#25 Highest Peak of All Time (McGrady '03 wins)

[ElGee]

Vote: 2003 Tracy McGrady

I have '98 Malone coming up next. I don't know what to do or how to comparatively judge Pettit, so if anyone has a take on him, I'd love to read it. Seems like he belongs around here.

Why is Howard leapfrogging McHale so easily? Is it an issue of McHale's health in '87 playoffs (just switch to '86 or '88) or do you guys believe Howard was just the better player?

McHale in a "4 around 1" offensive system built around him would be lethal, way more than Howard, because he could stabilize the high variance you get with such an offense with his unstoppable post scoring (and he has waay more moves to evade doubles). He is a better passer than Howard despite his reputation. His advantage on offense is HUGE. Dwight's offense is so much easier to make adjustments for and shut down in the playoffs. Peak McHale is a guy who had no trouble scoring on anyone or any type of defense. You couldn't slow him down with doubles either because he had shots/moves to turn away from them and score without taking a dribble (like the baseline jumphook which was not only super effective but impossible to block because of his long arms). No one had any measurable success at slowing him down in his prime, especially in the playoffs. With Dwight? I've seen teams put a help defender in the middle of the paint (to take away his rolling hook) and turn him into an offensive foul/turnover machine. Teams can and have made it look remarkably easy to neutralize his post offense. There is a big disparity in FT shooting as well. Dwight is a sub 60% FT shooter and very unreliable game to game (McHale was around ~80ish%). I actually feel fouling him a lot is a good strategy, especially in a game he's getting a lot of touches in the post, because it breaks down the rhythm of the Orlando offense, and he can't make you pay for it at the line (in 2011, Howard shot 15+ FTs 23 times...good right? But his team record in those games was only 12-11).

One thing people need to keep in mind: McHale does not need Larry Bird or any one to create his own offense, he was that damn good at scoring on his own. This is something I feel McHale gets penalized for no reason. In his '87 season he was having almost month long stretches of scoring like 30 ppg on close to 70% shooting , he is still the only player in history to put up a 25+ ppg season on 60+% shooting, was behind in MVP voting to only Magic, MJ and Larry (got about same amount of votes as him). He was unstoppable with or without Larry. I have no personal reason to overrate McHale, I wasn't initially even a fan of his, but after watching hundreds of old Celtic games over the years (Bird is my favorite player of all-time), he gained a lot of respect from me.

Howard is the better interior defender, I don't quite view him as a clear cut DPOY type player as most do. I have tremendous respect for his ability to shut down the lane though. McHale was more versatile (could guard 3-5 comfortably...and I'm not saying this, he has legitimately shut down SFs like Dominique, Dantley and others in the playoffs) and was an excellent shot blocker and team defender himself (staple on all defensive first teams). This is a top 5 defensive big in the league too, so don't underestimate what he brings defensively. The biggest difference is in their rebounding, Howard is better but it should be considered that he has played around stretch 4s who don't rebound while McHale played on a frontline of two other 10+ rpg guys. Regardless, McHale never showed himself close to being on Dwight's level as a rebounder, though I never saw rebounding as a weakness for him. But is rebounding and edge in paint defense (keep in mind that McHale too is a great defensive player) enough to make up for how much more McHale brings offensively?

McHale is a good player to bring up. It was indeed in 1988 that the Celtics posted a 117.1 ORtg (+9.4) with Bird and McHale in the lineup for 63 games. I do indeed consider 1988 to be McHale's offensive peak, and your description of him is correct. Here is my impression/question of him at this time though:

(1) By the end of 1987, he wasn't the same defender he was in the mid 80's (foot injury)
(2) In 1988, that physical limitation is evenly slightly more pronounced.

Do you not agree with this? Because while 87 and 88 McHale is one of the GOAT post scorers, and an underrated passer, I no longer consider him the versatile and mobile defender he was in prior seasons. It's possible I'm not remembering this correctly though, although re-watching the games from this year has only confirmed this thought. McHale's post work is simply astounding, and he's still a positive impact defender IMO (he was even as late as 1991 with his length and intelligence), but this is a classic issue of offensive and defensive peak not overlapping.

