Here's the data we have: RAPM: We have small samples of Hakeem's RAPM, thanks to Squared2020. We have ~136 prime games (14 games in 1988 + 25 games in 1991 + 19 games in 1996 = 58 games in his 10-year prime, plus full-season data in 1997). We also have full post-prime and 9 games from his rookie season. Small samples can be very noisy (so larger uncertainty range), but 25 games in 1991 is big enough to not be entirely noise (particularly when boosted by the context of data from 3 other prime years, and data from 6 non-prime years).
How does Hakeem look in prime RAPM? His values are +1.82 in 1985, +1.52 in 1988, +3.19 in 1991, +3.50 in 1996, +3.37 in 1997. In other words... Bird, Magic, Shaq, Duncan, Garnett, Kobe, Curry >> Hakeem (with no data for Russell, Wilt, Oscar, West). I.e., Hakeem's RAPM data is significantly lower than all the available players in this tier. But we're dealing with small samples, where Hakeem's teams underperformed vs their full-season rate.
What if we curve Hakeem's numbers up, based on his team's full-season play (so if Hakeem's teams performed 12% worse in the games we have vs their full-season rating, what if we assume the underperformance is equally from Hakeem and his teammates and so boost Hakeem's numbers by 12%)? Hakeem ends up having +1.7 in 1988, +3.4 in 1991, +4.8 in 1996. Which is an improvement!... that still isn't enough to get Hakeem over the better years of literally any of the other available players in this tier.
Okay, if we still think the measurement is too noisy, what if we only compare the full-season data at equivalent ages (so age 34+). This gives a handle on how players aged, and maybe can help us infer prime value based on the decline. Hakeem (age 34+): 3.37, 3.11, 2.62, 1.56, 0.5, 1.04 Shaq (age 34+): 1.97, 2.96, 0.62, -1.32, 0.43. Duncan (age 34+): 3.26, 5.1, 5.24, 4.03, 3.04 Garnett (age 34+): 5.73, 6.89, 6.3, 3.46, 1.53 Kobe (age 34+): 0.74, 1.89, 0.18, -0.86 Finally, at least he's not last again!
So in (limited) prime RAPM, Hakeem looks worse than every available player in this tier. As an older player, Hakeem looks better than Shaq and Kobe (but under Duncan and Garnett). But again, this data is not ideal, and it may not capture his playoff improvement (which I'll try to get to in the coming posts/days).
...
*note that RAPM is a per 100 possessions stat. I'll edit in per-season values in a sec:
Don’t small sample RAPM numbers tend to peak much lower in general? Like if the leaders for a regular season RAPM will be at +7, the leaders for postseason RAPM will be more like +3? I don’t think you can compare fragment data from a 14-25 game sample and expect them to be on remotely the same scale as full season data. I feel like the fact that his post-prime full season data tends to be higher than his peak numbers from small season fragments would tend to support this hypothesis.
I don’t actually *know* the answer to this, but intuitively I wouldn’t think that’d be the case given that RAPM is a rate stat. But of course different RAPM measures are scaled differently (as are the same RAPM measure but when it is measuring RS as opposed to playoffs), so comparing values across measures is apples and oranges.
OhayoKD wrote:Lebron contributes more to all the phases of play than Messi does. And he is of course a defensive anchor unlike messi.
f4p wrote: yes, i remember him trying to walk it back later, but too little too late as far as i'm concerned when your initial writeup for one of the best teams ever is "james harden choked! couldn't beat vastly more talented team with best teammate injured. also, rockets not really that good if you think about it." maybe write the "oops, my formula underrated the warriors" part first.
someone had also suggested to him at one point in the project, non-rockets related, that iterating the results could help. basically until they stabilize. so you don't get the 2018 warriors as the 5th best team ever and the team that took them to 7 as a -0.3 SRS team.
if you want to tell me the 2020 celtics are better than the 1995 rockets, then i'm not sure what to say. there has to be some benefit to beating 4 straight 6 SRS teams, even if the margins aren't crazy, over just beating a good raptors team by a lot in the 2nd round. at this point, it's even less the rockets being 100th and why would the 2020 celtics be the 47th best team ever? i think this probably shows that playoff SRS could use an adjustment. it tends to get in a self-reinforcing loop. the 2016 spurs are amazing because they were a 10+ SRS team. the thunder beat them so they must be even more amazing. the warriors, already a 10+ SRS team, beat the thunder so they're even more amazing. the cavs beat the warriors so they're even more amazing. some part of this chain probably involves significant underperformance from one of the great teams as opposed to continual outperformance by the teams ranked lower at the beginning of the playoffs. the 2020 celtics get a +15 SRS for beating the post-kawhi raptors and a +8 for losing to a heat team that was +2.6 when the playoffs started, while the 1995 rockets are languishing at a +7.8 for knocking off 4 straight teams with an average 6 SRS, all on the road.
With all this talk about injury I may as well as note that there are significant opposing injuries in every title-run. The only two exceptions I can think off going back all the way to 89 would be....
Miami Heat's run in 2012....and the Houston Rocket's run in 94. Injury is also an especially odd to fixate on if you're backing Steph when his team was a big beneficiary of opposition health in 2015, 2017, 2018, and 2022. Iow, every time He's won a title
Agreed there's injuries for basically every title run! This isn't meant to suggest otherwise. Like I said in... some post a few pages back?... the last thread?... I'm not someone to claim there's an asterisk to a championship run or anything like that.
I just bring injuries up as I think it's a potential bias for a metric like 'Titles compared Expected Titles', along with variability. If you're just counting rings (or rings vs expected rings), you may overrate teams that had favorable variance and opposing injuries, while underrating teams that had unfavorable variance / opposing players.
That's one of the reasons I tend to prefer SRS or other similar metrics. They're not perfect, but they certainly do a better job at capturing variance (a close series gets less credit than a blowout series, thus giving less credit to a team that may have benefited from variance). SRS doesn't capture injuries inherently, but there's to ways it can handle injuries better: 1) if an opposing player was injured in the regular season or earlier rounds of the playoffs, it will lower the value of the opponents, which is an imperfect/approximate correction for certain longer-timescale injuries. 2) We can manually calculate MoV or SRS only when a team's healthy or facing healthy opponents, and these are easier to calculate than rings over expected rings when both teams are healthy. E.g. we can look at the 2018 Warriors' performance vs 2018 Rockets when everyone was healthy and see that the Warriors were the better team. Compare this to 'Rings vs Expected Rings': if you just look at rings vs expected rings when everyone's healthy, you start dealing with even smaller sample sizes than we were already dealing with, making the measurement far noisier as a method for approximating team goodness in era.
As for Curry, sure, he had injured opponents when he won. His team was also injured every time they lost up until 2022 (16, 19, 20, 21). Regardless, SRS tends to portray Curry's teams as clearly better than Hakeem's, even if we just focus on the series when both teams were healthy. Obviously this is supporting cast driven, but again the impact metrics make it clear that Curry was the more valuable to his team's performance regardless.
