Nerlens Noel Thread

[76ciology]

My guess is that the Wolves decision to tank by resting their good players and others to the point of hardly having the required number of guys required to play a game (low overall impact with or without wiggins) and Wiggins' not excelling in other aspect aside from scoring, where it isn't good enough to compensate for it, did contribute for the poor advanced stats.

But nevertheless, Wiggins did show some glimpse of a promising future ahead of him. I like his potential to be a good individual defender, his ability to get to the line and his ability to finish around the rim.

[Sixerscan]

Wiggins' awful defensive numbers really don't pass the eye for me. It's fine info, but I haven't been convinced that these overarching defensive stats have been able to properly correct for quality of teammates yet. Regressions can only do so much.

Minny was the worst defensive team in the league, obviously Wiggins had some role in that, but so does that dreck of a roster. Dieng was pretty much the only other guy on their roster that played a meaningful amount of games that wasn't a sieve. He's no Noel and definitely has a lot to work on but he was a fine defensive player this year.

As far as his offensive role, Bodner already touched on the tough role he had. If Noel was responsible for that level of offense he would have shot 30% from the field.

I'm really excited about Nerlens, but the voters got this one right IMO.

[CoreyGallagher]

In what world? MCW's 48% true shooting percentage was far less than Wiggins' 51.7%.

Wiggins efficiency, considering age, usage, and dearth of offensive talent on that team, was actually not bad, and it improved as the year went on and Wiggins expanded his post-up game and got to the line more (53.3% on a 24.3% usage rate in the 2nd half).

Noel had the bigger impact this year IMO, but, for the role Wiggins played, his season was a success. Remember, even all-time greats like LeBron James (48.8% true shooting, 5.1 win shares, 1.9 bpm, 3.1 vorp) and Kevin Durant (51.9% true shooting, 2.3 win shares, -1.4 bpm, 0.4 vorp) struggled (relatively speaking) in their ability to score in volume during their rookie seasons. For kids that age, and in that role, usage is as much a predictor of future success as efficiency is, and Wiggins didn't fall on his face in terms of efficiency

BPM and VORP, noticeably. Certainly not their shooting and finishing.

I wasn't disparaging Wiggins' future by making the comparison, it was strictly numerical and of that of NBA Reference because I'm lazy. I also wasn't comparing their ROY candidacies in that there's usually more to consider than just the numbers themselves. Imo Wiggins' season was a success as well, just so happens that he had stiffer competition than MCW did last season.

[marcush]

Meh, thankfully this ROY thing is out of the way. Now we can just get back to being simple old Noel / Sixer homers without having tO talk down Wiggins at every opportunity .

[ABDIW76ers]

Always liked Wiggins, was hoping he'd be a Sixer last year, but I'm happy with the Embiid pick as of right now.

I do sometimes wonder what it would take to get Wiggins from the T-Wolves, but it'd take too much and probably wouldn't be worth it. I'd love his fit with the team though. Not really considering it, just a fun, random, impossible scenario.

[76ers 2020]

Wiggins' awful defensive numbers really don't pass the eye for me. It's fine info, but I haven't been convinced that these overarching defensive stats have been able to properly correct for quality of teammates yet. Regressions can only do so much.

I'm a fan of the defensive differential percentage tracker on NBA.com. My go to page for looking at an individual's defensive performance currently. The description is in the Leonard picture. A few screenshots.

[76ciology]

I'm probably the biggest critic of Wiggins in realgm.

Personally, I think the reason why there are conflicting opinions about Wiggins is because of the different standards of expectations of Wiggins and the raw stats vs advanced stats argument.

We are in this era in the NBA where in we are witnessing two all-time great swingman in KD and LBJ. It's just unfair, IMO, to expect that he will be in the same echelon with them.

Wiggins is limited on my eyes. He is simply a post and off ball scorer and a lock down defender. Not much facilitating, rebounding and the other stuffs.

He doesn't create and facilitate, doesn't shoot much from 3s, doesn't block or rack up steals, doesn't rebound at a good rating RB%. IMO, Wiggins is more in the same echelon with Derozan, Paul George and Melo (20-22 PER guys). He will never be in the same level with personally IMO, the greatest player in the NBA ever in LBJ. LBJ is just elite on aspects of the game. While KD has always been elite with his scoring even in college. And we've seen how limited Wiggins' offense was in college (see dean on draft; dean demakis). T-MAC always has point guard handles. Kobe always has his superstar instincts and moves. MJ can do it all like LBJ. Wiggins is limited. To me he's Melo on offense at the very best with lock down defense. And I don't know how many teams would pass up having a guy like that.

