RealGM Top 100 List #5

[MacGill]

Duncan was more mobile than Wilt or Shaq and if I need to post the video of Wilt walking/slow jogging up the court I will.

Hmm....
(view at 3:06)
[youtube]http://www.youtube.com/watch?v=zukRbA3y0qo#t=186[/youtube]

Also this: http://www.letsrun.com/forum/flat_read. ... ad=5124581

That someone walks or slow-jogs doesn't mean they're incapable of doing otherwise, fwiw.

I can't view the video from where I am at right now but I will respond to this after.

[colts18]

No, actually overrated is someone who (you believe) is rated over what's his actual value, so it inherently depends on the majority opinion on such player.
Unless I've been using this verb wrong my whole life, might very well be...
In such case, I'll rephrase my statement as "Shaq was really rated by most people over his actual value".

Who should have finished higher than Shaq in MVP voting? Shaq finished 2nd in xRAPM (behind Duncan). He was the lead catalyst in improving the Heat by 6 SRS despite the Heat losing Odom/Butler in the trade.

[Ryoga Hibiki]

Ok, so then tell me my good man, how do you determine who you believe is rated over their actual value? Isn't it by how that player actually played? Because if you think he was, and someone else disagrees, how do we determine who was right?

So let me ask you this. In your opinion, not those of others or whatever you deemed the perception. How was Shaq overrated on the basketball court duing his 2005 season?

seriously, where are we going here?
the overrated point came out because of the MVP thing being brought up, but as this concept is just causing useless controversies let's forget about it.
my point is that the Shaq we saw in 2005, even if he improved his conditioning vs his late Lakers version, was not anymore the dominant offensive force he used to be. Of course he was great, but all that MDE thing was over, he became a guy who could be contained on single coverage and could be played with more honest defensive schemes.
Once you take out from Shaq that dominance, Shaq clearly loses the comparison to Duncan who always had a significant defensive edge.
Hence, 2005 is in my view out of scope when it's time to bring out the peak Shaq years when he was better than Duncan.

[Ryoga Hibiki]

Who should have finished higher than Shaq in MVP voting? Shaq finished 2nd in xRAPM (behind Duncan). He was the lead catalyst in improving the Heat by 6 SRS despite the Heat losing Odom/Butler in the trade.

Duncan should have won the MVP without the injury and, using hindsight, Wade was actually the best player in his team.
The voting itself was obviously not a travesty. just I don't think it should be brought out in this topic when comparing the two players.

Anyway, I would have voted for Dirk, at the time, even over Nash.

[MacGill]

Ok, so then tell me my good man, how do you determine who you believe is rated over their actual value? Isn't it by how that player actually played? Because if you think he was, and someone else disagrees, how do we determine who was right?

So let me ask you this. In your opinion, not those of others or whatever you deemed the perception. How was Shaq overrated on the basketball court duing his 2005 season?

seriously, where are we going here?
the overrated point came out because of the MVP thing being brought up, but as this concept is just causing useless controversies let's forget about it.
my point is that the Shaq we saw in 2005, even if he improved his conditioning vs his late Lakers version, was not anymore the dominant offensive force he used to be. Of course he was great, but all that MDE thing was over, he became a guy who could be contained on single coverage and could be played with more honest defensive schemes.
Once you take out from Shaq that dominance, Shaq clearly loses the comparison to Duncan who always had a significant defensive edge.
Hence, 2005 is in my view out of scope when it's time to bring out the peak Shaq years when he was better than Duncan.

Where we are going is I am trying to understand your POV. Why are you comparing him to what he no longer was instead of the player he was at that time? I am asking you why he is overrated and you're telling me because he was no longer dominant like he once was. Of course he wasn't....same as Duncan now as compared to 05. You're tying in the MVP voting which I never even brought up. I simply asked why you believed him overrated and now I know.

05 Shaq was overrated because he was a lesser version of himself than in previous years.

If you go in with the understanding that he no longer was that player..but still a good one, maybe he isn't as overrated as one once thought.

