonna be worse, at least 08 you can see Boston's historic defense but with 06-07 there's not many team results on KG's side. I'm gonna throw out the middle linebacker stuff for old time's sake too

like I think Kobe gets underrated here too but it's not like KG is Jamal Crawford

I'm not really going to be the one to argue for KG in 07, though.

[Outside]

Where did you get the RAPM?

https://docs.google.com/spreadsheets/d/e/2PACX-1vSzp3G5rwP9xgCgluVGmR3Qj4-BMoGSYiuTKM6o_pzES6s95oQE1nQvB2CXed-4fRc_MMGgpULtDaJ_/pubhtml?gid=1825430955&single=true

Is this your own version of RAPM?

[Outside]

Actually it's not.

Find me a list of any statistic. It will be ranked on something. Therefore, any statistic is invalid because it can't rank a player. Don't ever use any statistic to measure a player, not PPG, not assists, not PER, nothing.

PER is no better I agree but the raw stats are consistent. If Harden leads the league in scoring, it's because he scores the most points. It doesn't claim otherwise, it's honest.

RAPM is deceptive af and can be used to push agendas like nobody's business.

RAPM doesn't claim anything otherwise either. It's a list of the players whose teams see the biggest difference in point differential when they're on the court, with some adjustments for team context. It's not pretending to be anything else either.

Just like with any other stat, it's up to the people using it to interpret it correctly. Raw stats are manipulated like crazy too. Like how TS% has been used in this thread to act like Westbrook isn't among the elite offensive players in the league.

Two points.

TS%

I never used TS% to claim that Westbrook isn't an elite offensive player. He clearly is. I also never claimed that TS% was the ultimate indicator of efficiency, that shooting efficiency is the only thing that matters, or that TS% captures a player's total offensive impact. My point is that Westbrook's low TS% is an indicator of lower efficiency that is a weakness in his POY resume compared to other candidates.

I don't see why that's controversial. I agree that his poor shooting efficiency is mitigated by other benefits that come with his aggressive style of play, but what we're talking about is how Westbrook compares to the very best players in the league. When several of them are also excellent at creating opportunities for teammates and have significantly better shooting efficiency, that shows a weakness in Westbrook's performance relative to other POY candidates.

How RAPM is used

I agree that RAPM isn't a straight ranking system and that it's up to people to interpret RAPM correctly and present RAPM data correctly in their arguments. Unfortunately, what RAPM says about a player requires quite a bit of context, and from what I've seen in the recent part of this thread is that context is rarely presented with the RAPM. Instead, I've seen things like E-Balla calling Jamal Crawford the worst 20 MPG player in the league because he's 489th out of 493 in RAPM. That violates what you say about RAPM on two counts: 1) it uses RAPM as a ranking system; and 2) it doesn't provide context.

Here's what I've learned from my introductory research on RAPM:

-- You apparently need to be a data analytics expert to fully understand it.

-- It's designed to use past data to predict future performance (kind of the opposite of the caveat on an investment brochure).

-- It uses ridge regression to increase the predictive accuracy of RAPM on future performance.

-- Jeremias Engelmann seems to be the godfather of RAPM and is the expert behind ESPN's RPM.

-- There are multiple versions of RAPM.

-- The formula used to calculate RAPM isn't made public.

-- RAPM may be calculated using non-performance factors such as height.

Please feel free to correct me on any of the above that I got wrong.

I have yet to find a fan-level explanation of what RAPM is and what it tells you. I've seen repeatedly that it requires context, but I haven't found an explanation of what that context is, how to determine what the context is for a particular player, or how to properly use RAPM when comparing players.

Given all of the above, RAPM comes across to a non-data analyst as a black box that spits out numbers that we're supposed to accept as the best metric to assess players and that it trumps everything else. Maybe it is that good, but without the necessary understanding that is apparently impossible for a non-data analyst to acquire and seeing how RAPM data is presented by those who promote it and claim to understand it, I'm reluctant to accept it blindly on faith, especially when it runs counter to other data that I have available.

I'll keep working to expand my understanding of RAPM, but those who create it and those who promote its use seem to make that really hard.

[ardee]

PER is no better I agree but the raw stats are consistent. If Harden leads the league in scoring, it's because he scores the most points. It doesn't claim otherwise, it's honest.