To be clear, most people think of McHale's 1987 RS of 26 ppg on 66% TS. But he did suffer through the PS a bit with the foot. In 1988, his ORtg went UP and he was part of a GOAT-level offense (117.1 ORtg would beat the all-time mark by 1.5 points!) and then in the PS, he averages 25 ppg on 67% TS! Even when Bird crumbles against Detroit, McHale averages 27 ppg on 63% TS. Keen observers will note that in the famous Wilkins-Bird duel in 88, McHale had 33 points and 13 reb on 84% TS.

I would vote 87 McHale as of now (perhaps due to RS missed games as a tie-breaker), but am open to the 88 season as well.

[mysticbb]

Ridge Regression (RAPM) will reduce the error more than standard linear regression, but part of that comes at the expense of outliers; It's kind of "mashing" the numbers closer to 0. Single-season RAPM thus has some ballparking value -- everyone should be aware of the numbers -- but it's still subject to problems. Note: this is also why I wanted to know the errors (basketballvalue posts it's 2-yr APM errors...Jerry does not with his RAPM), because the Stockton/Malone case being raised here is really telescoping on that one 2001 result (Stockton +5.6, Malone +2.1).

Ridge regression is biased due to the incorparation of that lambda. Thus, it makes not much sense to calculate a standard error, because you can't interpret that in the same way as the SE from OLS. Jerry actually calculated some sort of errors via bootstrap and the ballpark here is: After 1/4th of a season the error is in the range of the error of a 2yr APM, after half a season, the error is already as small as the values we know from the 6yr APM study from Iliardi. But then again, it is not that useful as information, because the important thing is actually the increased predictive power of RAPM. That is made sure by a mathematical theorem. And that was proven!

What we would interpret as outliers is probably related to the other mechanism Jerry is using, something I tried to explain only a little bit in an earlier discussion. Jerry has that ridge regression combined with a machine learn algorithm, which becomes more "confident" about a certain player the more informations are there. That's why the absolute values for players are increasing with a bigger sample size.

[ElGee]

Ridge Regression (RAPM) will reduce the error more than standard linear regression, but part of that comes at the expense of outliers; It's kind of "mashing" the numbers closer to 0. Single-season RAPM thus has some ballparking value -- everyone should be aware of the numbers -- but it's still subject to problems. Note: this is also why I wanted to know the errors (basketballvalue posts it's 2-yr APM errors...Jerry does not with his RAPM), because the Stockton/Malone case being raised here is really telescoping on that one 2001 result (Stockton +5.6, Malone +2.1).

Ridge regression is biased due to the incorparation of that lambda. Thus, it makes not much sense to calculate a standard error, because you can't interpret that in the same way as the SE from OLS. Jerry actually calculated some sort of errors via bootstrap and the ballpark here is: After 1/4th of a season the error is in the range of the error of a 2yr APM, after half a season, the error is already as small as the values we know from the 6yr APM study from Iliardi. But then again, it is not that useful as information, because the important thing is actually the increased predictive power of RAPM. That is made sure by a mathematical theorem. And that was proven!

In this context, how do you then interpret the large shifts in Garnett and LeBron this year after ~60 games to ~80 games?

I'm also not sure what you mean by the "increased predictive power of RAPM" in this context. Can you elaborate on that?

What we would interpret as outliers is probably related to the other mechanism Jerry is using, something I tried to explain only a little bit in an earlier discussion. Jerry has that ridge regression combined with a machine learn algorithm, which becomes more "confident" about a certain player the more informations are there. That's why the absolute values for players are increasing with a bigger sample size.

Does he discuss this methodology somewhere? If I were to do this, my first thought was always machine learning...

[ardee]

Vote: 2003 T-Mac

There doesn't seem to be enough Malone backing, and anyway, I'm not the biggest fan of 1998.