[Warspite]

Why Duncan over Shaq

Well because defense matters. Duncan is one of the GOAT defensive players who has anchored the best defensive teams that didnt come out of Detroit. He has more MVPs more rings and longevity. His impact on the game is far more reaching.

At the end of the day Duncans imprint on the game is much larger than Shaqs. Shaq should have been and could have been ranked higher but because he wasnt dedicated and wasnt coachable he simply started his prime too late and ended too early.

[lorak]

I. Does everyone understand the difference between npi rapm, pi rapm, and xrapm? Does anyone have specific questions about the metric(s)? I feel like we're throwing "rapm" out there too much without considering its context.

No, I do not.
If you plan on using them for evaluations, it would be helfpul for me (and probably others), if youd could start with a short paragraph or two (no more) explaining each one. Adding suggested reading will allow me (and hopefully others) to learn more.

Thanks

Ok, I'll try to keep this as basic and short as possible. Just in case it ends up being too long, I'll bold the especially pertinent sentences.

Adjusted Plus Minus ("APM")

APM is the logical place to start because rapm evolved from it. Raw plus minus is interesting in context, but it's basically useless over time because it doesn't isolate a player from his teammates. To try to correct this problem, Wayne Winston and Jeff Sagarin created APM (they called it WINVAL).

The goal is to isolate a player from his teammates by using regression analysis. To do this you take every 5v5 matchup stint that occurs during an nba season (there are usually ~30,000 different ones) and run a linear regression where the dependent variable is the point margin during each stint (home pts/100poss - away pts/100poss) and the indepedent variables are indications of whether each particular player was on the floor (1 for home players -1 for away). When you run the regression, you get coefficients that represent the ('isolated' plus minus) value of each player.

But there are problems with APM: low minute players often get absurd values (often you just take low mp guys out of APM regressions), the range of values is quite large, and there is very little consistency from year to year.

Non Prior Informed RAPM ("NPI RAPM")

In order to correct some of APM's biggest issues, Joe Sill developed RAPM, which is exactly the same as APM except that it uses a ridge regression instead of an ordinary least squares regression.

Simply put NPI RAPM is RAPM for one year only - single season RAPM

If anyone wants Sill's original paper, PM me and I'll send it along. To put it basically, ridge regression assumes all the coefficients are close to 0, and the more information we have about a coefficient (e.g., a player who played many minutes), the further it can stray from 0. The biggest advantage RAPM has over APM is that it prevents overfitting (low mp players no longer get absurd values, and the range of values is much more reasonable). It also helps multicollinearity to some degree - that is, it helps to disentangle players that play many minutes together. According to Sill's study, one year rapm is as accurate (similar r-squared) as three year apm.

The main disadvantage is that it introduces a bias - i.e., that each value is close to 0. It is also not particularly great (but obv much better than APM) at predicting future seasons and also isn't terribly stable.


Prior Informed RAPM ("PI RAPM")

I mentioned that ridge regression assumes the values are close to 0. We call 0 the "prior." You can change the prior and replace it with any value that you think is more reasonable.

PI RAPM uses RAPM from previous years in order to improve stability and predictive value.

Techincally, PI RAPM could use any prior, but in today's jargon we usually use it to refer only to RAPM where the prior is the previous season's RAPM (which used the previous season's RAPM and so on).

Again, it improves stability and predictive value (plus it looks way more reasonable), but it is influenced heavily by previous seasons (even with an aging curve, it has trouble giving up on certain players !Garnett).

xRAPM/RPM

Rather than using rapm from previous seasons as the priors, these forms use a combination of rapm from previous seasons, spm from previous seasons, and spm from the current season as a prior.

In other words, xRAPM uses RAPM and Box Score elements from current and previous seasons to improve predictive value and stability.

Again, this all helps xRAPM become the best out of sample predictor of the RAPM family, but it introduces biases of its own and heavily relies on the box score and previous seasons. And the way the spm (box score) regression is set up really matters - this is the (2014) SPM used in xRAPM, for example.