I certainly don't weigh team results as my only metric -- there's plenty of stuff that measures a player's value more directly -- but I have a hard time looking at the Rockets and the Warriors' performance and not seeing this as a point for Curry, if we're going to incorporate stuff like team results at all.
The first bolded: Yes, true, but then we have to consistent with listing injuries and what counts as “injuries.” There are obvious cases like KD in 2019, Magic in 1989, etc., but aside from those obvious cases, both sides will probably say they had injuries.
Second bolded: Who was injured in 2016 when they lost?
lessthanjake wrote:Kyrie was extremely impactful without LeBron, and basically had zero impact whatsoever if LeBron was on the court.
lessthanjake wrote: By playing in a way that prevents Kyrie from getting much impact, LeBron ensures that controlling for Kyrie has limited effect…
eminence wrote: He doesn't, I agree, but he does need to be close. And the perception of the time and the OP was that it wasn't - they said "I think most at the time would have been incredulous at the idea that Hakeen was better". Close is good enough to close that 'incredulous' gap.
On a personal opinion - '87/'88 the gap was as big as it ever was.
wasnt hakeem better in 87 and 88 than he was in 86?
i get he didnt win as much but wasnt that coz his team got worse? didnt hakeem go crazy in the po's anyway?
hakeem already whopped magic in 86 with mid. if he had a good team like magic did why wouldnt he just beat magic again?
tbh the more im readin the more hakeem seems like 80's bron.
I don't feel he was significantly better than '86 until he got into Rudy T's system.
Hakeem did play well in the playoffs those seasons.
Feels to me like you've over-indexed into one playoff series that fits your narrative. Magic and Hakeem met twice more, when Hakeem had a decent cast (16-10 without him in '91) and Magics cast was worse than it had been in the mid/late 80s, and the Lakers smacked the Rockets 3-1 and then 3-0.
That presents a question though. Was Hakeem actually worse before he Rudy came? Or was he just not being used properly? Isn't a popular argument for Garnett pointing to 2008 and then letting raising our appraisal of his Timberwolves defense?
Rockets do go 16-10 in 91. But then they go 2-10 in 92, and 9-27 from 92-97. For that-two year stretch are a 36-win team without, and notably, 91 is a bit of an anomaly:
Spoiler:
This is basically the only time in Hakeem's career where his team put together a decent stretch without him. Some of the players just...really stepped up (though not sure how sustainable it was), streaky Maxwell had the best scoring stretch of his career put up a game where he scored 51 and a bunch of other really good scoring games (shot 41% from 3 in the stretch, a career 32% shooter). Thorpe had a great stretch, maybe the best, of his career. Sleepy Floyd the streaky player he was had some great scoring games off the bench (including a 40 pointer). Rockets also had a good defensive backup in Larry Smith. Is this sustainable for a longer stretch? When Hakeem is first out of the lineup, they struggle, but then they put together a hot stretch in February, going 9-2 before Hakeem came back. And when he came back, Rockets were even better, they won 13 straight and went 17-2 over the next 19, so it appears Hakeem built on how well they were playing and took them to another level when he came back. That stretch without him might have actually served as the reason for why they weren't getting Hakeem the ball in the playoffs despite the fact he was really effective when he got it, and gave the streaky perimeter players the green light (Maxwell leading the team with 19 shot attempts a game in the playoffs on sizzling 41FG%/49TS%). Anyways, from '92-'96, his teams were 7-27 without him, in '86 they were 7-7, and in '87/'88 they were 3-7. The in/out change list that’s usually posted around here shows very impressive numbers for Hakeem in basically every other situation (some of the highest in/out changes IIRC), it’s only in ’91 when the difference without him was marginal. I mean the very next season they go 2-10 without him.
While the gap shrank, I do not think Magic was at a disadvantage in terms of help like Hakeem was in 86. Credit where it's due for taking care of buisness(emphatically in 90, not so emphatically in 91), and I do accordingly see the gap as bigger in 89-91(with 86-88 as a swing), but I don't really think that shoots down Hakeem being a peer or near peer early on. And while we're talking about narratives...
Spoiler:
Secondly, I don’t actually think Hakeem is having a huge offensive impact during the Chaney years because of how poorly the offense is organized/structured. There’s really not much synergy between him and his teammates. Watching the Lakers series in ’90, when he’s getting an outrageous amount of defensive attention, Rockets are basically clueless on how to actually take advantage of all the defensive attention he is drawing. Even when the first pass by Hakeem was a proper one, the second and third passes by his teammates weren’t (very lazy, slow, indecisive passing that allows defense to recover despite how compromised it was). There is very poor player movement, the floor spacing is puzzling at times and the Rockets were known for being a very poor half court passing team (the guards not exactly a smart, altruistic bunch). In contrast, Lakers know exactly where the ball is going after Magic or Worthy (who btw undressed Buck Johnson in the series) kicks it out of a double team and the second and third guys make the quick hitting plays to get the right guy the ball. To be fair, Hakeem iirc was also was frustrated by the attention and forced bad shots at times, but it's alarming how incapable Houston was of exploiting such aggressive double/triple teaming.
I do think the idea that Hakeem leaped to parity with the best of the best in 92-95 in his 30's when he was used properly but had platued or diminished after 86(even though his playoff box is improved over the next 2 postseasons) as his situation deteriorated is a bit of over-indexing inofitself.
its my last message in this thread, but I just admit, that all the people, casual and analytical minds, more or less have consencus who has the weight of a rubberized duck. And its not JaivLLLL
Here's the data we have: RAPM: We have small samples of Hakeem's RAPM, thanks to Squared2020. We have ~136 prime games (14 games in 1988 + 25 games in 1991 + 19 games in 1996 = 58 games in his 10-year prime, plus full-season data in 1997). We also have full post-prime and 9 games from his rookie season. Small samples can be very noisy (so larger uncertainty range), but 25 games in 1991 is big enough to not be entirely noise (particularly when boosted by the context of data from 3 other prime years, and data from 6 non-prime years).
How does Hakeem look in prime RAPM? His values are +1.82 in 1985, +1.52 in 1988, +3.19 in 1991, +3.50 in 1996, +3.37 in 1997. In other words... Bird, Magic, Shaq, Duncan, Garnett, Kobe, Curry >> Hakeem (with no data for Russell, Wilt, Oscar, West). I.e., Hakeem's RAPM data is significantly lower than all the available players in this tier. But we're dealing with small samples, where Hakeem's teams underperformed vs their full-season rate.
What if we curve Hakeem's numbers up, based on his team's full-season play (so if Hakeem's teams performed 12% worse in the games we have vs their full-season rating, what if we assume the underperformance is equally from Hakeem and his teammates and so boost Hakeem's numbers by 12%)? Hakeem ends up having +1.7 in 1988, +3.4 in 1991, +4.8 in 1996. Which is an improvement!... that still isn't enough to get Hakeem over the better years of literally any of the other available players in this tier.