RAPM is deceptive af and can be used to push agendas like nobody's business.

RAPM doesn't claim anything otherwise either. It's a list of the players whose teams see the biggest difference in point differential when they're on the court, with some adjustments for team context. It's not pretending to be anything else either.

Just like with any other stat, it's up to the people using it to interpret it correctly. Raw stats are manipulated like crazy too. Like how TS% has been used in this thread to act like Westbrook isn't among the elite offensive players in the league.

Two points.

TS%

I never used TS% to claim that Westbrook isn't an elite offensive player. He clearly is. I also never claimed that TS% was the ultimate indicator of efficiency, that shooting efficiency is the only thing that matters, or that TS% captures a player's total offensive impact. My point is that Westbrook's low TS% is an indicator of lower efficiency that is a weakness in his POY resume compared to other candidates.

I don't see why that's controversial. I agree that his poor shooting efficiency is mitigated by other benefits that come with his aggressive style of play, but what we're talking about is how Westbrook compares to the very best players in the league. When several of them are also excellent at creating opportunities for teammates and have significantly better shooting efficiency, that shows a weakness in Westbrook's performance relative to other POY candidates.

How RAPM is used

I agree that RAPM isn't a straight ranking system and that it's up to people to interpret RAPM correctly and present RAPM data correctly in their arguments. Unfortunately, what RAPM says about a player requires quite a bit of context, and from what I've seen in the recent part of this thread is that context is rarely presented with the RAPM. Instead, I've seen things like E-Balla calling Jamal Crawford the worst 20 MPG player in the league because he's 489th out of 493 in RAPM. That violates what you say about RAPM on two counts: 1) it uses RAPM as a ranking system; and 2) it doesn't provide context.

Here's what I've learned from my introductory research on RAPM:

-- You apparently need to be a data analytics expert to fully understand it.

-- It's designed to use past data to predict future performance (kind of the opposite of the caveat on an investment brochure).

-- It uses ridge regression to increase the predictive accuracy of RAPM on future performance.

-- Jeremias Engelmann seems to be the godfather of RAPM and is the expert behind ESPN's RPM.

-- There are multiple versions of RAPM.

-- The formula used to calculate RAPM isn't made public.

-- RAPM may be calculated using non-performance factors such as height.

Please feel free to correct me on any of the above that I got wrong.

I have yet to find a fan-level explanation of what RAPM is and what it tells you. I've seen repeatedly that it requires context, but I haven't found an explanation of what that context is, how to determine what the context is for a particular player, or how to properly use RAPM when comparing players.

Given all of the above, RAPM comes across to a non-data analyst as a black box that spits out numbers that we're supposed to accept as the best metric to assess players and that it trumps everything else. Maybe it is that good, but without the necessary understanding that is apparently impossible for a non-data analyst to acquire and seeing how RAPM data is presented by those who promote it and claim to understand it, I'm reluctant to accept it blindly on faith, especially when it runs counter to other data that I have available.

I'll keep working to expand my understanding of RAPM, but those who create it and those who promote its use seem to make that really hard.

Phenomenal post (other than the WB part but that's minor).

I am getting a masters in applied economics which is 50% statistics. I spend a lot of time on R, so I know my way around regressions to a certain extent. And I still don't know how they arrive at the numbers RAPM spits out, because I've never found a good explanation, or any explanation whatsoever. Yet RAPM is still this magnificent fountain of truth we aren't supposed to question at the risk of ridicule (which is very evident in this thread as you can see).

I believe the reason it rose to such prominence on this board is because a certain player whom I will not name had a contingent of fervent supporters who realized all other arguments for him weren't good enough, and thus located an obscure stat that placed this certain unnamed player on a pedestal and then began to spread its virtues to the masses so that this certain unnamed player would gain more notoriety.

Trust me I've been around for a while, this movement has been years in the making. The effort to normalize RAPM as the be all end all stat is primarily, I believe, so that more and more people will see said unnamed player as being far above his station.

Sent from my [device_name] using [url]RealGM mobile app[/url]

[Jaivl]

.

Most of the things you said are referring to RPM, but not RAPM. If you have the time, this video is a quite decent cursory explanation from the man himself. The main presentation it not really about RAPM, but close enough, and mentions RAPM.