I thought that T-Mac should've been voted in from before anyway so good pick

As for RAPM, just know that you're posting with someone who can tell you about all the mathematical bells and whistles and how it "adjusts" for everything. But it's still flawed.

And as long as it (or any number) is flawed, use with caution.

Yeah, I agree. And I stand by my earlier point:

The process of the 'adjusting' and the 'distribution of credit of the scoring margin' are procedures that were put together by a mathematician using his own ideas. Now, kudos to him for being able to create it, but as such, it's still a stat someone has well, to put it uneloquently, 'made up'.

My argument is not that I'm against all advanced stats, but that we simply cannot use a system developed by one man using his own mathematical ideas to judge players completely. By doing so, we are completely trusting that man's own procedures and ideas, in which some cases contradict a whole load of other evidence we have about a player.

Kobe had a low RAPM in 2012, but does anyone remember how the bigs struggled to score when he was out for a spell?

I can't in good conscience form my opinions on a stat developed by someone using his own mathematical techniques and ideas. It's the same way Hollinger put together his entire gigantic formula for PER and then started throwing out numbers to show how good a player was. I like PER more then most, but it's an extremely flawed stat that doesn't show us the true value of most players (consider Bill Russell never had a PER of more then 20). It's the same thing for RAPM.

Hope I haven't stepped on any toes with this argument, just my thoughts on a point that seems to be dividing a lot of posters right now.

[mysticbb]

In this context, how do you then interpret the large shifts in Garnett and LeBron this year after ~60 games to ~80 games?

As I said, you can't interpret those errors the same way as you would SE interpret for an OLS. In that context most of the "shift" comes from increased informations. That is related to the machine learn algorithm incoporated.

I'm also not sure what you mean by the "increased predictive power of RAPM" in this context. Can you elaborate on that?

In an out of sample prediction for ill-posed problems, the MSE(OLS) > MSE (RIDGE). You get a better result in out of sample tests for RIDGE than for OLS. That is a proven mathematical theorem, and the very basis of the Thikonov regularization.

Does he discuss this methodology somewhere? If I were to do this, my first thought was always machine learning...

http://www.sloansportsconference.com/wp ... iation.pdf

[mysticbb]

Kobe had a low RAPM in 2012, but does anyone remember how the bigs struggled to score when he was out for a spell?

And do you remember how the defense became better in those 7 games?

The point would be: RAPM remembers all those things, while you are just remembering certain things you think are important. You have cognitive biases and only a limited amount of informations in order to come up with a conclusion. The RAPM technique is using all minutes played during a season.

Also, ridge regression is nothing someone "made up", it is a proven technique used by many people in different subjects. I use that when solving some geophysical problems, it is a known technique in time series analysis. My gf is using it in Microbiology, and for many other people whether it is related to clinical reasearch, economy or other fields in which huge amount of data have to be analysed, ridge regression becomes a vital tool for dealing with ill-posed problems. Really, the math behind that is proven, that is nothing someone just "made up", and at that, very much different from the formulas Hollinger is using.

[colts18]

88 McHale wasn't that good defensively. the Bird-Wilkins duel should be really acknowledged as a duel where Wilkins completely destroys McHale.

[MisterWestside]

Yeah, I agree. And I stand by my earlier point:

The process of the 'adjusting' and the 'distribution of credit of the scoring margin' are procedures that were put together by a mathematician using his own ideas. Now, kudos to him for being able to create it, but as such, it's still a stat someone has well, to put it uneloquently, 'made up".

Well, it's not that RAPM was 'made up' out of the blue. It employs some neat math/statistics methods so there's science behind it. This along with the nature of +/- gives it a general edge over something like PER.

However, as with in every industry, using models with good, honest science behind them =/= perfection and error-free.

[ElGee]

In this context, how do you then interpret the large shifts in Garnett and LeBron this year after ~60 games to ~80 games?

As I said, you can't interpret those errors the same way as you would SE interpret for an OLS. In that context most of the "shift" comes from increased informations. That is related to the machine learn algorithm incoporated.

