Advanced Stats Knowledge Pooling

[Pharaoh]

As a guy who doesn't use advanced stats or even traditional stats to rate players I have a question:

Are there any stats that would indicate that Tobias Harris is "under-rated"?

He seems to suffer from the Chauncey Billups syndrome - traded a lot so until (unless) he does something monumental he's going to be viewed as "less than"

I'm wondering if my eyes deceive me

[Wigginstime]

What is the best advanced stat metric for comparing players: PER vs. BPM vs. WS/48 vs. VORP
I would rate them as follows:
1. BPM
2. VORP
3. WS/48
4. PER

For reference here is the top 10 players in 2017 for each stat:
PER:
1. Russell Westbrook • OKC 30.6
2. Kevin Durant • GSW 27.6
3. Kawhi Leonard • SAS 27.6
4. Anthony Davis • NOP 27.5
5. James Harden • HOU 27.4
6. LeBron James • CLE 27.0
7. Isaiah Thomas • BOS 26.5
8. Nikola Jokic • DEN 26.3
9. Chris Paul • LAC 26.2
10. Giannis Antetokounmpo • MIL 26.1

BPM:
1. Russell Westbrook • OKC 15.6
2. James Harden • HOU 10.1
3. Chris Paul • LAC 8.8
4. LeBron James • CLE 8.4
5. Nikola Jokic • DEN 8.4
6. Kevin Durant • GSW 8.0
7. Kawhi Leonard • SAS 7.9
8. Giannis Antetokounmpo • MIL 7.6
9. Stephen Curry • GSW 7.3
10. Jimmy Butler • CHI 6.9

WS/48:
1. Kevin Durant • GSW .278
2. Chris Paul • LAC .264
3. Kawhi Leonard • SAS .264
4. Rudy Gobert • UTA .250
5. James Harden • HOU .245
6. Jimmy Butler • CHI .236
7. Isaiah Thomas • BOS .234
8. Stephen Curry • GSW .229
9. Nikola Jokic • DEN .228
10. Russell Westbrook • OKC .224

VORP:
1. Russell Westbrook • OKC 12.4
2. James Harden • HOU 9.0
3. LeBron James • CLE 7.3
4. Giannis Antetokounmpo • MIL 6.9
5. Jimmy Butler • CHI 6.3
6. Stephen Curry • GSW 6.2
7. Kawhi Leonard • SAS 6.2
8. Rudy Gobert • UTA 5.4
9. Nikola Jokic • DEN 5.3
10. DeMarcus Cousins • TOT 5.3

[JXL]

Hot damn, does Jokic have a good season ahead of him... put it down that the Nuggets make the playoffs as a 5-6 seed.

[jinxed]

You left out the best stat. RPM. They've done studies on this based on their ability to predict outcomes. It's RPM.

In order
1. RPM
2. RAPM
3. BPM
4. WS
5. PER

VORP is simply BPM x Minutes played. They're really the 'same' stat. Like Win Shares ans WS/48 or RPM and RPM Wins.

[Cavsfansince84]

I don't think one stands out as being that much better than the others so long as we are talking about players who all play 25+ mpg for the sake of ws/48. I think all 4 combined tend to paint a pretty good picture though of the kind of season a player had.

[Wigginstime]

I don't think one stands out as being that much better than the others so long as we are talking about players who all play 25+ mpg for the sake of ws/48. I think all 4 combined tend to paint a pretty good picture though of the kind of season a player had.

It is interesting at how different the results are at times.

Take Golden State with Curry vs. Durant

Durant is #2 in PER, #6 in BPM, #1 in WS/48, & no in the top 10 in VORP
Curry is #6 in VORP, #8 WS/48, #9 BPM, not in the top 10 in PER

Affectively the combination of the advanced stats would say that Durant is the better player overall; however, Curry is most important to the team and the team struggles more with him off the court than Durant

[Cavsfansince84]

It is interesting at how different the results are at times.

Take Golden State with Curry vs. Durant

Durant is #2 in PER, #6 in BPM, #1 in WS/48, & no in the top 10 in VORP
Curry is #6 in VORP, #8 WS/48, #9 BPM, not in the top 10 in PER

Affectively the combination of the advanced stats would say that Durant is the better player overall; however, Curry is most important to the team and the team struggles more with him off the court than Durant

Durant's vorp was lower due to missing like 20 games though. Durant was really efficient. Curry is perhaps more valuable to them and you could even make the argument that Durant is the third most valuable player on that team but that's a hard thing for metrics to point out given the system they use there.

