lessthanjake wrote:I think this vote is over and has gone to Kareem (though I’ve not done a count myself), but I will note that the idea of comparing specific Kareem seasons to specific Jordan seasons is a bit odd. That sort of analysis is very very ripe for motivated reasoning (you can basically decide that anything is comparable if you want).
Which is why I noted every assumption being made and explained the reasoning for the extrap, after directly comparing their career-wide 1-year signals(wowy, and indirect) where Kareem also looks better. Yes it "can" be misused, like one could just
assume two-teams had equivalent casts with basically no justification and then use the result to claim the player with better results(with no attempt to account for health or even just the player in question) and then decide the player with better team results was a better cieling raisier. In fact, that is more or less what is happening when you list "dominant title seasons" without any sort of cast-assessment as if we have reason to think 1997 MJ's "impact" would be higher than the consistently better-looking(by box and snippets of non-box) 1988. As is we can also do this with other title-years, it just makes Mj
look worse. 1993-1994 or 1993-1995 does Jordan no favors but I guess we can get to that later. I
could have just taken the 86 net-rating that had the Bulls tick up to 31-wins, but I literally made things as favorable for Micheal as possible regarding a year which dominates the pack statistically.
Ultimately there is going to be some sort of "extrapolation" if you are making a call on which player offered more individual value(or "cieling raising" or whatever). That does not change if you are framing "production" with a certain definition or comparing skillsets. There are potential problems, but that is always going to be true. Even if we had "full-season" RAPM, RAPM does not actually function well as a single-season comparer for players who have value beyond a certain treshold because it curves outlier value down. Nothing we have is without possible issues. So we have to weigh things based on considerations like sample-size, inclusiveness, relevance, ect and try to use things best according to the strengths or weaknesses. That's just how it works.
We can have big samples for a probable peak, or we can have no assumptions. Either way Kareem looks better.
And, on the substance, it inevitably ends up ignoring a whole bunch of Jordan’s title seasons. Like, even if you thought 1971 Kareem is similar in quality to 1991 Jordan (which I think is actually plausible—that 1971 season was ridiculous for Kareem)
1991 is quite literally the Bulls
2nd best title run, connected to the year recently voted by the board as the #1 peak. It is shortly preceded by the year(1988) that is top 3 or top 1 in any box-or-non-box statistic, and it is immediately preceded by the 2nd most popular "peak" pick. What exactly are you expecting to happen if we include years with a worse "on"?
Also, while it did not end up in a championship, as far as
impact goes, if 71 Kareem is ~ 91 Micheal, 72 Kareem is strongly advantaged as he leads a better full-strength team with a diminished version of his co-star relative to 71. That is also an "extrapolation", but I'd say that's much better justified then something like 'best two teammates similar, team must be similar, yes health issues but so what". Kareem has won
6 rings. Was there some rigorous methodology you applied when assuming that Kareem's later wins were a result of better help?
and you thought Jordan’s early seasons were analogous to Kareem’s late 1970’s seasons (which I think is a little less plausible),
Don't think it really matters unless you have a reason to think 1988 is actually not a peak year for MJ in terms of situational lift. Where in the careers it lines-up is kind of secondary. If kareem is exceeding a high-water mark(if you think it is likely not, feel free to explain why) in different years you consider non-anagolous, then from an impact standpoint, that's just really good for Kareem.
All that said, since you think it is unfair of me to focus on Jordan's second best team ever, we can just work off a very clean 82-game sample for
93 Jordan. And here, you do not have to make --any-- assumptions. In fact, as a show of good-faith I will literally replace the Bulls disappointing 93 regular season with their better
full-strength three-peat average:
Yeah, doesn't help Jordan at all. And if you think that's a one-off fluke, I'll remind you they were at a
52-win pace in 95 with Pippen having filed a trade-request. Even in 94, Pippen was forced to spend minutes at
sg, and longley, Scottie, and Horace were beefing. As far as it goes for "data" pre-data ball, the Bulls without MJ being crazy good is about as "certain" as it gets.
Simply put, impact does not much like Micheal here. You can say it's all noise, and point to the PER and whatever production,
subjectively defined, as you want, but as far as "success-cast" goes, Jordan does not have a good case vs Kareem. Even WOWYR really doesn't help here, because with 70sfan's corrections, wowyr's base(wowy) favors
Kareem. So even if you wanted to deal with the noisy-small sample randomness there, we have no idea who wowyr likes better.
If you want me to just go off what I think like u did with miami and chicago, I would also say I don't think the 1980 Lakers had comparable help to all but the 97 or 98 Bulls and they were still very dominant. But I am trying to to go off what can be empirically justified without a bunch of complicated derivations or assumptions(which fwiw i think had me and anenigma estimating the lakers help at somewhere between 45 and 50), and I am
also doing so in a way where I make
favorable assumptions for Mike. If Chicago had won 66 games in 1988, Jordan would probably have a much stronger impact argument. But they won 50(53-full strength). If the Bulls had only won 30 games when Jordan left, he'd have a phenomenal peak argument(but they literally contended for a title). We work with what we have. And what we have does not support Jordan being as impactful as Kareem imo.
I should have seen it coming, but it's easy to get lost in code and results that "feel" right. The general framework of the approach I am working on is still promising, but it needs some tweaks. 