-----

All of this stuff can add value to these discussions as long as its used in context. Personally, if I say "RAPM" I'm talking about NPI RAPM, but it's probably best at this point to distinguish so everyone knows specifically which iteration you're talking about.

One important factor to keep in mind is that parsing play-by-play data is very difficult to do perfectly. As a result, different people will come up with different RAPM results for the same season (though they shouldn't be wildly different). Additionally, how people account for other factors (HCA, coaching, point margin when the stint begins, etc.) will change the results.

Hopefully this helps clear things up to some degree. If anyone has more questions, I'd be happy to answer them. AcrosstheCourt could also be helpful here, and it's really too bad mystic got banned because he's knowledgeable about this stuff too.

Write more often, please

[MacGill]

Duncan was more mobile than Wilt or Shaq and if I need to post the video of Wilt walking/slow jogging up the court I will.

Hmm....
(view at 3:06; I also like the block at 5:42)
[youtube]http://www.youtube.com/watch?v=zukRbA3y0qo#t=186[/youtube]

Also this: http://www.letsrun.com/forum/flat_read. ... ad=5124581

That someone walks or slow-jogs doesn't mean they're incapable of doing otherwise, fwiw.

Ok, had to time to watch and have seen that highlight reel footage before.

Question, while I certainly give credit to college Wilt and very early on Wilt, he was a far cry from the version of Wilt that everyone is saying has a case of Duncan, would you not agree? Post highlight reel footage of 67 Wilt, he wasn't doing that, or what about LA Wilt etc. The point, like everything else against O'Neal here is longevity right and Duncan has played much more mobile pf/c throughout his entire career.

This would be like a peak of Wilt's athleticism, where if it was Shaq they would say, he only played that way for a short period of time. And this is the crux of the argument for me. Wilt ballooned like Shaq from that version of a player but we only have very limited footage. However, in that footage there are signs that his weight gains presented the same disadvantages/advantages as with Shaq. However, most Wilt supporters think he was still able to play like he did in Kansas.

He was much slower, wasn't horizontally quick etc but had a great vertical. It sounds familiar to me and if Duncan's career with longevity & attitude bests Shaq's, then how can a player with similar weaknesses be superior? Only if those who champion him refuse to believe that physics is physics.

[RayBan-Sematra]

Here is another good statistical comparison for Shaq & Duncan.
These are all playoff stats.

First 9 years
Shaq (94-01) == 28 / 13 / 3.3apg / 2bpg on 57%TS (28.5 PER)
Duncan(98-10) == 24 / 13 / 3.5apg / 3bpg on 56%TS (26.6 PER)

Next 3 years (years 10-12)
Shaq (02-04) == 25 / 13 / 3apg / 3bpg on 57%TS (27.3 PER)
Duncan (08-10) == 20 / 12 / 3apg / 2bpg on 51%TS (22.3 PER)

Final 2 years (years 13-14)
Shaq (05-06) == 19 / 9 / 2apg / 2bpg on 57%TS (29% USG)
Duncan (12-13) == 18 / 10 / 2apg / 2bpg on 53%TS (25.8 USG)

________________________________________________________


First 9 years -
Both are extremely good. Both are extremely consistent.
Shaq was more productive. Shaq was more consistent (but not by much).
Shaq had the higher peaks and had the more impressive Finals performances.
Shaq was better offensively.
Duncan was better defensively.

Next 3 years
This is where Shaq really separates himself from Duncan.
He clearly performs at a much higher level in years 10 through 12 remaining close to his Peak level while Duncan declines big time.
Shaq remained a Top 2-3 player from 02-04 while Duncan from 08-10 was certainly not ranked that high.

Final 2 years
Not a huge gap between them now.
Shaq was a legit MVP candidate in 05 though and was clearly a greater offensive force in 05-06 then Duncan was from 12-13.
Shaq also was in a bigger role offensively while Duncan was in more of a supporting role on more balanced teams.