Yes...but you have no interpretation of this ITO of the stability/accuracy of the model?

I'm also not sure what you mean by the "increased predictive power of RAPM" in this context. Can you elaborate on that?

In an out of sample prediction for ill-posed problems, the MSE(OLS) > MSE (RIDGE). You get a better result in out of sample tests for RIDGE than for OLS. That is a proven mathematical theorem, and the very basis of the Thikonov regularization.

OK...not sure where you are going. No one was talking about RAPM compared to OLS APM.

Does he discuss this methodology somewhere? If I were to do this, my first thought was always machine learning...

http://www.sloansportsconference.com/wp ... iation.pdf

Can you verify that this is the technique for the posted stats on his website? Because it includes box score information...(Ironically, I had someone in a college class once do this using only box score data, well before this paper was published.)

[colts18]

Here are the RAPM numbers for the 2001-2003 period:

Top 10

Shaquille O'Neal 7.4 1.8 9.3 29821.0
Tim Duncan 3.7 4.4 8.1 35513.0
Kevin Garnett 4.1 3.5 7.6 36218.0
Dirk Nowitzki 4.6 2.8 7.4 34856.0
John Stockton 3.5 2.9 6.5 26862.0
Tracy McGrady 4.8 0.7 5.6 34101.0
Ben Wallace 1.0 4.3 5.3 31533.0
David Robinson -0.5 5.6 5.1 23440.0
Rasheed Wallace 1.2 3.7 5.0 31371.0
Steve Francis 3.1 1.8 4.9 32299.0

Karl Malone was at 0.5 (3.4 offense, -2.9 defense) in 32000+ possessions during that time span.

http://stats-for-the-nba.appspot.com/PBP/2003np.html

McGrady was 6th in that time span and everyone ahead of him except Stockton was voted already.

[lorak]

88 McHale wasn't that good defensively. the Bird-Wilkins duel should be really acknowledged as a duel where Wilkins completely destroys McHale.

Why you think so?

Wilkins:
G1 10/24 FG, 10.9 GmSc
G2 8/24, 10.0
G3 9/22, 12.9
G4 15/29, 25.2
G5 7/22, 13.6
G6 13/30, 22.4
G7 19/33, 35.5

So one great game, two good and four bad for player of that caliber.

[tsherkin]

DS, can you post the FTA as well?

Sent from my BlackBerry 9700 using Tapatalk

[lorak]

DS, can you post the FTA as well?

4/7, 6/8, 7/8, 10/13, 10/10, 9/11 and 8/9. But look at TS%: .462, .400, .490, .576, .473, .502 and .636. That's really bad and if McHale guarded him full time, he did very good job.

[tsherkin]

Hmmm.

Not super-different from his usual efficiency in the playoffs at that time, before he started to get the 3 a bit. But yeah, he certainly didn't dominate, just posted similar point totals. What was Bird's PPG/TS% in that series?

Nique... Bleh. Was never a huge fan to begin with. Exciting dunker and he peaked well in the early 90s but bleh.

Sent from my BlackBerry 9700 using Tapatalk

[mysticbb]

Yes...but you have no interpretation of this ITO of the stability/accuracy of the model?

The accuracy of the model is determined by the result of the out of sample tests.

OK...not sure where you are going. No one was talking about RAPM compared to OLS APM.

Well, you were asking for SE of RAPM while referring to bbv with the 2yr APM and the SE. I wanted to point out that the lower error J.E. got via bootstrap is not that important, because the advantage in terms of predictions RAPM has over APM is the more important thing. I should have probably made that more clear in the first answer to you.

Can you verify that this is the technique for the posted stats on his website? Because it includes box score information...(Ironically, I had someone in a college class once do this using only box score data, well before this paper was published.)

No, that is not the method used. I just wanted to show you the machine learning part, which is described in that paper. The RAPM values on his website are RAPM values. He now has an idea that a prior based on boxscore data might give a better prediction, but we have to wait to see the results. Jerry will upload all of his computed data in the next couple of weeks and I hope he will add some more informations about the used method too.