Okay, if we still think the measurement is too noisy, what if we only compare the full-season data at equivalent ages (so age 34+). This gives a handle on how players aged, and maybe can help us infer prime value based on the decline. Hakeem (age 34+): 3.37, 3.11, 2.62, 1.56, 0.5, 1.04 Shaq (age 34+): 1.97, 2.96, 0.62, -1.32, 0.43. Duncan (age 34+): 3.26, 5.1, 5.24, 4.03, 3.04 Garnett (age 34+): 5.73, 6.89, 6.3, 3.46, 1.53 Kobe (age 34+): 0.74, 1.89, 0.18, -0.86 Finally, at least he's not last again!
So in (limited) prime RAPM, Hakeem looks worse than every available player in this tier. As an older player, Hakeem looks better than Shaq and Kobe (but under Duncan and Garnett). But again, this data is not ideal, and it may not capture his playoff improvement (which I'll try to get to in the coming posts/days).
...
*note that RAPM is a per 100 possessions stat. I'll edit in per-season values in a sec:
Methodology:
Spoiler:
To get RAPM per Season, you need to convert from per 100 to per season. You can convert like this: 1. Start with your RAPM stat (in per 100 possessions) 2. Dividing by a per 100 stat (e.g. Possessions per 100 possession, Points per 100 Possessions). 3. Multiply by a per season stat (e.g. total possessions in the season, total points in the season, etc.). 4. Optional: Round to the nearest whole number if you want for legibility. I.e., the formula is: RAPM (per 100 possessions) * (stat per Season) / (stat per 100 possessions) = RAPM (per Season)
I will be using Basketball Reference for my Possession data. Then I'll be dividing by Points per 100 possessions and multiplying by total points in the season. The Points cancel out on top and bottom, so all you're doing is multiplying by a conversion factor that converts from per 100 to per season.
Note: different sources approximate possessions in different ways, so you may get slight differences depending on the source. OhayoKD and Enigma once got very mad that I used Basketball Reference for my possession data rather than asking Statmuse (e.g. asking "How many possessions did hakeem play in 1997?" to statmuse, then using (total Possessions)/(100) as my conversion factor).
The problem is... Statmuse possessions look blatantly wrong?? Statmuse says Hakeem played 1,593.1 possessions in the 1996-97 season. First off, possessions don't come in decimals so there's obviously some calculation behind the scenes going on. Second off... this suggests Hakeem played ~ 20.4 possessions per game in a year he was averaging 36.6 minutes per game, when most stars playing a lot of minutes are closer to ~ 75 possessions per game (certainly greater than 50). So the Statmuse possession data seems blatantly wrong.
For reference, Basketball Reference says Hakeem played ~69.9 possessions per game... which sounds far more reasonable! So I'l be using Basketball Reference for possessions.
Total RAPM in a Season (age 34+) Hakeem (age 34+): +184, +96, +87 (shortened season, on pace for +143), +32, +15, +26, Shaq (age 34+): +42, +101, +28, -31, +6, Duncan (age 34+): +135, +161 (shortened season, on pace for +200), +214, +172, +132, +136, Garnett (age 34+): +239, +242 (shortened season, on pace for +301), +243, +73, +28, +37, Kobe (age 34+): +44, +7, +4, -30,
If we sum their RAPM per season (using their 82-game pace RAPM for the shortened seasons), we get: Total RAPM across ages 34–36 season: Garnett (+856), Duncan (+721), Hakeem (+455), Shaq (+140), Kobe (+55) Total RAPM across ages 34–final season: Duncan (+989), Garnett (+921), Hakeem (+496), Shaq (+146), Kobe (+55)
The first bolded: Yes, true, but then we have to consistent with listing injuries and what counts as “injuries.” There are obvious cases like KD in 2019, Magic in 1989, etc., but aside from those obvious cases, both sides will probably say they had injuries.
Second bolded: Who was injured in 2016 when they lost?[/quote] Agreed! There's definitely a judgement call on what threshold injury counts where it gets tricky. For Hakeem, that's partly why I've been trying to emphasize *close series already subject to variability* + injury... basically everyone *should* agree that some of these final-game series for Hakeem were close enough to be subject to variability/luck. So any injuries to opponents just tilt the odds of variability/luck more in Hakeem's favor. This is something you could say about almost every championship, but my point is that Hakeem benefited more variability than other championships, and "Rings over expected Rings" is a measure that seems pretty biased to overrate a player who benefited from more variability/luck/opposing injuries.
I've been using *injured* as a shorthand for missed games (which could be due to other reasons like suspensions, sickness, mid-career retirement before coming back, etc... just having one word is a lot easier than having to spell out every reason).
For 2016: Curry was clearly injured. He missed 6 games & wasn't able to start in 7 games, so that will decrease the 2016 Warriors' Playoff SRS compared to what it would have been when healthy. When he came back, he was still playing through injury. It's blatantly obvious in the stats if you compare Curry's/team stats to any other point in 2016 or any of the neighboring years. You can also see it on film at times, where his speed isn't what it was that year when healthy and where is shooting rhythm wasn't what it was.
In their closest series, Draymond also was also suspended for Game 5 finals. Not saying he shouldn't have been, just that most metrics (rings vs expected rings, playoff SRS, etc.) will think the healthy Warriors underperformed in that game when in reality they were just injured. Starting center Bogut was also injured and out for the final 2.75 games of the finals. Small detail, but Iguodala also tweaked his back in the finals and wasn't able to dunk on a certain fast break at the end of a certain game 7... (though Iguodala was basically always ailing in the playoffs, so nothing new here).
As I've said before, I'm not a fan of championship asterisks. The Cavs won fair and square, and their playoff performance was crazy good. All I'm trying to say is that the context is important. So if the Rockets won 2 championships... but benefited more than most teams from variance... then maybe ring counting (rings vs expected rings) might overrate how good of a playoff performance Hakeem had. If the 16 Warriors played worse in the playoffs and lost in Game 7 by 4 points... maybe this overrates how much of a playoff faller Curry/Warriors are when healthy.
Here's the data we have: RAPM: We have small samples of Hakeem's RAPM, thanks to Squared2020. We have ~136 prime games (14 games in 1988 + 25 games in 1991 + 19 games in 1996 = 58 games in his 10-year prime, plus full-season data in 1997). We also have full post-prime and 9 games from his rookie season. Small samples can be very noisy (so larger uncertainty range), but 25 games in 1991 is big enough to not be entirely noise (particularly when boosted by the context of data from 3 other prime years, and data from 6 non-prime years).
How does Hakeem look in prime RAPM? His values are +1.82 in 1985, +1.52 in 1988, +3.19 in 1991, +3.50 in 1996, +3.37 in 1997. In other words... Bird, Magic, Shaq, Duncan, Garnett, Kobe, Curry >> Hakeem (with no data for Russell, Wilt, Oscar, West). I.e., Hakeem's RAPM data is significantly lower than all the available players in this tier. But we're dealing with small samples, where Hakeem's teams underperformed vs their full-season rate.
What if we curve Hakeem's numbers up, based on his team's full-season play (so if Hakeem's teams performed 12% worse in the games we have vs their full-season rating, what if we assume the underperformance is equally from Hakeem and his teammates and so boost Hakeem's numbers by 12%)? Hakeem ends up having +1.7 in 1988, +3.4 in 1991, +4.8 in 1996. Which is an improvement!... that still isn't enough to get Hakeem over the better years of literally any of the other available players in this tier.