Duncan also has one extra year (2014) (year 15) and I think he was better then 07/09 Shaq (which are the years I would use for 15th year Shaq) but not by a huge margin.

Overall it seems that Shaq pretty clearly comes out ahead in this comparison.

Even if you think Duncan was on-par with Shaq as a playoff performer through their first 9 years Shaq gives himself a huge edge with his superior play in years 10-12.

Further beyond that in their next 2-3 years neither one has a significant edge in career value as I have shown.

[An Unbiased Fan]

Overall it seems that Shaq pretty clearly comes out ahead in this comparison.

Even if you think Duncan was on-par with Shaq as a playoff performer through their first 9 years Shaq gives himself a huge edge with his superior play in years 10-12.

Further beyond that in their next 2-3 years neither one has a significant edge in career value as I have shown.

Only if you don't factor in defense. Most of Shaq's teams were below average defensively in the playoffs.

[colts18]

Good game to Bad game ratio for star players in their prime:
Shaq 95-03 16.00
Jordan 91-98 4.43
Hakeem 91-97 3.20
James 06-11 2.14
Barkley 91-96 1.46
Dirk 01-11 1.20
Duncan 99-07 1.20
Wade 05-11 1.17
Kobe 01-10 0.88
Robinson 91-98 0.88
Malone 91-01 0.76
Garnet 00-08 0.48

Good game is >25 Game score and bad game is <15 Game score.

Good shooting games to bad shooting games ratio:
Hakeem 91-97 2.19
James 06-11 1.86
Shaq 95-03 1.76
Dirk 01-11 1.67
Duncan 99-07 1.49
Wade 05-11 1.40
Jordan 91-98 1.32
Barkley 91-96 1.29
Robinson 91-98 0.76
Garnett 00-08 0.62
Malone 91-01 0.61
Kobe 01-10 0.60

Good shooting game is >60 TS% and bad shooting game is <50 TS%.

*Updated through 2011

[colts18]

Peak Shaq vs Peak Duncan vs Peak Garnett:

20 points/10 rebounds games:
Shaq 00-03: 60/70 (85.7%) games
Duncan 01-05: 48/79 (60.8%)
KG 00-04: 23/35 (65.7%)

25/10:
Shaq: 46/70 (65.7%) games
Duncan: 30/79 (38%)
KG: 13/35 (37.1%)

30/10:
Shaq: 31/70 (44.3%) games
Duncan: 17/74 (23.0%)
KG: 7/35 (20%)

30/15:
Shaq: 15/70 (21.4%) games
Duncan: 12/79 (15.2%)
KG: 6/35 (17.1%)

[RayBan-Sematra]

Most of Shaq's teams were below average defensively in the playoffs.

That isn't true.
The worst playoff defense Shaq ever anchored from 93-06 was ranked 13th which is still an above average defense.
He also anchored 7 playoff defenses that were inside the Top 10.

Once Shaq had a decent coach he consistently anchored Top 10 defenses.

Shaq was of course not Duncan on defense but he was still an elite defensive anchor throughout the 90's and early 00's.
The gap between the two defensively is only about one tier and is far smaller then the gap in their offensive ability.

[Exodus]

Count me in as Duncan being better than shaq, that's all

[colts18]

Only if you don't factor in defense. Most of Shaq's and Kobe teams were below average defensively in the playoffs.

Fixed.

Did you know that Kobe has just 1 playoff run post-Shaq where his team had a D rating better than the playoff average? That's 1 good defensive run in 7 tries. Why wasn't Kobe anchoring good defenses? Why did Peak Kobe (2007) anchor the worst playoff defense?

[ardee]

Good game to Bad game ratio for star players in their prime:
Shaq 95-03 16.00
Jordan 91-98 4.43
Hakeem 91-97 3.20
James 06-11 2.14
Barkley 91-96 1.46
Dirk 01-11 1.20
Duncan 99-07 1.20
Wade 05-11 1.17
Kobe 01-10 0.88
Robinson 91-98 0.88
Malone 91-01 0.76
Garnet 00-08 0.48

Good game is >25 Game score and bad game is <15 Game score.