On eyetest, being a scorer and lock down defender is enough. But on advanced stats, you need to be able to do the other stuffs on offense such as creating and facilitating (OBPM, ORTG). On defense, you have to be a presence overall not just a lockdown defender (DBPM and DRTG). You also need to do the miscellaneous stuffs like blocks, rebounding and steals (PER).

Wiggins effect in the game is limited, superstars effect the game almost everywhere. That's how I look at it. Superstars can either do it all or just dominate a game. Wiggins is more of a one on one guy on both ends of the floor (watch his struggles in scoring post all-star when defenders played physical on him).

Now, Wiggins is just 20 so I'm not burning villages to spread my message. I know there's a long road ahead of him. His length and athleticism is elite, thus most people see his potential to be infinite (I'm reading some posts saying he's going to be one of the top guys in the league). For now, I'm not going to close the book at him. But my projection at him after I did my valuation of him pre-draft remains unchanged, I think it's very unlikely for him to be a superstar and I do think Noel provides better overall presence than him (dominant defense and slightly below average to average offense over good offense and defense)

[tk76]

We are in this era in the NBA where in we are witnessing two all-time great swingman in KD and LBJ. It's just unfair, IMO, to expect that he will be in the same echelon with them.

This is a fair point, in that LBJ is one of the best all around players and KD one the most efficient scorers. Wiggins is unlikely to ever be in that conversation. But that is the case most years- that there is no generational talent in the ROY discussion. Same for Nerlens, who has a chance of being an all time defender, but will never be in the same conversation as Hakeem, David Robinson or Shaq.

I guess you are right in that there are no true all time greats at center that are currently in their prime the way LBJ and KD are. But there also are no all time great SG's in their prime... and so that opens a chance for Wiggins to be the best SG in the NBA in a few years, especially if his shooting improves... and you can play him next to a dominant do everything SF.

My point is that Wiggins lacking playmaking skills is not necessarily a negative. There is tremendous value in having a great player who is not ball dominant. Lebron and DWade were great enough talents to make it work in Miami- but in general you don't want a bunch of ball dominant stars fighting for point duties on the same team. And I see wiggins developing into an elite player who can still thrive on the same floor as an elite PG or point forward. As opposed to someone like Kobe, who you had to pair with complementary PG/SF's due to his ball dominant ways. Same with Melo and to an extent DWade. lebron and Durant are exceptions, partly because they are such special players.

[76ciology]

We are in this era in the NBA where in we are witnessing two all-time great swingman in KD and LBJ. It's just unfair, IMO, to expect that he will be in the same echelon with them.

This is a fair point, in that LBJ is one of the best all around players and KD one the most efficient scorers. Wiggins is unlikely to ever be in that conversation. But that is the case most years- that there is no generational talent in the ROY discussion. Same for Nerlens, who has a chance of being an all time defender, but will never be in the same conversation as Hakeem, David Robinson or Shaq.

I guess you are right in that there are no true all time greats at center that are currently in their prime the way LBJ and KD are. But there also are no all time great SG's in their prime... and so that opens a chance for Wiggins to be the best SG in the NBA in a few years, especially if his shooting improves... and you can play him next to a dominant do everything SF.

My point is that Wiggins lacking playmaking skills is not necessarily a negative. There is tremendous value in having a great player who is not ball dominant. Lebron and DWade were great enough talents to make it work in Miami- but in general you don't want a bunch of ball dominant stars fighting for point duties on the same team. And I see wiggins developing into an elite player who can still thrive on the same floor as an elite PG or point forward. As opposed to someone like Kobe, who you had to pair with complementary PG/SF's due to his ball dominant ways. Same with Melo and to an extent DWade. lebron and Durant are exceptions, partly because they are such special players.

I don't know if there's a shortage of star to superstar swingmans in the league or KD/LBJ has raised the standard at a ridiculous level. But if Wiggins is to be the best SG in the league, it is not because of talent but more of the lack of competition. But nevertheless, PER and BPM will continue to haunt Wiggins IMO, unlike he either becomes a dominant scorer/defender who can make his teammates better or a well rounded player. Shouldn't Wiggins be racking (padding) up ridiculous amount of stats with how blatant their tank job was to end the season? Will his rebounding and assists number be lower with a KAT/Pekovic and Rubio?