Watch on YouTube

[therealbig3]

RAPM doesn't claim anything otherwise either. It's a list of the players whose teams see the biggest difference in point differential when they're on the court, with some adjustments for team context. It's not pretending to be anything else either.

Just like with any other stat, it's up to the people using it to interpret it correctly. Raw stats are manipulated like crazy too. Like how TS% has been used in this thread to act like Westbrook isn't among the elite offensive players in the league.

Two points.

TS%

I never used TS% to claim that Westbrook isn't an elite offensive player. He clearly is. I also never claimed that TS% was the ultimate indicator of efficiency, that shooting efficiency is the only thing that matters, or that TS% captures a player's total offensive impact. My point is that Westbrook's low TS% is an indicator of lower efficiency that is a weakness in his POY resume compared to other candidates.

I don't see why that's controversial. I agree that his poor shooting efficiency is mitigated by other benefits that come with his aggressive style of play, but what we're talking about is how Westbrook compares to the very best players in the league. When several of them are also excellent at creating opportunities for teammates and have significantly better shooting efficiency, that shows a weakness in Westbrook's performance relative to other POY candidates.

How RAPM is used

I agree that RAPM isn't a straight ranking system and that it's up to people to interpret RAPM correctly and present RAPM data correctly in their arguments. Unfortunately, what RAPM says about a player requires quite a bit of context, and from what I've seen in the recent part of this thread is that context is rarely presented with the RAPM. Instead, I've seen things like E-Balla calling Jamal Crawford the worst 20 MPG player in the league because he's 489th out of 493 in RAPM. That violates what you say about RAPM on two counts: 1) it uses RAPM as a ranking system; and 2) it doesn't provide context.

Here's what I've learned from my introductory research on RAPM:

-- You apparently need to be a data analytics expert to fully understand it.

-- It's designed to use past data to predict future performance (kind of the opposite of the caveat on an investment brochure).

-- It uses ridge regression to increase the predictive accuracy of RAPM on future performance.

-- Jeremias Engelmann seems to be the godfather of RAPM and is the expert behind ESPN's RPM.

-- There are multiple versions of RAPM.

-- The formula used to calculate RAPM isn't made public.

-- RAPM may be calculated using non-performance factors such as height.

Please feel free to correct me on any of the above that I got wrong.

I have yet to find a fan-level explanation of what RAPM is and what it tells you. I've seen repeatedly that it requires context, but I haven't found an explanation of what that context is, how to determine what the context is for a particular player, or how to properly use RAPM when comparing players.

Given all of the above, RAPM comes across to a non-data analyst as a black box that spits out numbers that we're supposed to accept as the best metric to assess players and that it trumps everything else. Maybe it is that good, but without the necessary understanding that is apparently impossible for a non-data analyst to acquire and seeing how RAPM data is presented by those who promote it and claim to understand it, I'm reluctant to accept it blindly on faith, especially when it runs counter to other data that I have available.

I'll keep working to expand my understanding of RAPM, but those who create it and those who promote its use seem to make that really hard.

Phenomenal post (other than the WB part but that's minor).

I am getting a masters in applied economics which is 50% statistics. I spend a lot of time on R, so I know my way around regressions to a certain extent. And I still don't know how they arrive at the numbers RAPM spits out, because I've never found a good explanation, or any explanation whatsoever. Yet RAPM is still this magnificent fountain of truth we aren't supposed to question at the risk of ridicule (which is very evident in this thread as you can see).

I believe the reason it rose to such prominence on this board is because a certain player whom I will not name had a contingent of fervent supporters who realized all other arguments for him weren't good enough, and thus located an obscure stat that placed this certain unnamed player on a pedestal and then began to spread its virtues to the masses so that this certain unnamed player would gain more notoriety.

Trust me I've been around for a while, this movement has been years in the making. The effort to normalize RAPM as the be all end all stat is primarily, I believe, so that more and more people will see said unnamed player as being far above his station.

Sent from my [device_name] using [url]RealGM mobile app[/url]

Or, you know, people just like having more information and don't have a preconceived bias to a player.

[Texas Chuck]

I'm gonna throw out the middle linebacker stuff for old time's sake too

I think I love you.