Good shooting games to bad shooting games ratio:
Hakeem 91-97 2.19
James 06-11 1.86
Shaq 95-03 1.76
Dirk 01-11 1.67
Duncan 99-07 1.49
Wade 05-11 1.40
Jordan 91-98 1.32
Barkley 91-96 1.29
Robinson 91-98 0.76
Garnett 00-08 0.62
Malone 91-01 0.61
Kobe 01-10 0.60

Good shooting game is >60 TS% and bad shooting game is <50 TS%.

*Updated through 2011

All well and good to use Game Score until you factor in defense... games where Billups and Stockton were running 7-8 pick and rolls in a row and absolutely abusing the Lakers rotations in an almost comedic fashion do not show up here.

Sent from my GT-I9300 using RealGM Forums mobile app

[Jaivl]

Why wasn't Kobe anchoring good defenses? Why did Peak Kobe (2007) anchor the worst playoff defense?

Are you serious?

And Kobe is not being talking about yet, and no need to.

[drza]

Drza, how do you feel about 2004 KG (peak) against 2004 Shaq? According to RAPM, Shaq has a slight advantage. xRAPM says the same thing. Do you think there was much seperation between the two in the games they played?

2) RAPM itself

xRAPM has Shaq and Garnett essentially equal (10.4 for Shaq and 10.3 for Garnett). As already mentioned, Garnett also had over 2,500 more possessions than Shaq which should be used as a measuring stick when comparing the two numbers.

I'm confused with your second statement. I'm looking at J.E.'s 2004 RAPM and KG has a gigantic lead over Shaq in RAPM:
https://sites.google.com/site/rapmstats/2004-rapm

This is also supported by Doctor MJ's RAPM Chronology spreadsheet (shout-out to Doctor MJ for putting in the time for such a valuable spreadsheet) and I see Garnett with a fairly sizeable lead in 2004 in normalized RAPM (12.65 to 8.7).

In 2005, Shaq finished ahead of both KG and Duncan in MVP voting. He finished ahead of KG in xRAPM, prior informed, NPI RAPM. Did you know that Shaq finished ahead of KG in defensive RAPM in 2005?

For those of you questioning Shaq's defensive impact, here are some NPI RAPM's comparing Shaq to Garnett:

00: Shaq 1.27 KG 1.15
01: Shaq 0.7 KG 1
02: Shaq 3.4 KG 0.7
03: Shaq 0.7 KG 2.9
04: Shaq 1.8 KG 2.0
05: Shaq 1.4 KG -0.7 (KG was negative on defense this year)
06: Shaq 1.5 KG 2.2

Average: Shaq +1.54, KG +1.32

Shaq was not a defensive liability. Shaq was actually a solid defensive player

Who should have finished higher than Shaq in MVP voting? Shaq finished 2nd in xRAPM (behind Duncan). He was the lead catalyst in improving the Heat by 6 SRS despite the Heat losing Odom/Butler in the trade.

Dude, you have to stop this. And I'm not talking about arguing against KG (I appreciate and welcome that. It's what this project is about). I'm talking about the way you're mis-using stats in your rhetoric. I've called you on this before in this project, but it's getting worse. It's misleading, borderline dishonest, and is really dangerous to the acceptance of statistical analysis in these debates.

All of these posts of yours I quoted are from the last few pages of this thread. Do you notice how in every underlined statement, you're swapping which version of RAPM that you're referencing? That you're not giving any context in the posts where you use these numbers? That you're using them extremely carefully to argue points where the results fit your argument? Here's why this is a problem:

1) RAPM is already confusing enough. Just a few pages ago we had folks admitting they didn't know what it was, and we had someone attempting to help with a (downright excellent) explanation of the approach. But casually, a lot of folks aren't going to realize or intuitively understand the differences between those stats. So you mixing and matching which ones you use to fit your arguments are going to mislead many readers.

