OhayoKD wrote:Will keep all this in GB as to not derail the project but uh...
You know what, let's start with some housekeeping
iggymcfrack wrote:
I need to bookmark this for when it’s time to argue for CP3 and Stockton in the PC project. It’s amazing how well Stockton fares in the little bit of advanced analytics we have for the past. Everything I see about him makes me like him more and more. I’m starting to wonder if maybe Stockton > Wilt is actually a reasonable take.
RAPTOR does not have on/off pre-97(and even then it's heavily informed by the prior years). IOW is functionally a box-metric for the Jordan and Stockton and as has been noted during the project repeatedly...
Bill Russell — the only defender definitively better than Hakeem — finished higher than ninth in PER once. His average WS/48 placement was 8th (much smaller league). If we want to use someone more directly comparable in style, 1998-2001 Duncan averaged 8th in WS/48, 6th in PER, and 12th in BPM. Insightful stuff.
Just because you can see a few one-number metrics easily accessible on basketball-reference does not mean they have any real merit in these comparisons. Again, toy with formulas enough and you can get all sorts of results. We just discussed that with Engelmann’s “xRAPM”. IBM set up a formula that awarded the crown to Barkley three years running from 1986-88 (just outstanding Barkley years huh) and then to Dennis Rodman in 1992. Or we could use PIPM, which is imo the best pure box metric because it does a better job of weighing the defensive aspects of box inputs. There we see Hakeem start out his career 8th - 2nd - 3rd - 3rd - 2nd - 3rd - 5th - 6th. That reads a lot more honestly to me, but ultimately we just get back to these all being formulas toying with a few inputs not actually substituting for real impact.
If you are not confident enough to assess a player outside of what you can read off a stat-sheet, you can say so without trying to poison the well against those who actually took the time to develop their assessments.
....the box-score is a fickle mistress.
I will get to the non-box(aka impact) data present for Jordan and Lebron below (
spoiler: it mostly favors Lebron), but let's start with the lede:
TheGOATRises007 wrote:
On the GB board, I agree with you. I disagree regarding the PC board.
Gregoire wrote: There are definitely a group of LeBron superfans that are toxically aggressive, and it gets tiring.
VanWest82 wrote:Over there it's less than 10% with those making pro Jordan arguments getting shouted down and enduring personal attacks
So is said. Yet, the first "personal" attack(not to be confused with your arguments being criticized/challenged) I can find comes from...a
Jordan voter:
Also why even bother to quote me when you know I don't take you seriously as a poster anyway? Seems like you're just looking to muck up the project by starting meaningless arguments.
(note that when dquinn similarly challenged the criteria for voting someone for lebron on page 1, they were
not personally attacked?)
It was then, another
Jordan voter who said it would be a "bad look" if MJ was voted below Russell:
i kinda think it would be a bad look for the project if he drops to 4, making anyone from the outside looking in want to dismiss the whole thing. i suppose we can say we don't care about such things, but i'm not sure it furthers discussion if that happens. but we shall see i guess.
And it was someone who prefers
Jordan who decided to try and derail a thread by bringing up personal beef(after a rather embarrasing apology demand we'll get to later):
I'm not sure you're going to get it with this poster. I post this not to bring back old baggage, but to offer a bit of support -- You're not the only one who's struggled with this poster. Four posters including me left the latest Peaks project after verbal abuse and bad-faith arguments from this guy, and I would hate for people to start leaving this new project for the same reason.
Notably the poster above, has
outright lied about what the poster they "struggled" with claimed in order to potray them as dishonest:
https://forums.realgm.com/boards/viewtopic.php?p=104349742#p104349742Why?
Possibly to deflect from their calculations of playoff on/off swinging towards Lebron with a more accurate process and several posters realizing they had been "cooking"(chucking out inconvenient years, inconsistent process, ect)numbers to make Jordan look as good:
https://forums.realgm.com/boards/viewtopic.php?p=104353387#p104353387While various posters have complained I(and the "mob") were being unfair, none have bothered trying to challenge this account of events despite me offering them multiple opportunities to do so
That lie above seems to have spread with gregoire citing it as he has yet another meltdown spurned by Jordan being evaluated differently:
Gregoire wrote:try to falsificate plus-minus metrics (some of them claimed that manually counted 500 MJ games only to prove that Lebron is better lol
It was of course angry jordan fans who flooded the thread I'm referencing largely contributing to it getting locked, and it's jordan fans again who've jumped into the ongoing project complaining about anti-mj bias.