I know you guys hate PER. But I still think PER is a good stats to determine ordinary guys from stars/superstars. And I don't see wiggins breaking the 20-21 PER resistance (Paul George/Melo echelon) until some he improve in other aspects in his game.

[Sixerscan]

Wiggins' awful defensive numbers really don't pass the eye for me. It's fine info, but I haven't been convinced that these overarching defensive stats have been able to properly correct for quality of teammates yet. Regressions can only do so much.

I'm a fan of the defensive differential percentage tracker on NBA.com. My go to page for looking at an individual's defensive performance currently. The description is in the Leonard picture. A few screenshots.

Spoiler:

Does that adjust for the help defense behind the defender? Doesn't seem like it does. Or is it just about isolations? (Which is only a small part of defense)

I totally get how advanced stats can tell you how a 5 man unit defends. Have serious reservations about it proving much about individual defenders though. Moreover, I think a lot of people that push them are self interested (they want to be picked up by a team's or the NBA's analytical department, or they are a website trying to draw traffic) and so they or their boss overstate how valuable their formulas really are.

I certainly don't trust statistics where they don't explain the methodology behind them and I'm just supposed to blindly trust the data provider.

[HartfordWhalers]

I certainly don't trust statistics where they don't explain the methodology behind them and I'm just supposed to blindly trust the data provider.

There is tons of info out there about these stats and where they come from etc.

In terms of Wiggins defense:

1. The on-offs have him as a negative (for the 30th ranked defense), as does his Drating:

PPP Twolves allow with Wiggins off the court: 1.102
PPP Twolves allow with Wiggins on the court: 1.132 (3 points per 100 possessions worse)

(Calculated as straight how does the defense do with and without Wiggins)

Drating: 114 (19th on the Timberwolves who had a 112).
Calculated basically similarly, with a slight different on possessions etc.

2. The adjusted on-offs for opponent and teammate have him a net negative:
DRAPM: -1.26 points worse

(Calculated by regressing to see how every player does as a separate variable).

3. The Synnergy stats show him as a negative defender

"And Synergy Sports judged his capabilities as an individual defender only marginally better, ranking him in the 32nd percentile of defenders according to its video-scouting metrics."

Calculated by looking at video of every play in the NBA, and seeing how often his man scores. These are available as splits by possession type etc.

4. The box score based metrics* DBPM and DRPM all show him to be a negative defender.

DRPM: -2.16
DBPM: -1.8

DRPM* is a combination of the adjusted on offs,box score stats, and heuristic modeling rules including age, and height factors) that have fit the data best off historical data.

DBPM is a box score metric that has its weights derived from fitting what matched APM best. In essence, its the box score prediction for what someone's APM should be without the random noise the APM has, albeit also without capturing everything missed in the box score that is captured by APM.

Also, he looked bad per
5) The eye test.

I'm just not sure how anyone questions that he actually was bad, while playing bad defense and being forced to run more miles in teh NBA season as a 19-20 year old than anyone but Lillard.

[Sixerscan]

I certainly don't trust statistics where they don't explain the methodology behind them and I'm just supposed to blindly trust the data provider.

There is tons of info out there about these stats and where they come from etc.

In terms of Wiggins defense:

1. The on-offs have him as a negative (for the 30th ranked defense), as does his Drating:

PPP Twolves allow with Wiggins off the court: 1.102
PPP Twolves allow with Wiggins on the court: 1.132 (3 points per 100 possessions worse)

(Calculated as straight how does the defense do with and without Wiggins)

Drating: 114 (19th on the Timberwolves who had a 112).
Calculated basically similarly, with a slight different on possessions etc.

2. The adjusted on-offs for opponent and teammate have him a net negative:
DRAPM: -1.26 points worse

(Calculated by regressing to see how every player does as a separate variable).

3. The Synnergy stats show him as a negative defender

"And Synergy Sports judged his capabilities as an individual defender only marginally better, ranking him in the 32nd percentile of defenders according to its video-scouting metrics."

Calculated by looking at video of every play in the NBA, and seeing how often his man scores. These are available as splits by possession type etc.