2) You're blurring the names. You said that "According to RAPM, Shaq has a slight advantage (over KG)". Which obviously sparked confusion, as ReservoirDogs was forced to reply "I'm confused with your second statement. I'm looking at J.E.'s 2004 RAPM and KG has a gigantic lead over Shaq in RAPM:". Which, of course, you already know. But you tried to use the similarity in terminologies between the different RAPM types to mislead.

3) The way you're using the numbers sucks. I use +/- data to support my stances as much as anyone. But even I don't answer questions about who is better with only, "his RAPM says...". You've got to at least give some context for the numbers, you've got to be consistent with your use of the numbers, and for goodness sakes don't try to shape the argument by warping the names of the numbers.

4) Tell the whole story. If you're going to use any of the RAPMs to support your guy, you've got to also respect and report when it doesn't. If you know that there is a weakness in the stat that you're using, and that said weakness relates directly to the result you plan to report, then be very careful how you use it.

You decided to make a repository of RAPM results and put it in your sig for people to see with your every post. I applaud that, as it is great that you're providing that to folks. But when you do that, it makes you somewhat of an ambassadar of sorts. In a place where the stat is only partially understood, by doing this you've declared yourself an authority. So what you do is going to be noticed. There are a lot of people here that either distrust statistics or don't care to give them any credence. But there is good information to be found in the numbers, if you use them honestly and correctly. The way that you're doing it has to be making every statistician here cringe, because you're inviting all of the unwarranted criticisms that plague wider acceptance. Really, you have to do better than this in your posts.

*Note for the mods. I'm trying to criticize the posts and posting styles, not the poster. I don't know how good a job I did with that. But I wanted to do this publicly just in case this is affecting how others in the project might feel about the use of statistical analysis. I apologize in advance if this steps over the line of backseat modding, but I had to put it out there.

*Note for everyone else. I'm on my high horse in this post, and pre-reading it I fear I'll come off as pompous. And obviously, this opens me up to folks throwing stones at my glass house. I'm not perfect either, but just know this was weighing on me enough that even though it opens me up in ways I don't like, I felt I had to speak on it. I'm sorry if this derails or distracts in any way from the great discussions we're having

[RayBan-Sematra]

Did you know that Kobe has just 1 playoff run post-Shaq where his team had a D rating better than the playoff average? That's 1 good defensive run in 7 tries. Why wasn't Kobe anchoring good defenses? Why did Peak Kobe (2007) anchor the worst playoff defense?

Funny how he doesn't hold that against Kobe who he thinks was a consistently elite defender in most years.
With Shaq though...

[MacGill]

See, you can't deny Shaq's stats. Because that is black and white. He has the playoff performances, accolades, and head to head match-ups to back everything up.

So what are other things to consider:

Well, he's a lot more disliked than Duncan. He's also feuded with another ATG, where fan bases go back and forth as to who was more valuable. And well, Shaq is still on TNT, at times still making a fool of himself. What does this have to do with his ranking? Well to some, even DocMJ, it's warranted, and again to a point I agree. But I certainly cannot prove this fact so it's only my own speculation....but voting for Shaq now, not only says how close he was to Wilt, but how separated he was from Kobe. And that goes against both fan bases. It pretty simple to tell who they are as well.

And I am only saying this again, because consistency isn't being applied here. Because if it was both Hakeem & Duncan would be ahead right now given posters criteria of what's valuable.

Last part about defense. System and players play a part and Shaq was put into many systems over his career. But he did show the ability to anchor teams and the same advanced metrics that show Duncan and KG as very good, show Shaq as decent as well. When we know one player can't make the difference that the old skewl players could on the entire team, understanding that Shaq was capable of this clearly shows that under different earlier circumstances there is no reason to believe this couldn't happen. If it never did, I wouldn't bring it up but this to me is no different than luck of teammates as well. Especially when you understand how much better Shaq was offensively.