And then, last but not least...
lessthanjake wrote:Lol, just read through the last few pages of this thread, and was surprised to find that my own posts in the PC board and the responses to them were discussed a bunch in this thread. My two cents is somewhere in the middle to be honest: I do actually think that there’s a lot of hostility on the PC board to anything that is perceived as pro-Jordan or anti-LeBron. I think I can reasonably say that I myself experienced quite a bit of hostility—including a boatload of sarcastic and mocking remarks—while arguing in favor of Jordan (though I won’t claim to be a delicate flower that doesn’t punch back).
Hmmm.
Well you certainly
throw punches:
LessthanJake wrote:It might behoove you to step back for a moment and question whether there might actually be something to learn here from someone with more knowledge than you.
Among other things, I also come at this discussion as an independently knowledgeable person about Jordan’s career and his teams, since I watched virtually all their games, and I’ve tried to impart some of that knowledge on you, which you obviously do not independently have.
It’s all just completely motivated-reasoning-based data manipulation, and, again, you surely realize that.
I don’t really feel the need to go through all this, because there’s a lot of obviously motivated reasoning going on here that I don’t think many people would find persuasive on its face, but just a few minor things
I don’t really feel the need to go through all this, because there’s a lot of obviously motivated reasoning going on here that I don’t think many people would find persuasive on its face, but just a few minor things:
And even on the separate question about what is more impressive in some sense, the point being made was applying this logic even to Tim Duncan’s Spurs, who did not play in a less-expanded league (the opposite actually). The argument that Duncan’s Spurs were as dominant a team as Jordan’s Bulls despite having worse records and lesser SRS, because of some argument about standard deviations is pretty obviously just motivated reasoning.
To be clear(on that last quote)...
A. that was
not what was argued
Nor does this
Yeah, why don't we check that:
Yeah, I don't know about you, but that does not read to me like a blanket comparison between Jordan's Bulls and Duncan's Spurs. Rather one between specific teams from that run.
Probably could have worded that less definitively, (San's playoff-heavy method is hardly the only approach) but I do think these claims count as "reasonably arguable"...
B. turns out standard deviations
do correlate more with championships than srs
Let's look it it the other way, using data from 1984 onward. Here are the mean/median Z scores for champs and non-champs in different SRS bands.
SRS between 2 and 4:
Non-champs: Median 0.64, Mean 0.65 (163 teams)
Champs: Median 0.85, Mean 0.76 (3 teams)
SRS between 4 and 6:
Non-champs: Median 1.06, Mean 1.07 (102 teams)
Champs: Median 1.16, Mean 1.14 (13 Teams)
SRS between 6 and 8:
Non-champs: Median 1.44, Mean 1.48 (69 teams)
Champs: Median 1.60, Mean 1.58 (12 Teams)
SRS above 8:
Non-champs: Median 1.91, Mean 1.92 (11 teams)
Champs: Median 2.09, Mean 2.08 (11 teams)
Within each of these bands, the Z scores are higher for the champs than the non-champs.
What's more, building a model (logistic regression) for a title using either SRS or the Z score, the model with the Z score is more predictive (AIC of 120.07 vs 109.7 using SRS vs Z score for all league history, and AIC of 11.25 vs 8.09 using SRS vs Z score for teams since 1984).
Gap increases using pre-84
I bring this up because
in the very next thread...
lessthanjake wrote: The idea that you’d accuse me of having “sample size only matter[] when you want it to matter” is actually completely insane and wildly dishonest. You need to apologize for this promptly, otherwise it is clear you are discussing in bad faith.
Unbelievable. The fact that you can’t bring yourself to apologize for making such a blatantly dishonest attack is breathtaking. You don’t even address it, but instead go on some free-wheeling subject-changing maneuver to justify your stated refusal to apologize for something that was plainly unacceptable behavior. You were straightforwardly dishonest. Just apologize. It shouldn’t be difficult to do so if you’re a remotely mature human being. And it’s actually substantively important to discussion like this project that you apologize for behavior like this, because it will genuinely discourage people from honestly identifying appropriate caveats to the points they make and the evidence they provide if they know that people will just dishonestly attack them as if they did not do so. And that will ultimately lower the standard of discussion.
And you’re right—there was one particular part of your post I didn’t respond to. And that’s because all you did was take out half the data that was most inconvenient to you and then still only got to a conclusion that you yourself admitted wasn’t even particularly flattering for Hakeem. It was not something I felt the need to respond to, because I didn’t feel it made even a facially persuasive point.