4. The box score based metrics* DBPM and DRPM all show him to be a negative defender.

DRPM: -2.16
DBPM: -1.8

DRPM* is a combination of the adjusted on offs,box score stats, and heuristic modeling rules including age, and height factors) that have fit the data best off historical data.

DBPM is a box score metric that has its weights derived from fitting what matched APM best. In essence, its the box score prediction for what someone's APM should be without the random noise the APM has, albeit also without capturing everything missed in the box score that is captured by APM.

Also, he looked bad per
5) The eye test.

I'm just not sure how anyone questions that he actually was bad, while playing bad defense and being forced to run more miles in teh NBA season as a 19-20 year old than anyone but Lillard.

That quote was about that diff% stat, not any of those. Based on that quote, it's very ambiguous as to what it is really measuring. Is it just isolations? Does it adjust for help defense? Is it actually tracking who is guarding who even through switches or does it do an 82games-style lineup matchup and just assuming the "Shooting guard" is guarding the other "shooting guard"? I can't seem to find a more detailed qualitative explanation. If there is I would love to read it, and see if I agree with the methodology.

But as long as you bring it up, there actually is surprisingly little information about some of them. Sure, you can find a quick and easy write up about what those stats purport to represent. But when you try to find the actual formula for RPM? It's not publicly available, at least as of the last time I did a deep google search for it. Neither is any sort of confidence interval, error rate, relevant sample size, etc. Same issue with RAPM.

There's also the issue that, unlike PER, defense is inherently not based on box score stats. There's a lot of subjectivity, which makes having the actual data set even more important. We're basically being asked to put more and more trust in these people, who make money off of page clicks, not being right.

Does that not bother you? Basing your analysis on something where you have very little idea what the formula behind it is?

Sure, all of those stats seem to point in one direction. But what if they all have a common bias? Like maybe they don't properly balance out his awful teammates or the tough role he had to take properly. Net plus/minus and D rating absolutely don't. Maybe some of the more advanced ones do, but, again, without actually knowing the formula we can't know for sure. PER has its biases and we know it, so we know how to adjust. We don't with RPM.

It's not enough to say, "Oh they did a regression to adjust for that." There are tons of really crappy regressions done every day.

I would have to imagine that there are some stats that are very helpful for teams, but unless these guys want to release their formulae and data sets, as a public consumer of this stuff I have to take it with a pretty large grain of salt.

[HartfordWhalers]

I certainly don't trust statistics where they don't explain the methodology behind them and I'm just supposed to blindly trust the data provider.

There is tons of info out there about these stats and where they come from etc.

In terms of Wiggins defense:

1. The on-offs have him as a negative (for the 30th ranked defense), as does his Drating:

PPP Twolves allow with Wiggins off the court: 1.102
PPP Twolves allow with Wiggins on the court: 1.132 (3 points per 100 possessions worse)

(Calculated as straight how does the defense do with and without Wiggins)

Drating: 114 (19th on the Timberwolves who had a 112).
Calculated basically similarly, with a slight different on possessions etc.

2. The adjusted on-offs for opponent and teammate have him a net negative:
DRAPM: -1.26 points worse

(Calculated by regressing to see how every player does as a separate variable).

3. The Synnergy stats show him as a negative defender

"And Synergy Sports judged his capabilities as an individual defender only marginally better, ranking him in the 32nd percentile of defenders according to its video-scouting metrics."

Calculated by looking at video of every play in the NBA, and seeing how often his man scores. These are available as splits by possession type etc.

4. The box score based metrics* DBPM and DRPM all show him to be a negative defender.

DRPM: -2.16
DBPM: -1.8

DRPM* is a combination of the adjusted on offs,box score stats, and heuristic modeling rules including age, and height factors) that have fit the data best off historical data.

DBPM is a box score metric that has its weights derived from fitting what matched APM best. In essence, its the box score prediction for what someone's APM should be without the random noise the APM has, albeit also without capturing everything missed in the box score that is captured by APM.

Also, he looked bad per
5) The eye test.

I'm just not sure how anyone questions that he actually was bad, while playing bad defense and being forced to run more miles in teh NBA season as a 19-20 year old than anyone but Lillard.