[MisterWestside]

88 McHale wasn't that good defensively. the Bird-Wilkins duel should be really acknowledged as a duel where Wilkins completely destroys McHale.

Why you think so?

Wilkins:
G1 10/24 FG, 10.9 GmSc
G2 8/24, 10.0
G3 9/22, 12.9
G4 15/29, 25.2
G5 7/22, 13.6
G6 13/30, 22.4
G7 19/33, 35.5

So one great game, two good and four bad for player of that caliber.

Using usg/ortg (a bit more pure for offense than GmSc since we're just looking at offense instead of all-around game), Wilkins was "blergh" worthy in Games 1, 3, and 5. In Games 4 and 6 though he was great on offense. That Game 4 .576 ts% was at 40+ usg and he kept tov rates down (116 ortg). Game 6 was at a .502 ts% but that was again at 40 usg and tov rate was lower than Game 4 (113 ortg). Solid ast rates in both games as well. Game 2 was below-average. Game 7 was amazing. I'd say on offense Wilkins was solid in the series -- credit McHale for holding his own on defense but Wilkins wasn't MIA either.

[ElGee]

Yes...but you have no interpretation of this ITO of the stability/accuracy of the model?

The accuracy of the model is determined by the result of the out of sample tests.

OK...not sure where you are going. No one was talking about RAPM compared to OLS APM.

Well, you were asking for SE of RAPM while referring to bbv with the 2yr APM and the SE. I wanted to point out that the lower error J.E. got via bootstrap is not that important, because the advantage in terms of predictions RAPM has over APM is the more important thing. I should have probably made that more clear in the first answer to you.

Here's the full dialogue. Note the context of what you originally quoted me on was me explaining RAPM to Ardee and co.

Ridge Regression (RAPM) will reduce the error more than standard linear regression, but part of that comes at the expense of outliers; It's kind of "mashing" the numbers closer to 0. Single-season RAPM thus has some ballparking value -- everyone should be aware of the numbers -- but it's still subject to problems. Note: this is also why I wanted to know the errors (basketballvalue posts it's 2-yr APM errors...Jerry does not with his RAPM), because the Stockton/Malone case being raised here is really telescoping on that one 2001 result (Stockton +5.6, Malone +2.1).

Ridge regression is biased due to the incorparation of that lambda. Thus, it makes not much sense to calculate a standard error, because you can't interpret that in the same way as the SE from OLS. Jerry actually calculated some sort of errors via bootstrap and the ballpark here is: After 1/4th of a season the error is in the range of the error of a 2yr APM, after half a season, the error is already as small as the values we know from the 6yr APM study from Iliardi. But then again, it is not that useful as information, because the important thing is actually the increased predictive power of RAPM. That is made sure by a mathematical theorem. And that was proven!

I'm also not sure what you mean by the "increased predictive power of RAPM" in this context. Can you elaborate on that?

In an out of sample prediction for ill-posed problems, the MSE(OLS) > MSE (RIDGE). You get a better result in out of sample tests for RIDGE than for OLS. That is a proven mathematical theorem, and the very basis of the Thikonov regularization.

OK...not sure where you are going. No one was talking about RAPM compared to OLS APM.

Well, you were asking for SE of RAPM while referring to bbv with the 2yr APM and the SE. I wanted to point out that the lower error J.E. got via bootstrap is not that important, because the advantage in terms of predictions RAPM has over APM is the more important thing. I should have probably made that more clear in the first answer to you.

I was discussing variance in RAPM. It doesn't have to be formal SE, but I want to know the variance. I consider that to be incredibly relevant and valuable in all models -- who wouldn't want to know this in interpreting it? You seem to have nothing to say about it, nor can offer anything on the matter. I was hoping you would given your knowledge of the subject matter, but it doesn't seem to be important to you.