Yeah... I think
"delicate" might be an
understatement. Also, despite Jake's demands for an apology, I think Anenigma is on the money:
anenigma wrote:Cool, let me know when you feel like apologising for
actually doing everything that you just projected onto me. No interest in staying on subject,
showing all data rather than manipulated snippets, properly responding to points made, maintaining honest portrayals, advancing sincere engagement, responding with any real maturity… So much for high standards of discussion.
Let's start with the bold bit as, per usual, there are picks and cherries:
Jordan is a lot lower in raw WOWY (32nd in prime WOWY), so that’s one data point that doesn’t support his case—basically the only one.
Yeah, no.
Summary, extended samples, aka, "metrics"
There is very little of what "you" consider impact data but of course you are rather inconsistent with what is "good-faith" running with WOWYR while complaining about much more straightforward "adjustments". As it so happens, "WOWYR" is the one and only type of metric Jordan actually looks "good" in relative to Lebron. Using your standard of what counts as an impact metric we have:
-> Playoff on/off
(Lebron looks better)
-> On/off
(Lebron looks better, 97/98 rank below 17 and 18 lebron years respectively)
-> On+ON/off
(Lebron looks better, 2nd and 5th best regular season teams rank 8th and 11th respectively)
-> WOWY
(Lebron looks better, and Jordan ranks 4th amongst his contemporaries, literally does not matter what you use)
-> Indirect samples(what eminece outlined in the #2), Lebron looks much better
-> AUPM
(Lebron looks better with the exception of 3-year consecutive where MJ is a bit behind Duncan)
-> Squared RAPM
(Lebron looks better in the same set)
-> Full RAPM
(Lebron looks signficantly better with the potential for Jordan to close in if his early years score better(there is reason to think they might not))
Larger Samples, multi-year extraps(something WOWYR and its variant also use but in a much noisier manner with tiny samples of "off"), one-year indirect
Why don't we start with the large-samples which are not "limited"? We have 82 games telling us the Bulls were a 28-win team in 1984. They were a 30-win team without Jordan in 1986. That team then got better with additions such as oakley. And then, pre-triangle(a scheme that turned the Bulls from a good offense and a below average defense into a historically great offense and a -3 defense over the course of a season), the Bulls peaked at +2 at a 52 or 53-win pace(ben takes the former, eballa extrapolates the latter based on a 28 game-sample after a trade). If we unreasonably give Jordan all the credit for that improvement, out 82-game off sample leaves Jordan at +8.
That is
-> lower lift on a weaker team than what Kareem manages on the 77 Lakers if we ignore they traded significant pieces for Kareem.
-> within range of several kareem marks including when he was a rookie
(note the gap actually becomes quite large when you consider that srs was suppressed post-merger and championship teams of that era ranged from +4 to +6)
-> Dramatically worse than what we see with the same starting rotation(without Lebron) from the 11 cavs who go from 61-wins to 18-wins over a 21-game sample(19-win for the season overall), and worse than what we see over substantial without samples in the second cleveland stint(the most favorable appraisal of the cast gets lebron taking a 30-win team to 60 in the regular season) and small ones for Miami(40 to 60)
-> on the level of what we see with 69 Bill Russell, a retiring player-coach who had just knocked out a gauntlet tougher than any Micheal faced on the Bulls(marginally better replacement fwiw)
-> On par with what we see with 04/05 Duncan through injuries over a substantial sample(he would win a championship in 03 with a weaker version of what he had in 05)
-> On par with multiple substantial samples for Hakeem
-> Worse than substantial samples for Magic
We can also use 94 and 95 where the Bulls(with us inflating 93 by using the bulls full-strength rs-rating) only drop by 5.3-points in the regular season and 2-points in the playoffs making for an overall drop of "5" for the season(this is similar to what we get for 08/09 kg).
We can also look to 86 and 95 and the result remains the same. Over the largest possible samples, Jordan does not look on par with the other major Goat-candidates(Lebron, Russell, Kareem). He also does not gain separation over players he is assumed to be better than including his own contemporaries(magic is consistently advantaged, and Hakeem is competitive in the rs to go along with nigh unrivalled team-level playoff elevation)
If you are looking for the largest sample-sizes, that is what we have. I can make the most favorable assumptions and Jordan still does not reach the same heights as players he is considered comparable with. That is the lede. If you are going to "summarize all the available impact data for Jordan", there's no reason to bury it.