Okay, if we still think the measurement is too noisy, what if we only compare the full-season data at equivalent ages (so age 34+). This gives a handle on how players aged, and maybe can help us infer prime value based on the decline. Hakeem (age 34+): 3.37, 3.11, 2.62, 1.56, 0.5, 1.04 Shaq (age 34+): 1.97, 2.96, 0.62, -1.32, 0.43. Duncan (age 34+): 3.26, 5.1, 5.24, 4.03, 3.04 Garnett (age 34+): 5.73, 6.89, 6.3, 3.46, 1.53 Kobe (age 34+): 0.74, 1.89, 0.18, -0.86 Finally, at least he's not last again!
So in (limited) prime RAPM, Hakeem looks worse than every available player in this tier. As an older player, Hakeem looks better than Shaq and Kobe (but under Duncan and Garnett). But again, this data is not ideal, and it may not capture his playoff improvement (which I'll try to get to in the coming posts/days).
...
*note that RAPM is a per 100 possessions stat. I'll edit in per-season values in a sec:
Don’t small sample RAPM numbers tend to peak much lower in general? Like if the leaders for a regular season RAPM will be at +7, the leaders for postseason RAPM will be more like +3? I don’t think you can compare fragment data from a 14-25 game sample and expect them to be on remotely the same scale as full season data. I feel like the fact that his post-prime full season data tends to be higher than his peak numbers from small season fragments would tend to support this hypothesis.
I don’t actually *know* the answer to this, but intuitively I wouldn’t think that’d be the case given that RAPM is a rate stat. But of course different RAPM measures are scaled differently (as are the same RAPM measure but when it is measuring RS as opposed to playoffs), so comparing values across measures is apples and oranges.
Just looking at the Github database for the last 5 seasons available:
I believe that without sufficient samples of on and off minutes, RAPM tends to credit the players on the floor pretty equally for the time they've played together, making it tough for an individual to stand out. So a lot of Hakeem's score is just going to be "how did the Rockets do when he played". If you look, even when Hakeem gets a 19 or 25 game sample, he tends to score better than the smaller fragments even though '91 was one of his worst WOWY years and '96 was post-prime. I really don't think you can get anywhere comparing 9 and 14 game samples to full season data. That's just poor methodology leading to garbage results.
I don't think it's clear at all that Curry is a more capable offensive anchor in the PS than Kobe. Kobe's 08-10 Lakers squads were consistently better offensively than most of Curry's Warriors teams. I think Kobe was more resilient as a playoff performer actually. RS wise, Curry may very well be the GOAT, but his dropoff in PS play has always been the issue. Not to say he isn't great or isn't worthy of discussion at this level, but he goes from a clear offensive GOAT to clearly not the offensive GOAT, and more in line with guys like Kobe and Dirk (which is still an incredible offensive talent).
The fact that Curry would have the entire D built around stopping him seems relevant. There were 2-3 guys on him at times and he was getting defended/doubled 35 feet out. The guy broke the game in ways Kobe could only dream of.
Warspite wrote:Billups was a horrible scorer who could only score with an open corner 3 or a FT.
Here's the data we have: RAPM: We have small samples of Hakeem's RAPM, thanks to Squared2020. We have ~136 prime games (14 games in 1988 + 25 games in 1991 + 19 games in 1996 = 58 games in his 10-year prime, plus full-season data in 1997). We also have full post-prime and 9 games from his rookie season. Small samples can be very noisy (so larger uncertainty range), but 25 games in 1991 is big enough to not be entirely noise (particularly when boosted by the context of data from 3 other prime years, and data from 6 non-prime years).
How does Hakeem look in prime RAPM? His values are +1.82 in 1985, +1.52 in 1988, +3.19 in 1991, +3.50 in 1996, +3.37 in 1997. In other words... Bird, Magic, Shaq, Duncan, Garnett, Kobe, Curry >> Hakeem (with no data for Russell, Wilt, Oscar, West). I.e., Hakeem's RAPM data is significantly lower than all the available players in this tier. But we're dealing with small samples, where Hakeem's teams underperformed vs their full-season rate.
What if we curve Hakeem's numbers up, based on his team's full-season play (so if Hakeem's teams performed 12% worse in the games we have vs their full-season rating, what if we assume the underperformance is equally from Hakeem and his teammates and so boost Hakeem's numbers by 12%)? Hakeem ends up having +1.7 in 1988, +3.4 in 1991, +4.8 in 1996. Which is an improvement!... that still isn't enough to get Hakeem over the better years of literally any of the other available players in this tier.
Okay, if we still think the measurement is too noisy, what if we only compare the full-season data at equivalent ages (so age 34+). This gives a handle on how players aged, and maybe can help us infer prime value based on the decline. Hakeem (age 34+): 3.37, 3.11, 2.62, 1.56, 0.5, 1.04 Shaq (age 34+): 1.97, 2.96, 0.62, -1.32, 0.43. Duncan (age 34+): 3.26, 5.1, 5.24, 4.03, 3.04 Garnett (age 34+): 5.73, 6.89, 6.3, 3.46, 1.53 Kobe (age 34+): 0.74, 1.89, 0.18, -0.86 Finally, at least he's not last again!
So in (limited) prime RAPM, Hakeem looks worse than every available player in this tier. As an older player, Hakeem looks better than Shaq and Kobe (but under Duncan and Garnett). But again, this data is not ideal, and it may not capture his playoff improvement (which I'll try to get to in the coming posts/days).
...
*note that RAPM is a per 100 possessions stat. I'll edit in per-season values in a sec:
Don’t small sample RAPM numbers tend to peak much lower in general? Like if the leaders for a regular season RAPM will be at +7, the leaders for postseason RAPM will be more like +3? I don’t think you can compare fragment data from a 14-25 game sample and expect them to be on remotely the same scale as full season data. I feel like the fact that his post-prime full season data tends to be higher than his peak numbers from small season fragments would tend to support this hypothesis.
I don’t actually *know* the answer to this, but intuitively I wouldn’t think that’d be the case given that RAPM is a rate stat. But of course different RAPM measures are scaled differently (as are the same RAPM measure but when it is measuring RS as opposed to playoffs), so comparing values across measures is apples and oranges.
I haven't heard the small sample stuff before! My intuition agrees with jake that it's sample independent, but if you can provide a source for the small sample claim, that would obviously make our (limited) RAPM sample for Hakeem look much better.
I also agree with Jake that RAPM scales differently depending on your RAPM metric / prior. Obviously it would be best to have a single metric with the same methodology for all these players, but without that, we do the best with what we have. I intentionally chose a pretty standard/traditional source of RAPM (Goldstein, one of the two original/traditional sources of RAPM) which doesn't use very complicated priors or anything like that with the hope it would be on a similar scale to Squared2020. If you just want to compare the Squared2020 data to itself, Hakeem comes out clearly below available data for Jordan/Magic/Kareem/Bird/Robinson, which would presumably scale him down below the best post-97 RAPM players too (e.g. LeBron, Duncan, KG, Curry, etc.) if we were to believe the numbers... although as always the historical RAPM comes with increased uncertainty from smaller samples.