[ardee]

lmao if we ever do another RPOY project I can't wait to get to the prime KG years

2008 is gonna be fun.

If that does get redone one of these years I really think it should go forwards next time, progression feels so much more natural.

I can tell you this much, Dikembe is getting votes next go-round.

Dr Spaceman and I talked about it, this is the project that should be done this summer. Peaks was done 3 years ago, this was done 8 years ago.

And forward is fine, do the current year, then go back to '57 and get rolling.

[eminence]

2008 is gonna be fun.

If that does get redone one of these years I really think it should go forwards next time, progression feels so much more natural.

I can tell you this much, Dikembe is getting votes next go-round.

Dr Spaceman and I talked about it, this is the project that should be done this summer. Peaks was done 3 years ago, this was done 8 years ago.

And forward is fine, do the current year, then go back to '57 and get rolling.

Hmm, I'd prefer '55 as a starting point, but it could be fun.

[Bad Gatorade]

Spoiler:

RAPM doesn't claim anything otherwise either. It's a list of the players whose teams see the biggest difference in point differential when they're on the court, with some adjustments for team context. It's not pretending to be anything else either.

Just like with any other stat, it's up to the people using it to interpret it correctly. Raw stats are manipulated like crazy too. Like how TS% has been used in this thread to act like Westbrook isn't among the elite offensive players in the league.

Two points.

TS%

I never used TS% to claim that Westbrook isn't an elite offensive player. He clearly is. I also never claimed that TS% was the ultimate indicator of efficiency, that shooting efficiency is the only thing that matters, or that TS% captures a player's total offensive impact. My point is that Westbrook's low TS% is an indicator of lower efficiency that is a weakness in his POY resume compared to other candidates.

I don't see why that's controversial. I agree that his poor shooting efficiency is mitigated by other benefits that come with his aggressive style of play, but what we're talking about is how Westbrook compares to the very best players in the league. When several of them are also excellent at creating opportunities for teammates and have significantly better shooting efficiency, that shows a weakness in Westbrook's performance relative to other POY candidates.

How RAPM is used

I agree that RAPM isn't a straight ranking system and that it's up to people to interpret RAPM correctly and present RAPM data correctly in their arguments. Unfortunately, what RAPM says about a player requires quite a bit of context, and from what I've seen in the recent part of this thread is that context is rarely presented with the RAPM. Instead, I've seen things like E-Balla calling Jamal Crawford the worst 20 MPG player in the league because he's 489th out of 493 in RAPM. That violates what you say about RAPM on two counts: 1) it uses RAPM as a ranking system; and 2) it doesn't provide context.

Here's what I've learned from my introductory research on RAPM:

-- You apparently need to be a data analytics expert to fully understand it.

-- It's designed to use past data to predict future performance (kind of the opposite of the caveat on an investment brochure).

-- It uses ridge regression to increase the predictive accuracy of RAPM on future performance.

-- Jeremias Engelmann seems to be the godfather of RAPM and is the expert behind ESPN's RPM.

-- There are multiple versions of RAPM.

-- The formula used to calculate RAPM isn't made public.

-- RAPM may be calculated using non-performance factors such as height.

Please feel free to correct me on any of the above that I got wrong.

I have yet to find a fan-level explanation of what RAPM is and what it tells you. I've seen repeatedly that it requires context, but I haven't found an explanation of what that context is, how to determine what the context is for a particular player, or how to properly use RAPM when comparing players.

Given all of the above, RAPM comes across to a non-data analyst as a black box that spits out numbers that we're supposed to accept as the best metric to assess players and that it trumps everything else. Maybe it is that good, but without the necessary understanding that is apparently impossible for a non-data analyst to acquire and seeing how RAPM data is presented by those who promote it and claim to understand it, I'm reluctant to accept it blindly on faith, especially when it runs counter to other data that I have available.

I'll keep working to expand my understanding of RAPM, but those who create it and those who promote its use seem to make that really hard.

Phenomenal post (other than the WB part but that's minor).

I am getting a masters in applied economics which is 50% statistics. I spend a lot of time on R, so I know my way around regressions to a certain extent. And I still don't know how they arrive at the numbers RAPM spits out, because I've never found a good explanation, or any explanation whatsoever. Yet RAPM is still this magnificent fountain of truth we aren't supposed to question at the risk of ridicule (which is very evident in this thread as you can see).