[dhsilv2]

What is the best advanced stat metric for comparing players: PER vs. BPM vs. WS/48 vs. VORP
I would rate them as follows:
1. BPM
2. VORP
3. WS/48
4. PER

For reference here is the top 10 players in 2017 for each stat:
PER:
1. Russell Westbrook • OKC 30.6
2. Kevin Durant • GSW 27.6
3. Kawhi Leonard • SAS 27.6
4. Anthony Davis • NOP 27.5
5. James Harden • HOU 27.4
6. LeBron James • CLE 27.0
7. Isaiah Thomas • BOS 26.5
8. Nikola Jokic • DEN 26.3
9. Chris Paul • LAC 26.2
10. Giannis Antetokounmpo • MIL 26.1

BPM:
1. Russell Westbrook • OKC 15.6
2. James Harden • HOU 10.1
3. Chris Paul • LAC 8.8
4. LeBron James • CLE 8.4
5. Nikola Jokic • DEN 8.4
6. Kevin Durant • GSW 8.0
7. Kawhi Leonard • SAS 7.9
8. Giannis Antetokounmpo • MIL 7.6
9. Stephen Curry • GSW 7.3
10. Jimmy Butler • CHI 6.9

WS/48:
1. Kevin Durant • GSW .278
2. Chris Paul • LAC .264
3. Kawhi Leonard • SAS .264
4. Rudy Gobert • UTA .250
5. James Harden • HOU .245
6. Jimmy Butler • CHI .236
7. Isaiah Thomas • BOS .234
8. Stephen Curry • GSW .229
9. Nikola Jokic • DEN .228
10. Russell Westbrook • OKC .224

VORP:
1. Russell Westbrook • OKC 12.4
2. James Harden • HOU 9.0
3. LeBron James • CLE 7.3
4. Giannis Antetokounmpo • MIL 6.9
5. Jimmy Butler • CHI 6.3
6. Stephen Curry • GSW 6.2
7. Kawhi Leonard • SAS 6.2
8. Rudy Gobert • UTA 5.4
9. Nikola Jokic • DEN 5.3
10. DeMarcus Cousins • TOT 5.3

VORP is just BPM x minutes...ok not exactly but it's the same stat.

As for the question, I'd currently say of stats that are on major websites, WINS from espn is the best metric we have.

http://www.espn.com/nba/statistics/rpm/_/sort/WINS

[dhsilv2]

You left out the best stat. RPM. They've done studies on this based on their ability to predict outcomes. It's RPM.

In order
1. RPM
2. RAPM
3. BPM
4. WS
5. PER

VORP is simply BPM x Minutes played. They're really the 'same' stat. Like Win Shares ans WS/48 or RPM and RPM Wins.

RAPM has two forms. Prior informed and non prior informed. RPM is non prior informed (as best I understand it) XRAMP. But there is still prior informed XRAPM.

I have a pretty hard time believing RPM beat out prior informed XRAMP or even prior informed RAPM consistently.

Though even hollinger's value add stuff had a few good years, which was based on his PER.

[dhsilv2]

I don't think one stands out as being that much better than the others so long as we are talking about players who all play 25+ mpg for the sake of ws/48. I think all 4 combined tend to paint a pretty good picture though of the kind of season a player had.

It is interesting at how different the results are at times.

Take Golden State with Curry vs. Durant

Durant is #2 in PER, #6 in BPM, #1 in WS/48, & no in the top 10 in VORP
Curry is #6 in VORP, #8 WS/48, #9 BPM, not in the top 10 in PER

Affectively the combination of the advanced stats would say that Durant is the better player overall; however, Curry is most important to the team and the team struggles more with him off the court than Durant

Advanced "box score" stats. Curry has a pretty large lead in impact stats that I've seen.

[blabla]

RAPM has two forms. Prior informed and non prior informed. RPM is non prior informed (as best I understand it) XRAMP. But there is still prior informed XRAPM.

I have a pretty hard time believing RPM beat out prior informed XRAMP or even prior informed RAPM consistently.

Though even hollinger's value add stuff had a few good years, which was based on his PER.