ShaqAttac wrote:Okay so imma vote 6. HAKEEM didnt realize he looks better than shaq in the rapm, but his "impact" looks worse than kg n magic i guess. thing is he goes crazy in the playoffs and he won 2 chips with weak help and made a final punking magic along the way with no help. Wilt bottled over and over even on stacked supersquads. And shaqs impact just too low and hakeem punked him too. probably was the bitw for his era but mj n magic got lucky his teammates always sucked.
7. KG so he kills all the rapm stuff and apparently got the best longetvity so ig i'll just go with him. cant take him over hakeem coz hakeem had no help and won anyway. Magic prob was better but hiv hurt him.
I'm also going to nom: Mikan
I wanna vote MIKAN for 2 but imma keep my vote in case i need to use it for bron.
This is also p simple. He was waay better than everyone else in a waay no one else was, was the best on o and d, and won 7 rings.
ik we dont got data, but he won the 2nd most and he was way better than every1 else. Seems like a simple 2 to me.
Hope that was good!
thats my vote!
Do you have a source for Hakeems RAPM?
Squared Circle has 58-games from 5 seasons I think(years i don't think we would expect his best data from). Should note that is a very small sample(remember the "off" here is only a portion of a game"), but it doesn't like Hakeem much. Full-RAPM from JE exists for old Hakeem and fwiw he actually looks rather good there(ahead of shaq on more minutes, ahead of kobe on as many, trails behind duncan and kg on less). There's also pollock's plus-minus where Hakeem looks decent in a set that doesn't include what would likely be his two most valuable regular-seasons.
And then there is all the "real-world" stuff where over concetrated and extended samples he looks top-10 worthy(peer for mike, not so much for magic, but he's the biggest riser(team-wide or in terms of individual box) of the trio).
there's also wowyr which basically takes teammate off and then uses it to adjust other player's wowy over a duration of years. It's an attempt at emulating RAPM but it's much much noisier as it works off much smaller samples. There Hakeem looks bad.
Generally though, Hakeem does well in stuff where the samples are bigger and/or the data is more inclusive and he's a similar player to other impact-darlings like Duncan so I'd say it generally bolsters his case
PS: Github stuff is unsourced and should probably not be taken seriously
its my last message in this thread, but I just admit, that all the people, casual and analytical minds, more or less have consencus who has the weight of a rubberized duck. And its not JaivLLLL
f4p wrote:When people started nominating Mikan, my initial thought was "way too early". Mikan is an oddball situation. He's the guy you get to after you get to everybody else. After we clear the decks of the top 10, after we knock out some Kobe's and West's and Steph's and Malone's. When there's no one else left and we just have to acknowledge his dominance.
After all, it's hard to watch those old Mikan tapes and think you are watching modern basketball. If I had to say how most people categorize basketball eras, I think a lot of people think of the "modern era" starting in 1979-80. Magic and Larry entered the NBA, the 3 point line was introduced. The game started looking like the current game. They didn't take 3's, but games on are national TV, big stars are on great teams, the athleticism seems higher, the jump shots seem smoother. It's just a hop, skip, and a jump to the 90's when I think defense took a major leap forward. And beyond Magic and Larry, the extended definition of the "modern era" usually starts with Bill Russell. Russell and the Celtics start dominating just as the shotclock is also introduced. This feels like the first time the NBA is the NBA.
So of course Mikan comes before all of this. Before the shotclock, before even the minimal amount of 60's tape that we have. But then I think of Jerry West and Oscar Robertson and how we have to "get to them first" before we can get to Mikan, and I think why? They feel like they are from a much more modern NBA, but their careers only start 12 years after Mikan's. Are we really to believe the NBA was that different just 12 years before? And of course, Mikan obliterates them in terms of impact in his era.
His first 3 years, they only have WS because they don't track minutes played, and he leads all 3 years. The first 3 years they have PER, he leads the league. Given that all of his per game stats look better the 3 years before, he almost certainly leads PER for 6 straight seasons. And in the playoffs, he leads in PER the first 3 years they have it and WS48 in 2 of the 3 years they have it. In 1949, even if he played all 48 minutes of every game, his 4.2 WS in 10 games would give him the playoff record of 0.420 WS48.
And as some of my Hakeem posts have shown, the only guy who keeps Hakeem from topping some of these "playoff riser" lists is George Mikan. He goes up in the box score more than Hakeem. His team's actual vs expected titles could only be explained without the word "playoff riser" if Mikan was 13.5 wins and 5 SRS better in every regular season. So we have not only the most dominant regular season player of his era, but a player who arguably is the best playoff riser of all time. And he only started 12 years before West and Robertson.
Now of course, Mikan compounds the confusion by not only playing in a weak era but also only giving us 6 seasons to work with. But they would appear to be the most impactful stretch of 6 seasons ever.
Where does this mean Mikan should go? I don't really have an idea. But I'm starting to think a Top 10 placement isn't as crazy as I thought it was a few weeks ago.
I think this is quite understandable, and yes, depending on one's individual approach, I can totally see putting Mikan in the Top 10. Further, with my most recent run through of history and the criteria that felt most reasonable coming out of it, Mikan is higher on my list than before.
The line that particularly strikes me here is the one I've bolded about the 12 year gap. Did that much really change in that time?
So first, let me say that yes, a ton was changing in the 50s. I tend to point to an S-curve growth pattern to point out what should be expected in a situation like this:
If you look at theleague season averages for the NBA over on bkref, you see some indicators that the 50s was a time of steep growth. To me the one that stands out the most is the FG%.
To point the general pictures:
'46-47: 27.9% (first year of BAA, first year of data) '53-54: 37.2% (Mikan's final year as a star) '63-64: 43.3% '73-74: 45.9% '83-84: 49.2% ... '22-23: 47.5%
Noting up front that the mature adoption of the 3 really changes what the stat means, what we can say is that the gap between Mikan's final year as a star and 10 years later in '63-64 is bigger than the gap between '63-64 and the entire rest of NBA history.
And of course, the canyon between '46-47 & '53-54 dwarfs said gap.
This is among the factors that has led me to conclude that the gap between the '60s and the '00s is really pretty small compared to earlier eras.
Next thing I feel is important to point out is the difference drop off in Mikan's TS Add between '50-51 & '51-52. Mikan had an elite TS Add up through '50-51 averaging around +300. There were only 2 other guys racking up numbers like Mikan in that time frame: Alex Groza & Ed Macauley.
From '51-52 on, he drops down to 69.0, 105.2, 67.2, which place him out of the Top 5 in 9-team leagues. Still effective, but considerably less so. What happened? 2 things to point out:
1. The NBA widened the key from 6 to 12 feet specifically to stop Mikan before the '51-52 season.
2. The previous playoffs saw Mikan break his leg, which led to the Lakers not winning the title that year.
I think (1) is the big factor here, but I bring up (2) because I think people should consider for themselves. I absolutely believe that (2) is why the Lakers lost that title...but if it truly crippled Mikan's game for the rest of his career, it's hard to imagine the Lakers would have kept winning titles as a matter of course.