I believe the reason it rose to such prominence on this board is because a certain player whom I will not name had a contingent of fervent supporters who realized all other arguments for him weren't good enough, and thus located an obscure stat that placed this certain unnamed player on a pedestal and then began to spread its virtues to the masses so that this certain unnamed player would gain more notoriety.

Trust me I've been around for a while, this movement has been years in the making. The effort to normalize RAPM as the be all end all stat is primarily, I believe, so that more and more people will see said unnamed player as being far above his station.

Sent from my [device_name] using [url]RealGM mobile app[/url]

Outside - I think most of what you're talking about is RPM, rather than RAPM, although I think you're definitely in the ballpark with regard to understanding RPM.

RAPM is essentially a regression that equates a lineup (comprising of both teams) to scoring margins. So, for example, a game between two teams would have an equation such as -

x1 + x2 + x3 + x4 + x5 - y1 - y2 - y3 - y4 - y5 = m

where m is the scoring margin, x1, x2... etc are players from one team, and y1, y2... etc are players from another team. A regression is run so that each of these x and y variables (i.e. the players) is allotted a numeric value which attempts to estimate how much these players influence the scoring margin. RAPM uses ridge regression (aka regularisation) which attempts to reduce standard errors that happen in the raw adjusted plus minus (APM) numbers, and provide more accurate results.

There's quite a bit of documentation out there nowadays which should help the understanding of RAPM.

Now, RPM is a different beast, because the plus minus component contains a box score related prior. A prior is essentially a distribution that attempts to emulate the distribution you're trying to form before it's actually calculated. The box score prior is basically, "this is our best estimate of what RAPM should be, without actually calculating RAPM." Then, RAPM is calculated, but since RAPM is volatile, the uncertainties that go along with RAPM are shaped by the prior (i.e. the box score, and things such as height) and this is what forms RPM, and why RPM is a bit of a black box. It's also why the results of RPM seem to better showcase the "stars" of the NBA rather than RAPM - the stars of the game are often those who put up the biggest box score stats, so these players are working with a higher prior.

ardee, I think that, with regards to RAPM, RPM and its cousins, the important thing to understand is what RAPM can and can't do, before making assessments on it.

RAPM is excellent at incorporating things that aren't otherwise observed in the box score. Things such as positioning, off ball movement, leadership etc can be incredibly hard, if not impossible to quantify strictly using the box score. This is the real selling point of RAPM - there are many players who are great in box score statistics but not so great in RAPM, and vice versa. A guy like say, Klay, looks to be a much worse defender than Carlos Boozer by the box score, but that's because the defensive box score is highly incomplete. And really, this is a huge selling point. Comparing the box scores of say, Kobe and Durant might imply that Durant is a better offensive player, but it's statistics such as RAPM that show that offensively, Kobe has some degree of impact that Durant hasn't had. Is it that Kobe is more dynamic, and is better at warping defences (and creating open looks for his teammates + increasing offensive rebound opportunities)? Is it that he's a better passer, but doesn't necessarily pass more than Durant? RAPM can't answer this, but it can provide that little bit of food for thought that prompts further analysis of players.

Another point to note is that RAPM looks at elements that are in the box score, but shouldn't necessarily be tied to a single player. For example, a guy like Westbrook might be seen as "turnover prone" but when he's taking up so much offensive responsibility, it's only natural that he'll have a lot of turnovers (and the other Thunder players, much less). Ditto for guys like Adams and Nene (who are incredible at boxing out, which boosts the rebounds of other players), Steve Nash (whose playmaking is partially tied to the efficiency of other players) and players that gamble for steals (it's easier to gamble for steals when you've got an already elite defence around you).

Now, some of the weaknesses of RAPM include -

Sample size - even entire year results of RAPM can be somewhat fluky, or face giant collinearity issues. Attempts to mend this include using multi-year and prior-informed RAPM (i.e. using results from previous years in order to influence results from the assessed year) but this obviously has its own limitations too. Not just thanks to the player himself being different quality over multiple seasons, but also the teammates of a player improving/dropping off in quality.