That quote was about that diff% stat, not any of those. Based on that quote, it's very ambiguous as to what it is really measuring. Is it just isolations? Does it adjust for help defense? Is it actually tracking who is guarding who even through switches or does it do an 82games-style lineup matchup and just assuming the "Shooting guard" is guarding the other "shooting guard"? I can't seem to find a more detailed qualitative explanation. If there is I would love to read it, and see if I agree with the methodology.

But as long as you bring it up, there actually is surprisingly little information about some of them. Sure, you can find a quick and easy write up about what those stats purport to represent. But when you try to find the actual formula for RPM? It's not publicly available, at least as of the last time I did a deep google search for it. Neither is any sort of confidence interval, error rate, relevant sample size, etc. Same issue with RAPM.

There's also the issue that, unlike PER, defense is inherently not based on box score stats. There's a lot of subjectivity, which makes having the actual data set even more important. We're basically being asked to put more and more trust in these people, who make money off of page clicks, not being right.

Does that not bother you? Basing your analysis on something where you have very little idea what the formula behind it is?

Sure, all of those stats seem to point in one direction. But what if they all have a common bias? Like maybe they don't properly balance out his awful teammates or the tough role he had to take properly. Net plus/minus and D rating absolutely don't. Maybe some of the more advanced ones do, but, again, without actually knowing the formula we can't know for sure. PER has its biases and we know it, so we know how to adjust. We don't with RPM.

It's not enough to say, "Oh they did a regression to adjust for that." There are tons of really crappy regressions done every day.

I would have to imagine that there are some stats that are very helpful for teams, but unless these guys want to release their formulae and data sets, as a public consumer of this stuff I have to take it with a pretty large grain of salt.

The Diff figure grabbed off nba.com is based upon the play by play data, which is literally every shot taken with him the defender (whether its after a switch, on an isolation, in the post etc. You can get a breakdown of just isolation plays, just post plays, spot up shooting plays etc. Its basically the exact opposite of teh 82games -- who is opposite you -- and is pretty amazing.

To capture the data, they installed a series of 8 (iirc, I'm not looking it up) cameras in each stadium, and then dissect every play. Again, going off memory, but the service was sold to individual NBA teams at a sizable install cost, and pretty soon, there were 8 or so teams using it. Then 15 or so. And then the league stepped in and said that it was clearly a needed project, and they would buy it in bulk for the entire league -- and make a subset of its data available to the public. Honestly, its one of the coolest things I can think of the league doing in a while. There are some things to be careful with in it, but in general its a pretty amazing resource and has proven to be so.

In terms of adjusted on-offs, and straight RAPM, it is hard to get errors as you basically need to bootstrap them. But everything in it is just the actual play data, and there is no hidden variables of any kind, and if you are so inclined, you could spend an hour to learn how to run one and run it yourself even. It isn't a black box stat at all, but a straightforward output based upon play data and who scores when.

In terms of RPM, well... I'm not a fan for a few reasons, one of them being precisely because there are a bunch of 'hidden' variables. If Wiggins is taller than the average sf, that bumps his score. If Wiggins is younger than a prime player, that lowers his score. Its about getting the best priors in the regressions, so what you are asking the regressions to improve upon is not a blank slate, but rather a 'reasonable starting point'. In contrast, a lot of RAPM uses prior year RAPM's to get a better starting point than assuming Dwight Howard and Kevin martin are equally good at defense. But back on RPM, the precursor to it which is very close is publicly available, so if you are really wondering what goes into it, its not hard to get damn close and this is the only one that you can say 'I don't know what it is so I don't trust it'.

In terms of a common bias factor, I would submit it is that he was bad at d.

Basically the stats pull from 2 places -- how his team does when he is on the floor, and how the guy he actually is guarding does. Sometimes you get players who one sets of those favors while the other doesn't -- think that center that is never a help defender at all, but actually decently defends just his man, in part because he never helps the team. With WIggins, the stats say that both his team is worse with him, and his guy he is guarding lights him up. If it were a small sample that might be one thing, but it is not.

[Sixerscan]

There is tons of info out there about these stats and where they come from etc.

In terms of Wiggins defense:

1. The on-offs have him as a negative (for the 30th ranked defense), as does his Drating:

PPP Twolves allow with Wiggins off the court: 1.102
PPP Twolves allow with Wiggins on the court: 1.132 (3 points per 100 possessions worse)

(Calculated as straight how does the defense do with and without Wiggins)

Drating: 114 (19th on the Timberwolves who had a 112).
Calculated basically similarly, with a slight different on possessions etc.