MJ:
'84 to '85 Bulls (w/MJ): +11 wins
'93 (w/MJ) to '94 Bulls: +4 wins
'95 (pre/MJ) to '95 Bulls (w/MJ): +20
LeBron:
'03 to '04 Cavs (w/LeBron): +17 wins
'10 (w/LeBron) to '11 Cavs: +46 wins
'10 to '11 Heat (w/LeBron): +12 wins
'14 (w/LeBron) to '15 Heat: +18 wins
'14 to '15 Cavs (w/LeBron): +26 wins
'18 (w/LeBron) to '19 Cavs: +31 wins
'18 to '19 Lakers (w/LeBron): +7 wins
As has repeatedly established(and ignored by those with positions the evidence is inconvenient for), non-box data
consistently favors Lebron. In fact,
with larger and more inclusive samples, Lebron's best signals
look even better while Jordan's look worse. Even with WOWYR(which excludes much of the data for Lebron listed above and compounds a sample-size issue by tossing around noisy adjustments across the duration of a player's prime(EX: Bulls without 89 Pippen factors largely into how we assess 91 Pippen)), "Jordan being ahead of every goat candidate" is misleading. His advantage over Kareem is questionable as corrected data puts Kareem's WOWY ahead(the base of WOWYR). His advantage over Russell is entirely a matter of framing(do you value championships or srs?) as WOWYR suggests that Russell's 11 championships came with
average help.
There is
no non-box stat which generally puts Jordan #1, even that where Jordan is competing against a narrowed field(almost no on/off pre-97 for any player besides Micheal). Those
unsourced github-sets(something Lessthanjake has refused to acknowledge while calling whatever doesn't favor Jordan "flawed") only really advantages Jordan over the likes of Malone and of course full RAPM for 97/98 places him well below Lebron under the same creators:
(Cheema, the spax)
(keep in mind, it is not clear that second-three-peat MJ's "empirical impact" is lower than first-three peat MJ with the snippets we have arguably favoring the
latter years). Lebron also dominates a field of players who can scale favorably against Mike when we use data that is available for both(Duncan, KG, Steph, Shaq, ect):
James is, arguably, the king of overall plus-minus stats. 2018 is the 25th season of league-wide plus-minus data, which covers nearly 40 percent of the shot-clock era and touches 12 of the top-20 players on this list. None have achieved LeBron’s heights: He holds four of the top-five scaled APM seasons on record, and six of the top eight. Since 2007, 10 of his 11 years land in the 99th percentile.
(scaled-apm, JE)
Among contemporaries, prime Magic(mostly excluded in the data less-than-jake pulled up), is consistently advantaged. So is David-Robinson(pollock, wowy, wowyr, concentrated), with Hakeem mostly grading out as a regular-season peer(and then sees the most team-level elevation in the postseason of anyone from that time-period).
Lessthanjake, bastion of fairness and honesty, repeatedly seeing this sort of non-box data, found various excuses to disregard it(and then attack the credibility of those who provided it) on grounds that were better applied to a metric they preferred(WOWYR):
It’s logical extrapolation using basically completely unrelated data points and trying to draw a meaningful conclusion from them. You might as well pick a completely random team that you happen to think is similarly or less talented than the Bulls without Jordan and look at their wins/SRS and say that the difference is the upper bound of Jordan’s impact.
They are different rosters in that the 88 Bulls were better. I am comfortable giving Jordan an unfair amount of credit, yes. You quite literally just threw out 3-variants of WOWYR, a metric which uses 89 Pippen to assess 91 Pippen as a teammate. This is what "impact" comparisons pre-97 is. You look at the with, you look at the without, and you use your understanding of the game to try and adjust for context and isolate variables. The difference here is I'm using 82-games while WOWYR is using either a handful(overall) or 8 games a season(it is not at all clear what is counted and Ben explicitly makes a distinction between 1994 and "wowy") for one guy and 2.2 games for another(that's bill russell) and THEN decides to make "corrections" based on the 3 games a teammate missed 5 years back or the 70 games another teammate missed 10 years forward. This is why Ben explcitly outlines using "shorter-time periods" as a solution and then uses "shorter-time periods" as the backbone of his impact write-ups.
WOWYR is extrapolated from "WOWY" utilizing "corrections" that are then applied to every year within a stretch. IOW. They take teammate "impact" from different versions of a team, multiple years removed from each other(the thing you said was "useless!", "ridiculous!", and "bad-faith!") to "correct" WOWY. The difference being the sample here is much smaller than with the extrapolations you didn't like for that same reason.
Whether it's deliberate, or simply a byproduct of jake mostly having no clue what they're talking about...
...the end-result is often an nonsensical jumbling of buzz-words, silly assumptions, and implicit or explicit hostility disguised with relatively sophisticated sentence structure.