Role - RAPM looks at how point margins change within a specific role, and are NOT a measure of goodness within a vacuum. In fact, it's arguable that it's impossible for RAPM alone to assess players, because players can benefit/suffer from things such as position played, different levels of offensive primacy, and so on.

Teammates - even though RAPM attempts to parse impact from teammates, there will be issues based on playing style and fit that occur. For example, although Nash was better as a Sun than as a Maverick, his RAPM scores exploded in a way things such as his scoring, assist/turnover ratio and many of his advanced statistics (eg PER) didn't. I think that some of this explosion is due to him being a better player/being more "enabled" by D'Antoni, but also because those Suns teams were helpless without Nash. A guy like Marion/Stoudemire would be INCREDIBLY efficient with Nash on the court, but drop off considerably without him. Whereas whilst Nash was playing with Dirk (who is an offensive anchor himself), the effect that Nash has is reduced, because... well, Dirk is awesome. In other words, Nash had a higher RAPM with the Suns, partially because his playmaking skill increased the efficiency of certain players (finishers that aren't shot creators) over others such as Dirk.

Versatility - RAPM does funny things with versatility. A guy who can defend the perimeter and the rim at a +2 level might have value that exceeds that of a player that defends the perimeter at a +2.5 level, but can't defend the rim (and vice versa) because his presence might allow for a greater variety of lineups to be played. Depending on the rest of the team, and lineups, this could either give his RAPM a huge boost, or a hit relative to other defenders (this kind of ties in with role).

Variability - a guy like Klay Thompson has a solid offensive RAPM, but you never know which Klay is going to show up that day. He relies heavily on the 3 ball, so Klay could go for 37 in a quarter, or shoot 2/14. Compare him to a guy like OKC Durant, who was consistently good for 20+ points, but seldom had those "explosion games." In other words, even if two players have the same average effect on scoring margins (which is what RAPM will consider), it doesn't tell you how likely they are to be close to that level of impact. After all, a player whose impact oscillates between -5 and +10 isn't going to be the same as a player whose impact oscillates between 0 and +5.

Replaceability and diminishing returns - a guy who is +6 on offence and -1 on defence has the same RAPM as a guy that is +2.5 on offence and +2.5 on defence, but this doesn't necessarily mean that they have the same impact. One could argue that the second guy is more valuable because there are less diminishing returns on offence from him, but one could also argue that the first guy is SUCH a rare commodity in the NBA that it's easier to build a team around a +6 offence guy, because, well, it's easier to find slightly above average defensive guys than it is to find a +6 offence guy.

Fouls - as a Rockets fan, I'd like to take a moment to highlight the two main guys on my team right now. Harden is stupidly good at drawing fouls on offence, and CP3 is stupidly good at ticky tacky fouls that can bench the player he's guarding. I've seen CP3 take Curry off the court because he draws 2 quick fouls at the start of the game, which provides a boost to his team. But, RAPM doesn't consider that fouls can force lineup changes, so the value in being able to minimise the minutes of a strong opposing player goes untouched.

Minutes - self explanatory, I think. Not all players can play at the same level for the same amount of minutes.

Honestly, there's so much more. But in closing, I say this -

RAPM is incredibly useful, and really, what matters in the NBA is how much a player can impact winning. But there are most definitely limitations to how much RAPM can explain on its own. The box score + scrutinising actual play with the eye test can work wonders in understanding the RAPM results and helping RAPM shape ones opinion of a player. I don't think that it should necessarily be thrown out because a player is putting up stupid box score numbers, or because they look amazing, but it's an extremely valuable piece of the puzzle that can prompt extra thought, extra analysis, and possibly highlight certain elements of ones game that might otherwise go unnoticed without its influence.

It's not about preconceived bias, but rather, aiming to improve the understanding of a player. Honestly, KG's gargantuan RAPM scores might overrate him. You're not necessarily wrong at all in thinking this. But it's worth noting that he does a LOT of things (elite defensive versatility, floor spacing from a big) that aren't captured in the box score, and I think that without RAPM, our understanding of what makes KG (and many other players) so special wouldn't be quite as thorough.

In fact, speaking to a Kobe guy, I think it's worth noting that RAPM portrays Kobe in a better light than most "advanced" box score statistics do, and it's partially because things such as Kobe's offensive gravity are captured by RAPM, and aren't captured in most other metrics.