This isn't, nor ever was, a discussion of RAPM versus APM, so when you say the prediction advantage is "the more important thing" or it's "proven" it's totally irrelevant. If OLS APM never existed, I'd STILL want to know how much 10g can impact a player's score...independent of the average prediction of the overall model.

Here's what I see ITO of variance:

After the May 10 update, 14 players of the 373 who didn't play had their scores change by more than one standard deviation. They never played a minute -- the 103 players left and their 3 PS rounds had that large of an impact on them. (Average change in that group was 0.33 std.)

Of the 103 guys who did play, the AVERAGE change in score was 0.4 standard deviations with a max of 1.5, with 8 players (7.7%) changing by more than 1 Std. Five of the 41 players (12.2%) who played less than 90% of their season minutes before the update changed by at least 1 standard deviation. The average change in std among those 42 players was 0.5 with a max of 1.36 (James Jones). Jones absolute change was -2.5 points by playing 18% of his minutes since the update. LeBron changed by 2.5 points too. The largest absolute change was Marshon Brooks, who dropped 3.2 points (1648 RS MP)...without playing a minute in the postseason.

Better, if we filter by high-minute players only (min 1000 MP prior to May 10), we see:

-of the group (233) a std change of 0.4. max of 1.68. 93.1% w.in 1 std
-of those w/100% of MP before update (169), a std change of 0.4. max of 1.68. 93.4% w/in 1 std
-of those w/90%+ of their min before the update (33) a std change of 0.42. max of 1.50. 94% w/in 1 std
-of those w/80%+ of their min before the update (24) a std change of 0.50. max of 1.22. 91.7% w/in 1 std
-of those w70-90% (31) a std change of 0.50. max of 1.27. 90.3% w.in 1 std

Again, all those numbers are the changes in standard deviations, not the absolute point/pos change. A broad interpretation might be: after ~1700 MP for an individual (avg of this sample), almost 95% of players will be within a std of where they will be after increasing the league's minutes by 4% (and individual MP by no more than 30%).

Can you verify that this is the technique for the posted stats on his website? Because it includes box score information...(Ironically, I had someone in a college class once do this using only box score data, well before this paper was published.)

No, that is not the method used. I just wanted to show you the machine learning part, which is described in that paper. The RAPM values on his website are RAPM values. He now has an idea that a prior based on boxscore data might give a better prediction, but we have to wait to see the results. Jerry will upload all of his computed data in the next couple of weeks and I hope he will add some more informations about the used method too.

So in a discussion about the variance of RAPM, you interject a point about another method (APM). And in a discussion about the variance of one method (RAPM), you point to a completely different method (machine learning) and discuss the variance of that? What are you doing here that I'm missing?

[Doctor MJ]

Well we can look at Kobe Bryant, in 2009 he is a +6.1, in 2011 +2.7, in 2012 +1.5.
He regressed a bit, especially in 2012, but not really that much in boxscore stats.
How can we tell that Malone wasn't some +7 player in 1998, when Kobe went from +6.1 to +1.5 in 3 years ?
I don't think he was +7, but we don't have any evidence so why should we start guessing ?
That's my whole point

Never did I say that the data from Year X proved Year Y has to be a certain way, but when the year's are close it'd be silly not to really ponder what results mean that would require unusual career trajectories in order to explain without saying something about a guy's peak.

I'll also note that Kobe's APM falloff in the last two years is by no means something that has nothing to do with his game at his peak. Yes at peak he had impact, but the issues rising up in the past two years have everything to do with why he didn't have the kind of peak impact so many people think he did.

[therealbig3]

Using non-prior informed data for T-Mac's defense from 01-05:

01: +0.3
02: +0.9
03: -1.2
04: -2.0
05: +0.6

As you can see, he's a slight positive in the years he's not totally carrying a poor supporting cast. When he had to do everything offensively, his defense suffered. But as you can see in years like 01, 02, and 05, he was still having strong offensive impact while being a net positive on defense as well.