2. The adjusted on-offs for opponent and teammate have him a net negative:
DRAPM: -1.26 points worse

(Calculated by regressing to see how every player does as a separate variable).

3. The Synnergy stats show him as a negative defender

"And Synergy Sports judged his capabilities as an individual defender only marginally better, ranking him in the 32nd percentile of defenders according to its video-scouting metrics."

Calculated by looking at video of every play in the NBA, and seeing how often his man scores. These are available as splits by possession type etc.

4. The box score based metrics* DBPM and DRPM all show him to be a negative defender.

DRPM: -2.16
DBPM: -1.8

DRPM* is a combination of the adjusted on offs,box score stats, and heuristic modeling rules including age, and height factors) that have fit the data best off historical data.

DBPM is a box score metric that has its weights derived from fitting what matched APM best. In essence, its the box score prediction for what someone's APM should be without the random noise the APM has, albeit also without capturing everything missed in the box score that is captured by APM.

Also, he looked bad per
5) The eye test.

I'm just not sure how anyone questions that he actually was bad, while playing bad defense and being forced to run more miles in teh NBA season as a 19-20 year old than anyone but Lillard.

That quote was about that diff% stat, not any of those. Based on that quote, it's very ambiguous as to what it is really measuring. Is it just isolations? Does it adjust for help defense? Is it actually tracking who is guarding who even through switches or does it do an 82games-style lineup matchup and just assuming the "Shooting guard" is guarding the other "shooting guard"? I can't seem to find a more detailed qualitative explanation. If there is I would love to read it, and see if I agree with the methodology.

But as long as you bring it up, there actually is surprisingly little information about some of them. Sure, you can find a quick and easy write up about what those stats purport to represent. But when you try to find the actual formula for RPM? It's not publicly available, at least as of the last time I did a deep google search for it. Neither is any sort of confidence interval, error rate, relevant sample size, etc. Same issue with RAPM.

There's also the issue that, unlike PER, defense is inherently not based on box score stats. There's a lot of subjectivity, which makes having the actual data set even more important. We're basically being asked to put more and more trust in these people, who make money off of page clicks, not being right.

Does that not bother you? Basing your analysis on something where you have very little idea what the formula behind it is?

Sure, all of those stats seem to point in one direction. But what if they all have a common bias? Like maybe they don't properly balance out his awful teammates or the tough role he had to take properly. Net plus/minus and D rating absolutely don't. Maybe some of the more advanced ones do, but, again, without actually knowing the formula we can't know for sure. PER has its biases and we know it, so we know how to adjust. We don't with RPM.

It's not enough to say, "Oh they did a regression to adjust for that." There are tons of really crappy regressions done every day.

I would have to imagine that there are some stats that are very helpful for teams, but unless these guys want to release their formulae and data sets, as a public consumer of this stuff I have to take it with a pretty large grain of salt.

The Diff figure grabbed off nba.com is based upon the play by play data, which is literally every shot taken with him the defender (whether its after a switch, on an isolation, in the post etc. You can get a breakdown of just isolation plays, just post plays, spot up shooting plays etc. Its basically the exact opposite of teh 82games -- who is opposite you -- and is pretty amazing.

To capture the data, they installed a series of 8 (iirc, I'm not looking it up) cameras in each stadium, and then dissect every play. Again, going off memory, but the service was sold to individual NBA teams at a sizable install cost, and pretty soon, there were 8 or so teams using it. Then 15 or so. And then the league stepped in and said that it was clearly a needed project, and they would buy it in bulk for the entire league -- and make a subset of its data available to the public. Honestly, its one of the coolest things I can think of the league doing in a while. There are some things to be careful with in it, but in general its a pretty amazing resource and has proven to be so.

In terms of adjusted on-offs, and straight RAPM, it is hard to get errors as you basically need to bootstrap them. But everything in it is just the actual play data, and there is no hidden variables of any kind, and if you are so inclined, you could spend an hour to learn how to run one and run it yourself even. It isn't a black box stat at all, but a straightforward output based upon play data and who scores when.