Meanwhile (1) the shooting efficiency data for Mikan looks basically just like I'd imagine the NBA hoped for when they made a move to curtail him.
Now, given that Mikan still led the team to the title one could be forgiven for thinking it's not that relevant in this project, but for anyone trying to judge players here based on "dominance", there's something kinda sticky:
Mikan's dominance through '50-51 is something categorically different from what it is from '51-52 onward. Even ignoring that the NBA in '51-52 was quite primitive compared to what it was in the '60s, how do you judge a man's contemporary dominance when you see a split like that?
I'll leave the questions there for folks to chew on.
Let me hit something quick before I get to my last point:
If Mikan stopped being as dominant as a scorer but the team still won the title, perhaps what that says is that Mikan's defensive impact was through the roof like Russell.
There I'd say: I do think Mikan's defensive impact was central to the Lakers success...but we need to be careful not to look at Mikan as a not-quite-Russell. Mikan's contemporary Bob Kurland was the Russell of that generation, and by all accounts I've seen this got significantly curtailed once goaltending was implemented. Mikan certainly blocked some shots, but it wasn't his "thing" the way it was for Russell, and the way it became the thing for defensive anchors from then on out.
When we talk about league degree-of-difficulty in this context, there's some ambiguity to the term that I feel it's important to separate out into two interpretations:
a) the average quality of basketball in the league
b) top competition faced
While (a) isn't irrelevant here, I think (b) is most relevant to a project like this, and here I'm informed by my experience with tennis GOAT rankings, and you could think of it as the Venus Williams Frustration.
Venus Williams won 7 major titles, and lost in the finals 9 times. Who did she lose to?
Martina Hingis once in 1997. Garbine Muguruza once in 2017. The other 7 times came in between, and came to her sister Serena.
Now, if you're ranking Venus based on Grand Slam success, Serena isn't the competition because she's in a higher tier and would have also kept Venus' rank-rivals from winning their slams. So if you're comparing Venus' 7 to someone else's 7, you're underrating her. Minus Serena, Venus arguably wins 14 slams.
Move over to basketball:
When you bring up Oscar Robertson, you implicitly acknowledge that you don't think Mikan tops Russell or Wilt...which means we should think about how Oscar would stack up if Russell & Wilt didn't exist.
I have Oscar as the #1 player in zero seasons, #2 in 2 seasons, and #3 in 5 seasons. Only 2 top 2 seasons, but clearly a lot more seasons bump up to Top 2 sans Russell & Wilt.
If I do take away Russell & Wilt, Oscar goes up to 8 Top 2 seasons. That's 4 seasons at #1. 3 seasons at #2 behind Jerry West. 1 season at #2 behind Kareem Abdul-Jabbar.
Given that Kareem is clearly in a higher tier, and that West is a legend comparable to West, that's really 8 seasons that can be argued to be comparable as POY candidates as Mikan's 8 such seasons.
And note that this doesn't include a season like Oscar's rookie season where he came in and had arguably that GOAT Offensive season the NBA had ever seen.
So, was Oscar truly less dominant than Mikan, or was that just how it appeared because of the existence of Russell, Wilt, West & Kareem?
Now, as I say that, of course it makes sense to point out that Oscar's teams were necessarily "next in line" to win the title if we take out these other players. I can see the perspective that Mikan made all his teams champions or even top contenders as a matter of course whereas Oscar wasn't able to do anything like that.
Do consider though that when Mikan retired, the following year in '54-55 the Lakers had the #2 DRtg in the league and a 40-32 record. And now consider how that Laker team would have done if they had even rookie Oscar from '60-61. I would certainly favor them to win the title.
Of course that's what people call "time-machining", but as you said, the timescales here are small. If you're skeptical that 12 years can make that much of a difference, is it really time-machining a mere 6 years really beyond the pale?
Once you open the context floodgates with even even a smattering of era adjustment/time-machining, then it is logically inconsistent to not be doing it full hog. We should do our best to apply all the context. To me, that context says Mikan and Cousy would be G-leaguers today, or bench players at best. They will never get my vote on this list.
Warspite wrote:Billups was a horrible scorer who could only score with an open corner 3 or a FT.
Here's the data we have: RAPM: We have small samples of Hakeem's RAPM, thanks to Squared2020. We have ~136 prime games (14 games in 1988 + 25 games in 1991 + 19 games in 1996 = 58 games in his 10-year prime, plus full-season data in 1997). We also have full post-prime and 9 games from his rookie season. Small samples can be very noisy (so larger uncertainty range), but 25 games in 1991 is big enough to not be entirely noise (particularly when boosted by the context of data from 3 other prime years, and data from 6 non-prime years).
How does Hakeem look in prime RAPM? His values are +1.82 in 1985, +1.52 in 1988, +3.19 in 1991, +3.50 in 1996, +3.37 in 1997. In other words... Bird, Magic, Shaq, Duncan, Garnett, Kobe, Curry >> Hakeem (with no data for Russell, Wilt, Oscar, West). I.e., Hakeem's RAPM data is significantly lower than all the available players in this tier. But we're dealing with small samples, where Hakeem's teams underperformed vs their full-season rate.
What if we curve Hakeem's numbers up, based on his team's full-season play (so if Hakeem's teams performed 12% worse in the games we have vs their full-season rating, what if we assume the underperformance is equally from Hakeem and his teammates and so boost Hakeem's numbers by 12%)? Hakeem ends up having +1.7 in 1988, +3.4 in 1991, +4.8 in 1996. Which is an improvement!... that still isn't enough to get Hakeem over the better years of literally any of the other available players in this tier.
Okay, if we still think the measurement is too noisy, what if we only compare the full-season data at equivalent ages (so age 34+). This gives a handle on how players aged, and maybe can help us infer prime value based on the decline. Hakeem (age 34+): 3.37, 3.11, 2.62, 1.56, 0.5, 1.04 Shaq (age 34+): 1.97, 2.96, 0.62, -1.32, 0.43. Duncan (age 34+): 3.26, 5.1, 5.24, 4.03, 3.04 Garnett (age 34+): 5.73, 6.89, 6.3, 3.46, 1.53 Kobe (age 34+): 0.74, 1.89, 0.18, -0.86 Finally, at least he's not last again!
So in (limited) prime RAPM, Hakeem looks worse than every available player in this tier. As an older player, Hakeem looks better than Shaq and Kobe (but under Duncan and Garnett). But again, this data is not ideal, and it may not capture his playoff improvement (which I'll try to get to in the coming posts/days).
...
*note that RAPM is a per 100 possessions stat. I'll edit in per-season values in a sec:
Don’t small sample RAPM numbers tend to peak much lower in general? Like if the leaders for a regular season RAPM will be at +7, the leaders for postseason RAPM will be more like +3? I don’t think you can compare fragment data from a 14-25 game sample and expect them to be on remotely the same scale as full season data. I feel like the fact that his post-prime full season data tends to be higher than his peak numbers from small season fragments would tend to support this hypothesis.