You got a lot of things mixed up.
When people say "RAPM" they generally refer to the "pure" form, which is non prior informed. The prior informed version is called RPM, which is identical to xRAPM

[dhsilv2]

RAPM has two forms. Prior informed and non prior informed. RPM is non prior informed (as best I understand it) XRAMP. But there is still prior informed XRAPM.

I have a pretty hard time believing RPM beat out prior informed XRAMP or even prior informed RAPM consistently.

Though even hollinger's value add stuff had a few good years, which was based on his PER.

You got a lot of things mixed up.
When people say "RAPM" they generally refer to the "pure" form, which is non prior informed. The prior informed version is called RPM, which is identical to xRAPM

I have never heard people reference RAPM in terms of being 1 year season regression thus when I see the stats posted places I see RAPM or NPI RAPM. RPM is xRAPM which uses a mix of BOX SCORE data in with RAPM. I guess you could call that "prior" informed too, but as I understand what ESPN puts out, it's based on a single season, so I don't see any "prior" unless the box score data is the prior year and the plus minus data is current year, and then that's just weird. So I would assume there is a prior informed xRAPM and a non prior informed xRAPM. ESPN using a non prior informed one as it's a single season metric.

[WarriorGM]

None of those stats or the way they are presented closely match the commonly held to be best players in the league. I think there are currently 4 really transcendent players. An accurate stat or methodology for me would be able to do a good job of identifying those 4 ahead of the rest or give an understandable depiction of how they stand relative to the other players. The above stats for me overrate Davis, CP3, and Westbrook.

[dhsilv2]

None of those stats or the way they are presented closely match the commonly held to be best players in the league. I think there are currently 4 really transcendent players. An accurate stat or methodology for me would be able to do a good job of identifying those 4 ahead of the rest or give an understandable depiction of how they stand relative to the other players. The above stats for me overrate Davis, CP3, and Westbrook.

Who are the 4? And why are you so certain? Also which stats? A lot of stats have been disccussed, many don't think much of westbrook, I can't recall davis being ranked that highly on anything to be honest.

[WarriorGM]

None of those stats or the way they are presented closely match the commonly held to be best players in the league. I think there are currently 4 really transcendent players. An accurate stat or methodology for me would be able to do a good job of identifying those 4 ahead of the rest or give an understandable depiction of how they stand relative to the other players. The above stats for me overrate Davis, CP3, and Westbrook.

Who are the 4? And why are you so certain? Also which stats? A lot of stats have been disccussed, many don't think much of westbrook, I can't recall davis being ranked that highly on anything to be honest.

In alphabetical order I would say Curry, Durant, James, and Leonard. All 4 have shown they can win. All 4 have won rings or carried a team to the finals or have led a team to 60+ games and win against everyone else. Just look at the most recent playoffs to see why Leonard>Harden>Westbrook.

I am not dogmatic though. This is just a conceptual framework. If a stat or combination of stats can draw a picture that has players not in this top tier looking better than these four that makes logical sense then of course it should be considered. If these players are not represented though in an understandable way then I must suspect the stats in question probably aren't very representative.

Westbrook leads 3 of these stats we are talking about and Davis is sometimes mentioned as a top five player despite making the playoffs only once where his team was swept in the first round so I find him a particularly egregious case.

[dhsilv2]

None of those stats or the way they are presented closely match the commonly held to be best players in the league. I think there are currently 4 really transcendent players. An accurate stat or methodology for me would be able to do a good job of identifying those 4 ahead of the rest or give an understandable depiction of how they stand relative to the other players. The above stats for me overrate Davis, CP3, and Westbrook.

Who are the 4? And why are you so certain? Also which stats? A lot of stats have been disccussed, many don't think much of westbrook, I can't recall davis being ranked that highly on anything to be honest.

In alphabetical order I would say Curry, Durant, James, and Leonard. All 4 have shown they can win. All 4 have won rings or carried a team to the finals or have led a team to 60+ games and win against everyone else. Just look at the most recent playoffs to see why Leonard>Harden>Westbrook.

I am not dogmatic though. This is just a conceptual framework. If a stat or combination of stats can draw a picture that has players not in this top tier looking better than these four that makes logical sense then of course it should be considered. If these players are not represented though in an understandable way then I must suspect the stats in question probably aren't very representative.