In terms of RPM, well... I'm not a fan for a few reasons, one of them being precisely because there are a bunch of 'hidden' variables. If Wiggins is taller than the average sf, that bumps his score. If Wiggins is younger than a prime player, that lowers his score. Its about getting the best priors in the regressions, so what you are asking the regressions to improve upon is not a blank slate, but rather a 'reasonable starting point'. In contrast, a lot of RAPM uses prior year RAPM's to get a better starting point than assuming Dwight Howard and Kevin martin are equally good at defense. But back on RPM, the precursor to it which is very close is publicly available, so if you are really wondering what goes into it, its not hard to get damn close and this is the only one that you can say 'I don't know what it is so I don't trust it'.

In terms of a common bias factor, I would submit it is that he was bad at d.

Basically the stats pull from 2 places -- how his team does when he is on the floor, and how the guy he actually is guarding does. Sometimes you get players who one sets of those favors while the other doesn't -- think that center that is never a help defender at all, but actually decently defends just his man, in part because he never helps the team. With WIggins, the stats say that both his team is worse with him, and his guy he is guarding lights him up. If it were a small sample that might be one thing, but it is not.

I completely spaced on the SportsVU cameras. Right, I should have guessed that is what they used. Still, there is some sort of a black box as to how they translate the data from the cameras to this hard number, no?

I'm going to make it a side project over the next few weeks to get a better sense of what goes into RAPM and RPM. I had been under the impression that it wasn't really possible, but if it's really as tangible as you say it is to get close, it would be nice to get a better grasp of them.

I will however say that I think it's somewhat ridiculous that I even have to do that. They've done a ridiculously good job covering their tracks. For example, there was a paper from the 2010 Sloan conference that explained RAPM, and it's been mysteriously removed from the site. Can't find it anywhere, or an equivalent direct explanation.

Between these and QBR, opaqueness has really become the norm with what I would call "pop stats." I understand why people do it, if you have something valuable don't give it away for free. But, if we can really reverse engineer 98% of it, it's not like it's safe anyway. Regardless, I think it would pretty bad for the game to have these stats being regularly quoted in public discussion when 90% of people (including many of those quoting it) have even a broad sense of how it gets there.

[dbodner]

BPM and VORP, noticeably. Certainly not their shooting and finishing.

The part that I was responding to was when somebody said that MCW was actually more efficient. BPM and VORP aren't efficiency stats.

[HartfordWhalers]

I completely spaced on the SportsVU cameras. Right, I should have guessed that is what they used. Still, there is some sort of a black box as to how they translate the data from the cameras to this hard number, no?

I'm going to make it a side project over the next few weeks to get a better sense of what goes into RAPM and RPM. I had been under the impression that it wasn't really possible, but if it's really as tangible as you say it is to get close, it would be nice to get a better grasp of them.

I will however say that I think it's somewhat ridiculous that I even have to do that. They've done a ridiculously good job covering their tracks. For example, there was a paper from the 2010 Sloan conference that explained RAPM, and it's been mysteriously removed from the site. Can't find it anywhere, or an equivalent direct explanation.

Between these and QBR, opaqueness has really become the norm with what I would call "pop stats." I understand why people do it, if you have something valuable don't give it away for free. But, if we can really reverse engineer 98% of it, it's not like it's safe anyway. Regardless, I think it would pretty bad for the game to have these stats being regularly quoted in public discussion when 90% of people (including many of those quoting it) have even a broad sense of how it gets there.

The play data is literally lumped in based as result, I'm not sure what could be black box about it. You can see how often a guy fouls, how often its an and 1, etc, on individual play types.

There is tons of public stuff on how to do rapm, its more a question of do you want to and have the statistical comfort to do it. This has a few guides:
http://apbr.org/metrics/viewtopic.php?f=2&t=8859

In terms of being mysterious, if you get the idea of regression fitting, its just that -- with a choice of priors and picking how you penalize errors. So I wouldn't call RAPM opaque at all. Complicated perhaps, but on the statistical side not the math side. Just think of it as glorified adjusted plus minus, and you have the gist and know exactly what its faults are.

RPM is a different beast, and I will just leave it at that.

[76ciology]

Any chance Wiggins was playing with a bunch of defensive liabilities with the rationale of him being able to compensate for their defensive shortcomings then having a bunch of defensive guys off the bench?