For example...
lessthanjake wrote:Actually kind of yeah. I’ve explained this a bit above. But if you have a really ball-dominant style, when you go to the bench, your teammates will have not really touched the ball much in the flow of the offense beyond just finishing some plays with a shot, so they’ll be less in rhythm overall.
Wow you "explained" this, huh. Gee, thanks! Just one problem:
letskissbro wrote:Both Luka Doncic and James Harden have actually played under the exaggerated caricature that people have made "LeBron ball" out to be, yet their teams fare significantly better when they're off the court than LeBron's teams and often field positive net ratings with their stars on the bench.
LeBron's average times of possession in that 16-21 stretch from Ben's graph (minus 2019): 5.3, 6.4, 6.7, 7.4, 6.4
Corresponding net ratings for those teams while he sat: -4.3, -8.8, -0.5, -0.9, -2.0
Luka (20-23): 8.9, 8.9, 9.3, 9.1
Off NetRtg: +4.2, +0.3, +3.4, -2.7
Harden (17-23): 9.3, 8.8, 9.3, 8.6, 8.6, 9.2, 8.6
Off NetRtg: +3.7, +5.2, +1.1, -3.4, +2.4, +1.2, +2.8
If LeBron spends approximately six seconds with the ball in his hands, and that leads to over-reliance on him, why is it that Harden and Luka, who hold the ball for 50% longer, dribble the ball significantly more per touch, play under a similar team setup, and have comparable usage rates, see their teams perform well in their absence?
AEnigma wrote:Yet somehow his exit to the bench dwarfs the bench effects we see from “the really heliocentric ball-dominant guys” like Harden, Luka, Wall, Morant, Parker, DWill, Rose… Even Chris Paul and Steve Nash and peak helio Westbrook (which does need to be specified because for whatever mysterious reason the team survived his absences much better outside of that 2016-18 period…) see less of a drop-off than Steph Curry. Very strange, given this very real and very legitimate theory about how “helio” stars routinely ruin their benches.
Fascinating. I wonder if you might be onto something there: not being easily replaced means your team struggles more without you??? Going to need to double-check the numbers on that, because I was under the impression stars controlled their replacements and that having replacements incapable of recreating their effect was in fact a failing on them.
It's
bullsh-t,
That is why you've been "mocked", not because you voted for Jordan(plenty of people did without facing "toxic" pushback), but because you present assumptions you cannot actually back-up as if they are obvious truths, questioned for centering your arguments around these assumptions, and instead of responding like a reasonable person, you lash out and accuse other of being toxic and "drinking their kool-aid":
But you’re really just repackaging preconceived notions into boldly-pronounced arguments that are based on absurdity, like that we can measure Jordan’s impact by comparing the results of two essentially completely different rosters.
This an example of
projection.
It was after all,
you who assumed that every team Russell has ever played on was stacked because a player who by 1969 was a 6th-man was voted in the top 100.
You who assumed the Bulls were no better than the Heat in terms of help because the players at the top of the rosters were similar(while actively trying to dismiss health as a factor). And you who disregarded actual results without Jordan to confidently claim they were "bottom line: not a significant contender" because you had "watched enough to know their talent level" while apparently not picking up on how the triangle worked.
You have literally spent the last four threads spamming
unsourced data while accusing others of statistical malpractice. You then had the audacity to demand an apology when you were rightly called out. If you have any self-awareness, I'd suggest you stop crying wolf. The backlash you've faced is
on you, no one else.
PS: posters in the project have literally told me they were going to obscure their preference for peak as to not "spark controversy", so let me, on their behalf, say the quiet part out loud:
Yank3525 wrote:
Exactly. The longevity argument would hold more weight to me if LeBron was still cranking out years like 2019-20 instead of what he has done the last couple of years. He has basically become a part time player who has accumulated nice counting stats but doesn’t have the impact on winning that he once did.
Lebron is better at basketball, as the data(when it's not comically cherrypicked) suggests he is. Longevity isn't necessary. Lebron was better young, he was better old, and he was better at his best. Jordan's stock in PC has not fallen because kareem played extra-years. It's fallen because people are now more interesting in winning than subjective definitions of "production", and the idea that it was simply
longevity that dragged Jordan down is just a misrepresentation on what's been happening over the last year. It wasn't longevity that has seen multiple people argue for Hakeem, Duncan, Kareem, and Russell on the basis of how good they were. So let's stop reducing this debate to a regurgitation of Zach Lowe. It wasn't just "lebron fans", it was a bunch of posters who like different players coming together and thinking about basketball a bit differently. Why don't we analyze that, instead of hitting the same beats as first things first.