Now, why use non-prior informed, and not prior-informed? Because T-Mac doesn't just have an outlying peak, he also has a weird career trajectory in terms of his playing level, and it coincides directly with his effort. He went from a balanced player on either side of the ball in 02 to focusing entirely on offense in 03, and his defense suffered. This was exacerbated in 04, when the team he was with was crap, and he stopped caring about anything but purely scoring. Using the prior informed RAPM for 05 when his effort level on defense was night and day between 04 and 05 is unfair, imo. Using prior informed RAPM for 05, T-Mac barely makes a positive contribution at all on defense, at +0.1, even though my eye test tells me he's playing good defense this year, and is certainly doing more than that for Houston. Although still a small positive, +0.6 using non-prior informed makes more intuitive sense to me, and it is pretty decent for a wing...for example, that +0.6 ranks him well ahead of Kobe that year. His +0.9 in 02 ranks him well ahead of Kobe as well btw.

So I do see evidence that T-Mac was able to make a solid contribution defensively when allowed to. Being a slight positive defensively is more Kobe/Wade territory than Pippen/LeBron, I'll admit, but I guess what I was more getting at was that because of his physical gifts, when T-Mac "turned it on", he was closer to LeBron/Pippen level defense than Wade/Kobe when they "turned it on", because of his physical gifts.

[Doctor MJ]

Vote: Malone '98

This is one of those votes that I don't feel very confident about. The truth is I haven't thought in detail about which of Malone's years I'd pick so I'm just siding with everyone else for now. If Malone's still around next time though, I might switch years.

Me switching back to Malone here is basically acknowledging that I'd had him ahead of the rest of the guys before we started, and the stuff that made me hesitate isn't trustworthy enough to justify a switch now. With that said, I still don't see Malone as light years ahead of Howard, and it's possible I could get switched back.

To ramble a bit more, I wanted to put some thoughts down on other candidates. Looking at it by position:

Center: I see Howard as the next guy, and after that Gilmore is the guy I'm likely to favor. I can see Cowens and Reed joining the fray, but I'd really have to be persuaded.

Power Forward: Malone, and then I think both McHale and McAdoo need serious discussion. I think I'm comfortable with Pettit behind those 3, though I won't say that's set in stone.

Small Forward: Barry & King need to seriously start getting in the discussion now. Durant too I suppose. I'm tempted to talk Connie Hawkins, although his competition was so weak, it's tough to feel confident. I'm expecting Baylor to get in the fray as well, but it's hard for me to imagine seriously considering him before others pick him.

Jawdropping fact for me about Pettit next to Baylor:

As the league's efficiency increased, Pettit was one of those who improved with it. This was a guy who had been a volume scorer at 49% TS typically considered the best scorer in the game and yet despite the fact that his stature fell toward the end, he was actually continuing to become an even more effective scorer. New competition was coming in, some of it better than Pettit (hence his reduced prominence), but he was only feeding on this and learning immensely even as his body fell off. In his 2nd to last year he was 4th in PPG and 8th in TS%.

Baylor on the other hand basically showed no capability at all of learning and finding a way to be more efficient even though with a superior scorer next to him, efficiency was of paramount importance. Now, you might say, "Yeah, but early on, in his peak, he was plenty efficient".

Okay, in '61-62, the year where people get romantic on his abilities that year, Baylor's TS was 49%. The "old world" Pettit by that point was shooting 3% more efficient, and was 4th in scoring volume to Baylor's 2nd. He did this with comparable non-scoring skills. Literally, that's Baylor at his absolute best and all he's got on Pettit is scoring volume...playing on a team with a more efficient scoring superstar who he should be passing more to.

Okay, let me end my rant. My apologies for being so negative on Baylor guys, I just really, really think there are a lot of guys to consider before you side with him.

Shooting guard: I feel like I need others to champion guys here, I'm just not that enthusiastic. Penny and Iceman I suppose?

Point guard: Interesting Frazier has been mentioned. He needs to be discussed. KJ also. Maybe even Stockton although that makes me think of Payton and others.