[HartfordWhalers]

Any chance Wiggins was playing with a bunch of defensive liabilities with the rationale of him being able to compensate for their defensive shortcomings then having a bunch of defensive guys off the bench?

Adjusted on-off and rapm tries to control for that, so long as its not a situation where the lineups rarely if ever overlap. Given Wiggins massive minutes, the lack of a solid defender on their bench, and that he did poorly defending his own man, it seems very unlikely. But in general, something like that is why you wouldn't ever want to take on-off without any adjustments or skepticism and run with them.

There is a big thread on the gb about wiggins and the 538 article. My take there was basically:

I would be shocked if its an issue, and expect Wiggins to be a plus defender (long term). Short term, Wiggins deserved ROY even with his defense. But, if you are saying his defense was still secretly awesome and everyone and stats are wrong, well, then you are just wrong.

[Sixerscan]

The play data is literally lumped in based as result, I'm not sure what could be black box about it. You can see how often a guy fouls, how often its an and 1, etc, on individual play types.

There is tons of public stuff on how to do rapm, its more a question of do you want to and have the statistical comfort to do it. This has a few guides:
http://apbr.org/metrics/viewtopic.php?f=2&t=8859

In terms of being mysterious, if you get the idea of regression fitting, its just that -- with a choice of priors and picking how you penalize errors. So I wouldn't call RAPM opaque at all. Complicated perhaps, but on the statistical side not the math side. Just think of it as glorified adjusted plus minus, and you have the gist and know exactly what its faults are.

RPM is a different beast, and I will just leave it at that.

RPM definitely a bigger concern than RAPM (And, unfortunately, RPM is the one that ESPN seems to be pushing)

But, this is my bigger point, it's weird to me that these sites can throw numbers out without providing explicitly how they got it. I know what regression analysis is. It's been a little while, but after an hour or two I could get myself back up to speed. Regardless, it is somewhat ridiculous that the only way I can see what goes into their numbers is by taking the data and doing it myself. It's not hard, just show how you did it for LeBron or whoever, and let people plug in other players.

I don't understand how that is an effective way to get people interested in the stat. I guess they are exclusively worried about the actual stat community taking it seriously? Seems very insular IMO. It's not like that for baseball. WAR, VORP etc can all be found in a simple google search. Granted baseball is a much simpler game statistically, but the difference in user-friendliness is jarring. It would be like if basketball-reference just started throwing out True Shooting Percentage and just assuming that people are going to figure out that Fts are adjusted by .44

[HartfordWhalers]

The play data is literally lumped in based as result, I'm not sure what could be black box about it. You can see how often a guy fouls, how often its an and 1, etc, on individual play types.

There is tons of public stuff on how to do rapm, its more a question of do you want to and have the statistical comfort to do it. This has a few guides:
http://apbr.org/metrics/viewtopic.php?f=2&t=8859

In terms of being mysterious, if you get the idea of regression fitting, its just that -- with a choice of priors and picking how you penalize errors. So I wouldn't call RAPM opaque at all. Complicated perhaps, but on the statistical side not the math side. Just think of it as glorified adjusted plus minus, and you have the gist and know exactly what its faults are.

RPM is a different beast, and I will just leave it at that.

RPM definitely a bigger concern than RAPM (And, unfortunately, RPM is the one that ESPN seems to be pushing)

But, this is my bigger point, it's weird to me that these sites can throw numbers out without providing explicitly how they got it. I know what regression analysis is. It's been a little while, but after an hour or two I could get myself back up to speed. Regardless, it is somewhat ridiculous that the only way I can see what goes into their numbers is by taking the data and doing it myself. It's not hard, just show how you did it for LeBron or whoever, and let people plug in other players.

I don't understand how that is an effective way to get people interested in the stat. I guess they are exclusively worried about the actual stat community taking it seriously? Seems very insular IMO. It's not like that for baseball. WAR, VORP etc can all be found in a simple google search. Granted baseball is a much simpler game statistically, but the difference in user-friendliness is jarring. It would be like if basketball-reference just started throwing out True Shooting Percentage and just assuming that people are going to figure out that Fts are adjusted by .44

RAPM you can find with a simple google search, and is more intuitive than war, vorp etc.

Espn's love for rpm and its attempt to control it are one issue. What goes into it, is another.

But if you are looking for transparent stats about defense, that are intuitive and make sense, there are a bunch of them as mentioned.