[Dr Spaceman]

.

Great to see you back man, and great to see you haven’t lost a step

[E-Balla]

Where did you get the RAPM?

https://docs.google.com/spreadsheets/d/e/2PACX-1vSzp3G5rwP9xgCgluVGmR3Qj4-BMoGSYiuTKM6o_pzES6s95oQE1nQvB2CXed-4fRc_MMGgpULtDaJ_/pubhtml?gid=1825430955&single=true

Is this your own version of RAPM?

Not at all but it is up to date RAPM.

Two points.

TS%

I never used TS% to claim that Westbrook isn't an elite offensive player. He clearly is. I also never claimed that TS% was the ultimate indicator of efficiency, that shooting efficiency is the only thing that matters, or that TS% captures a player's total offensive impact. My point is that Westbrook's low TS% is an indicator of lower efficiency that is a weakness in his POY resume compared to other candidates.

I don't see why that's controversial. I agree that his poor shooting efficiency is mitigated by other benefits that come with his aggressive style of play, but what we're talking about is how Westbrook compares to the very best players in the league. When several of them are also excellent at creating opportunities for teammates and have significantly better shooting efficiency, that shows a weakness in Westbrook's performance relative to other POY candidates.

Its controversial because after mentioning his shooting efficiency the next step is to explain why that is important. You failed to mention that at all or explain it. You just said he has a 51 TS%, that "lowers his case", and left it at that. If you can't explain why it matters why mention it?

How RAPM is used

I agree that RAPM isn't a straight ranking system and that it's up to people to interpret RAPM correctly and present RAPM data correctly in their arguments. Unfortunately, what RAPM says about a player requires quite a bit of context, and from what I've seen in the recent part of this thread is that context is rarely presented with the RAPM. Instead, I've seen things like E-Balla calling Jamal Crawford the worst 20 MPG player in the league because he's 489th out of 493 in RAPM. That violates what you say about RAPM on two counts: 1) it uses RAPM as a ranking system; and 2) it doesn't provide context.

Here's what I've learned from my introductory research on RAPM:

-- You apparently need to be a data analytics expert to fully understand it.

-- It's designed to use past data to predict future performance (kind of the opposite of the caveat on an investment brochure).

-- It uses ridge regression to increase the predictive accuracy of RAPM on future performance.

-- Jeremias Engelmann seems to be the godfather of RAPM and is the expert behind ESPN's RPM.

-- There are multiple versions of RAPM.

-- The formula used to calculate RAPM isn't made public.

-- RAPM may be calculated using non-performance factors such as height.

Please feel free to correct me on any of the above that I got wrong.

I have yet to find a fan-level explanation of what RAPM is and what it tells you. I've seen repeatedly that it requires context, but I haven't found an explanation of what that context is, how to determine what the context is for a particular player, or how to properly use RAPM when comparing players.

Given all of the above, RAPM comes across to a non-data analyst as a black box that spits out numbers that we're supposed to accept as the best metric to assess players and that it trumps everything else. Maybe it is that good, but without the necessary understanding that is apparently impossible for a non-data analyst to acquire and seeing how RAPM data is presented by those who promote it and claim to understand it, I'm reluctant to accept it blindly on faith, especially when it runs counter to other data that I have available.

I'll keep working to expand my understanding of RAPM, but those who create it and those who promote its use seem to make that really hard.

I used Jamal Crawford playing like ass cheeks to say he plays like ass cheeks. You mentioned his PER as if that meant something and the fact that he is playing over Shabazz so I rebuttaled with the RAPM rankings of Crawford and Shabazz to show you both of them are on the shortlist of players with the most negative effect on their team's point differential. No one is using it as a ranking system just as evidence that what I see when I watch games is obviously true and Jamal Crawford is trash. I could've easily pointed to he and Shabazz's horrible on/off and got the same point across.

Also you're completely mistaken about RAPM. You described RPM. RAPM is a completely different thing. Height has nothing to do with it, the factors in RAPM are extremely simple (it only uses +/- data), calculating RAPM is not that hard with the right software (especially with multiple tutorials available online), and all RAPM does is determine how a player affects the point differential in their currently role on their current team.