Yeah, account for minutes played and Hakeem's full-rapm looks pretty good considering his age and milage at that point
its my last message in this thread, but I just admit, that all the people, casual and analytical minds, more or less have consencus who has the weight of a rubberized duck. And its not JaivLLLL
There's some proof of concept of players working on better teams/in better environments, but I largely don't upgrade past/future years like that. You reasonably could if you'd like, I imagine it'd be quite beneficial to several of the guys in this thread (Wilt/Hakeem/KG).
18-20 would be 39 win pace, not 36, not sure where that # came from, what were you thinking? You could reasonably combine '91/'92 into that one larger more mediocre sample (mediocre for guys we're discussing for this thread). '95/'96, and you may as well be looking at the career WOWY numbers, team had completely changed for the without samples.
I agree Magic was still advantaged in '90/'91, but that wasn't the position being argued against - "hakeem already whopped magic in 86 with mid. if he had a good team like magic did why wouldnt he just beat magic again?" Magic had a bit better cast and 'whopped' Hakeem back-to-back seasons (you seem to value scoring margin more than me relative to win%). I have serious doubts the '86 upset was getting a repeat (note the Rockets failing to beat anyone other than a young Blazers down 2 PFs until Rudy takes charge). They captured the magic in '86 (heh), but it wasn't particularly representative of Hakeem's time with the Fitch/Chaney Rockets.
iggymcfrack wrote: Don’t small sample RAPM numbers tend to peak much lower in general? Like if the leaders for a regular season RAPM will be at +7, the leaders for postseason RAPM will be more like +3? I don’t think you can compare fragment data from a 14-25 game sample and expect them to be on remotely the same scale as full season data. I feel like the fact that his post-prime full season data tends to be higher than his peak numbers from small season fragments would tend to support this hypothesis.
I don’t actually *know* the answer to this, but intuitively I wouldn’t think that’d be the case given that RAPM is a rate stat. But of course different RAPM measures are scaled differently (as are the same RAPM measure but when it is measuring RS as opposed to playoffs), so comparing values across measures is apples and oranges.
Just looking at the Github database for the last 5 seasons available:
I believe that without sufficient samples of on and off minutes, RAPM tends to credit the players on the floor pretty equally for the time they've played together, making it tough for an individual to stand out. So a lot of Hakeem's score is just going to be "how did the Rockets do when he played". If you look, even when Hakeem gets a 19 or 25 game sample, he tends to score better than the smaller fragments even though '91 was one of his worst WOWY years and '96 was post-prime. I really don't think you can get anywhere comparing 9 and 14 game samples to full season data. That's just poor methodology leading to garbage results.
Yeah I think it’s definitely clear that the RS and PS data on GitHub are scaled differently. You provide a plausible explanation for that, but it may be caused by something else that makes the scaling different. For instance, AuPM/g in the 2023 postseason is pretty clearly on a different scale from the regular season, but the postseason AuPM/g numbers are *higher.* So I kind of think we just don’t really know what is going on with the scaling sometimes. Which is probably a good reason to not really put a lot of stock in comparisons across different measures, years, or game type (i.e. regular season vs. playoffs). The actual RAPM data we have for Hakeem isn’t all that impressive, even if we just compare within the individual measures, though. It’s limited data, but I don’t think it could be portrayed as really supporting a #6 ranking (even if it’s such limited data that we probably can’t say it rules it out either).
OhayoKD wrote:Lebron contributes more to all the phases of play than Messi does. And he is of course a defensive anchor unlike messi.
eminence wrote:(note the Rockets failing to beat anyone other than a young Blazers down 2 PFs until Rudy takes charge).
Their playoff rotation had the same top seven as their regular season minutes leaders. Yeah, Carr was injured: he was gone for half the season and by your own choice of metric was barely missed. Hardly on the level of, oh, spitballing here, losing your starting point guard because of a cocaine addiction and then making the Finals anyway.
Why would someone backing Garnett suddenly feel inspired to slam someone for only winning four series in the first eight years of his career. To the extent 1986 was not representative, it was because — like with the 2004 Wolves — the 1986 roster was an outlier in its overall complementarity.
MyUniBroDavis wrote:Some people are clearly far too overreliant on data without context and look at good all in one or impact numbers and get wowed by that rather than looking at how a roster is actually built around a player
Bklynborn682 wrote: Note: different sources approximate possessions in different ways, so you may get slight differences depending on the source. OhayoKD and Enigma once got very mad that I used Basketball Reference for my possession data rather than asking Statmuse (e.g. asking "How many possessions did hakeem play in 1997?" to statmuse, then using (total Possessions)/(100) as my conversion factor).
Possibility 4: Enigma just lied. They lied about manually tracking LeBron’s 200+ playoff hours by hand. They got the numbers from some website or made them up and lied about manually tracking. If it’s from a website, they haven’t as of yet provided that website so we’re unable to cross-reference that website with PBPStats to see where the disagreement lies.
aenigma wrote:No, because me calculating it by hand is not a source. You do not need statmuse specifically; you can do the same thing more slowly on stats.nba.com.
...
Because when I am trying to replicate a point of data, my first thought is to make sure that the process… replicates the data. A novel thought I am sure to someone who has made it increasingly apparent they just wants to twist data to suit their priors regardless of logical accuracy.
Ben (approximately) gave us Lebron’s 2016-21 on/off per 48 minutes. On that graph, eyeball says it is a bit higher than 18 (you can draw a line clearly placing the dot higher). But when I do the basketball-reference method, I end up with 17.87. And then pbpstats seemingly has lower net rating assessments than basketball-reference. However, when I do the point differential method? +18.18 per 48 minutes.
Is it biased toward Lebron? Well, we can do the same thing with Robinson. His point on the graph looks like 25 (or at least high 24). The bbr method gives 23.9. However, using point differentials gives +25.13. There again, much closer correspondence.
Now, it is possible those graphs are meaningless and poorly placed. Maybe the dots were thrown randomly and the scales are not correct. But if so, that probably raises some questions about all the eyeball scales you used for Jordan…
TLDR: You used per possessions for something countable that was then turned into a per minute stat, and when more accurate data was provided(for different reasons, but let's boil it down to "using the same process" for lebron and jordan), you assumed aenigma was claiming to have film-tracked it and made assumptions about their process, and threw in the possiblity they were lying before trying to clarify.
its my last message in this thread, but I just admit, that all the people, casual and analytical minds, more or less have consencus who has the weight of a rubberized duck. And its not JaivLLLL
OhayoKD wrote:Full-RAPM from JE exists for old Hakeem and fwiw he actually looks rather good there(ahead of shaq on more minutes, ahead of kobe on as many, trails behind duncan and kg on less).
In JE’s RAPM, Hakeem ranks 28th in the NBA in the 1996-1997 season. Hakeem ranks 42nd in the NBA in the 1997-1998 season. In the 1998-1999 season, Hakeem ranks 52nd. In the 1999-2000 season, Hakeem ranks 68th. And he’s so far down in the two years after that that I am too lazy to even count how far back he is. In those same years, Shaq finished: 17th, 1st, 2nd, 1st, 2nd, and 2nd. Old Hakeem definitely does not finish “ahead of Shaq.” He’s not even close. Nor does Hakeem look very good at all in the JE RAPM in general. It’s mostly years where he’s old (though 1996-1997 is a tail-end prime season and he’s still ranked 28th), so it’s not very meaningful, but I just want to clarify for people reading that he was not ahead of Shaq (in part since there’s seemingly been some confusion from other posters here on this precise issue).