Westbrook leads 3 of these stats we are talking about and Davis is sometimes mentioned as a top five player despite making the playoffs only once where his team was swept in the first round so I find him a particularly egregious case.

But westbrook finishes fairly poorly in some of the stats as well, or at least outside of the top 5. Davis was first team all nba, but that's due to calling him a center, not him being top 5. Though he's got a coach who i'm pretty sure is trying to lose, so it's hard for me to judge Davis on winning. Making the playoffs two years ago imo was pretty impressive though. That said Davis is normally ranked more like 8th or further back.

Harden, Westbrook, and Paul are almost always ranked ahead of Davis.

As for KD....westbrook was on his team when he did all those things you mentioned, unless you're giving him credit for the title last year which seems pretty amazing. I think KD is a great player, but your argument doesn't really work for him and not for westbrook.

[WarriorGM]

As for KD....westbrook was on his team when he did all those things you mentioned, unless you're giving him credit for the title last year which seems pretty amazing. I think KD is a great player, but your argument doesn't really work for him and not for westbrook.

You are right that KD being paired with Westbrook on OKC makes isolating their respective contributions a bit more complicated. But the eye test and various other indicators and outcomes point to KD being superior than Westbrook. I wasn't paying much attention to OKC so I might have some facts wrong but my impression is KD scored more and more efficiently during OKC's run to the finals in 2012 and that was the case in general on the Thunder. The season KD had an injury the Thunder missed the playoffs. While the Warriors being such a strong a team even without KD puts a cap on how much credit KD can rightfully claim, the all-time best playoff record and great finals performance doesn't hurt either.

One can also just choose to ignore KD and focus on Curry, James and Leonard in calibrating or judging the accuracy or representativeness of given stats.

I am simply advocating a process of using a stat (wins) to confirm an eye test (these players) to confirm other stats (advanced stat of your choice).

[blabla]

I have never heard people reference RAPM in terms of being 1 year season regression thus when I see the stats posted places I see RAPM or NPI RAPM. RPM is xRAPM which uses a mix of BOX SCORE data in with RAPM. I guess you could call that "prior" informed too, but as I understand what ESPN puts out, it's based on a single season, so I don't see any "prior" unless the box score data is the prior year and the plus minus data is current year, and then that's just weird. So I would assume there is a prior informed xRAPM and a non prior informed xRAPM. ESPN using a non prior informed one as it's a single season metric.

When people call it "prior informed" it has nothing to do with the number of years used. That'd be the difference between e.g. "single season RAPM" and "multiseason RAPM".
RPM is generated by creating a BoxScore metric (like BPM), then using that as "prior information" before plugging it into an RAPM-type regression.
RAPM that's not prior informed would just ignore all the BoxScore information

[dhsilv2]

I have never heard people reference RAPM in terms of being 1 year season regression thus when I see the stats posted places I see RAPM or NPI RAPM. RPM is xRAPM which uses a mix of BOX SCORE data in with RAPM. I guess you could call that "prior" informed too, but as I understand what ESPN puts out, it's based on a single season, so I don't see any "prior" unless the box score data is the prior year and the plus minus data is current year, and then that's just weird. So I would assume there is a prior informed xRAPM and a non prior informed xRAPM. ESPN using a non prior informed one as it's a single season metric.

When people call it "prior informed" it has nothing to do with the number of years used. That'd be the difference between e.g. "single season RAPM" and "multiseason RAPM".
RPM is generated by creating a BoxScore metric (like BPM), then using that as "prior information" before plugging it into an RAPM-type regression.
RAPM that's not prior informed would just ignore all the BoxScore information

Math maybe needs new words. RAPM is normally a reference to a multi year regression. rpm is not. If same year box score is "prior", we should not use that term. It makes understanding the stat more difficult for no reason.

[blabla]

Math maybe needs new words. RAPM is normally a reference to a multi year regression. rpm is not. If same year box score is "prior", we should not use that term. It makes understanding the stat more difficult for no reason.

You can keep on believing what you want to believe, but a regression technique doesn't get a new name because there's a different number of years used.
When people say "prior" informed, it's always with a BoxScore based metric
RAPM, in it's pure form, doesn't have any priors, and the fact that it gets computed over several years doesn't mean it has a prior

There's also single year RAPM, and multiyear RPM