[dhsilv2]

Phenomenal post (other than the WB part but that's minor).

I am getting a masters in applied economics which is 50% statistics. I spend a lot of time on R, so I know my way around regressions to a certain extent. And I still don't know how they arrive at the numbers RAPM spits out, because I've never found a good explanation, or any explanation whatsoever. Yet RAPM is still this magnificent fountain of truth we aren't supposed to question at the risk of ridicule (which is very evident in this thread as you can see).

I believe the reason it rose to such prominence on this board is because a certain player whom I will not name had a contingent of fervent supporters who realized all other arguments for him weren't good enough, and thus located an obscure stat that placed this certain unnamed player on a pedestal and then began to spread its virtues to the masses so that this certain unnamed player would gain more notoriety.

Trust me I've been around for a while, this movement has been years in the making. The effort to normalize RAPM as the be all end all stat is primarily, I believe, so that more and more people will see said unnamed player as being far above his station.

Sent from my [device_name] using [url]RealGM mobile app[/url]

ESPN has their own version of RAPM in their stats and their writers use it constantly...the idea that it's something that this board is trying to "normalize" is absurd. It's being used everywhere, your grandmother will be using it before long!

[bondom34]

PER is no better I agree but the raw stats are consistent. If Harden leads the league in scoring, it's because he scores the most points. It doesn't claim otherwise, it's honest.

RAPM is deceptive af and can be used to push agendas like nobody's business.

RAPM doesn't claim anything otherwise either. It's a list of the players whose teams see the biggest difference in point differential when they're on the court, with some adjustments for team context. It's not pretending to be anything else either.

Just like with any other stat, it's up to the people using it to interpret it correctly. Raw stats are manipulated like crazy too. Like how TS% has been used in this thread to act like Westbrook isn't among the elite offensive players in the league.

Two points.

TS%

I never used TS% to claim that Westbrook isn't an elite offensive player. He clearly is. I also never claimed that TS% was the ultimate indicator of efficiency, that shooting efficiency is the only thing that matters, or that TS% captures a player's total offensive impact. My point is that Westbrook's low TS% is an indicator of lower efficiency that is a weakness in his POY resume compared to other candidates.

I don't see why that's controversial. I agree that his poor shooting efficiency is mitigated by other benefits that come with his aggressive style of play, but what we're talking about is how Westbrook compares to the very best players in the league. When several of them are also excellent at creating opportunities for teammates and have significantly better shooting efficiency, that shows a weakness in Westbrook's performance relative to other POY candidates.

Just replying to this as it's all I've been involved with here, but when comparing him to these players, I don't thin kanyone says he's better offensively than Curry. Lebron this season hasn't been particularly effective, and Harden's impact was greater to start the season then declined a bit more recently. Even when comparing him to these 3, he's comparable to at the least 2 of them. Not really sure anyone has an argument strictly offensively over Curry in the last 3 years.

[Dr Spaceman]

Hey guys, so after a long hiatus I have decided to get back to writing a little bit to scratch that creative itch, and my first post is dealing mostly with Harden and the Rockets and POY-relevant stuff:

https://rootingforgoliath.wordpress.com/2018/01/31/the-harden-paradox/

I’m hoping to be much more consistent this time around, but since its working for free I make no guarantees

But yeah, feedback, discussion, everything is welcome.

[bondom34]

Hey guys, so after a long hiatus I have decided to get back to writing a little bit to scratch that creative itch, and my first post is dealing mostly with Harden and the Rockets and POY-relevant stuff:

https://rootingforgoliath.wordpress.com/2018/01/31/the-harden-paradox/

I’m hoping to be much more consistent this time around, but since its working for free I make no guarantees

But yeah, feedback, discussion, everything is welcome.

Good read, and I agree with the thoughts behind it (I don't think anyone touches GSW in general though). Interesting thoughts on the roboticness (is that a word?) of the offense and how it may sort of backfire come postseason, along w/ making mistakes helping to learn. I think OKC's shot went out the window w/ Roberson so they're a shoe in to the WCF barring injury, but if Kawhi were around I'd still lean Spurs myself. 2 notes, first they do have a loss (NOLA a few games ago) with CP3, Harden, and Capela, and second and my apologies but every time I see your name I think of Boy Meets World .