Unless you’re talking about comparing old Hakeem’s RAPM to old Shaq’s RAPM? Which I assume is what you must’ve meant here, such that the above is more a clarification for others. But that’s just a cross-year comparison that is just comparing stuff with different scaling, so comparing the numbers isn’t very meaningful. The bottom line is that Hakeem doesn’t really look very good in JE’s RAPM, with the 1996-1997 season only ranking 28th being the most meaningful data point IMO (though even that isn’t all that meaningful). The best we could say is that he looks pretty good given his old age, but even that’s not exactly true with regards to the 1996-1997 season (the one that is probably the most relevant season here in terms of being one we actually value much), where he was only 34.
OhayoKD wrote:Lebron contributes more to all the phases of play than Messi does. And he is of course a defensive anchor unlike messi.
eminence wrote:(note the Rockets failing to beat anyone other than a young Blazers down 2 PFs until Rudy takes charge).
Their playoff rotation had the same top seven as their regular season minutes leaders. Yeah, Carr was injured: he was gone for half the season and by your own choice of metric was barely missed. Hardly on the level of, oh, spitballing here, losing your starting point guard because of a cocaine addiction and then making the Finals anyway.
Why would someone backing Garnett suddenly feel inspired to slam someone for only winning four series in the first eight years of his career. To the extent 1986 was not representative, it was because — like with the 2004 Wolves — the 1986 roster was an outlier in its overall complementarity.
So as a scalp - not particularly comparable to the '86 Lakers ehh?
Because Hakeem didn't only have bad casts as best I can tell. Hakeem from '90-'92 had an '04 Wolves level cast and took them nowhere (additionally in '85/'86). Hakeem only had actually poor casts in that '87-'89 run.
I was indifferent to you correcting ShaqAttak until you started trying to undermine the other wins. Were the Blazers some stout team? No, but they did lose, and it was not just because their real SRS was more like 2.5 than 2.7. Nor was there any particular “magic” in beating the Kings and Nuggets. Those were series they should have won based on their full strength SRS, and they did.
Anyway, I do not think the 1990/91 Rockets are a 2004-level cast at all. Where is the Sam Cassell equivalent? 1986 without Lucas seems like the most comparable situation there. And I am not saying that to denigrate Garnett, because what he did with that abysmal 2003 cast is more impressive as far as first round exits go, but if we are asking who could win in the playoffs with less, I would not say anything suggests that edges Garnett.
MyUniBroDavis wrote:Some people are clearly far too overreliant on data without context and look at good all in one or impact numbers and get wowed by that rather than looking at how a roster is actually built around a player
lessthanjake wrote: I don’t actually *know* the answer to this, but intuitively I wouldn’t think that’d be the case given that RAPM is a rate stat. But of course different RAPM measures are scaled differently (as are the same RAPM measure but when it is measuring RS as opposed to playoffs), so comparing values across measures is apples and oranges.
Just looking at the Github database for the last 5 seasons available:
I believe that without sufficient samples of on and off minutes, RAPM tends to credit the players on the floor pretty equally for the time they've played together, making it tough for an individual to stand out. So a lot of Hakeem's score is just going to be "how did the Rockets do when he played". If you look, even when Hakeem gets a 19 or 25 game sample, he tends to score better than the smaller fragments even though '91 was one of his worst WOWY years and '96 was post-prime. I really don't think you can get anywhere comparing 9 and 14 game samples to full season data. That's just poor methodology leading to garbage results.
Yeah I think it’s definitely clear that the RS and PS data on GitHub are scaled differently. You provide a plausible explanation for that, but it may be caused by something else that makes the scaling different. For instance, AuPM/g in the 2023 postseason is pretty clearly on a different scale from the regular season, but the postseason AuPM/g numbers are *higher.* So I kind of think we just don’t really know what is going on with the scaling sometimes. Which is probably a good reason to not really put a lot of stock in comparisons across different measures, years, or game type (i.e. regular season vs. playoffs). The actual RAPM data we have for Hakeem isn’t all that impressive, even if we just compare within the individual measures, though. It’s limited data, but I don’t think it could be portrayed as really supporting a #6 ranking (even if it’s such limited data that we probably can’t say it rules it out either).
Every RAPM I've seen split into RS and PS components has the RS on a much larger scale since there is more off data for the numbers to be confident in who to attribute credit to. Either way, instead of using numbers we don't understand the scale of, it would be much more useful to just look at the raw data which is much easier to understand. According to the raw data Eminence posted from Pollack, Hakeem was at about a +12 from '94-'96 which would be right up there with the very best of Shaq (the absolute best 3-year stretch you can find for Shaq is +13).
I believe that without sufficient samples of on and off minutes, RAPM tends to credit the players on the floor pretty equally for the time they've played together, making it tough for an individual to stand out. So a lot of Hakeem's score is just going to be "how did the Rockets do when he played". If you look, even when Hakeem gets a 19 or 25 game sample, he tends to score better than the smaller fragments even though '91 was one of his worst WOWY years and '96 was post-prime. I really don't think you can get anywhere comparing 9 and 14 game samples to full season data. That's just poor methodology leading to garbage results.
Yeah I think it’s definitely clear that the RS and PS data on GitHub are scaled differently. You provide a plausible explanation for that, but it may be caused by something else that makes the scaling different. For instance, AuPM/g in the 2023 postseason is pretty clearly on a different scale from the regular season, but the postseason AuPM/g numbers are *higher.* So I kind of think we just don’t really know what is going on with the scaling sometimes. Which is probably a good reason to not really put a lot of stock in comparisons across different measures, years, or game type (i.e. regular season vs. playoffs). The actual RAPM data we have for Hakeem isn’t all that impressive, even if we just compare within the individual measures, though. It’s limited data, but I don’t think it could be portrayed as really supporting a #6 ranking (even if it’s such limited data that we probably can’t say it rules it out either).
Every RAPM I've seen split into RS and PS components has the RS on a much larger scale since there is more off data for the numbers to be confident in who to attribute credit to. Either way, instead of using numbers we don't understand the scale of, it would be much more useful to just look at the raw data which is much easier to understand. According to the raw data Eminence posted from Pollack, Hakeem was at about a +12 from '94-'96 which would be right up there with the very best of Shaq (the absolute best 3-year stretch you can find for Shaq is +13).
I don’t think that that’s right up there with Shaq though. That’s about a +12 on-off for Hakeem, with about a +5-6 net rating while on. That +13 time period you refer to for Shaq has a +10.1 net rating while on. Considering how much harder it is to raise a team above a certain point, I’d say that those numbers are substantially better for Shaq.
OhayoKD wrote:Lebron contributes more to all the phases of play than Messi does. And he is of course a defensive anchor unlike messi.
(though Iguodala was basically always ailing in the playoffs, so nothing new here). 
