f4p wrote:DraymondGold wrote:.
Compare this to Harden, whose box stats (including EPM which has a box component) look best, but whose pure impact lags behind. This is consistent with the heliocentric playstyle, which tends to overrate pure impact compared to the box estimate. In fact, we can actually see this in Harden’s impact evolution. He’s at his most impactful per-possession pre-heliocentric ball, then becomes less impactful per-possession as he raises his load to record-breaking amounts. Some of this was discussed in prior threads by DoctorMJ, and I agree with some of his conclusions. I do think some of the lineups were set to reduce Harden’s raw plus minus (most notably Harden and Paul’s minutes were offset more than most star pairings, with Paul leading the bench lineups). But a) these lineups effects should be corrected in adjusted metrics RAPM, where Harden still comes out behind the competition, and b) this speaks to Harden’s scalability issues (and the scalability issues of heliocentrism in general)
What exactly are those issues? Harden became super heliocentric and the 2018 rockets are one of the best teams ever. Seems like a kind of obvious non-issue. In fact, seems like the increased usage made increased success. With nash we brainstorm reasons why certain numbers might look low but with harden we're just like "the team was amazing, hardens impact must have decreased as apparently the rest of the team not only offset that decrease but went so far up they created an all time team in spite of him.". Imagine if hardens teammates had been so impactful when he was good. The 2018 rockets won like 90% of their games when healthy (44-5). What does 2014 or 2016 harden do with that cast? 95%, 100%? Do the warriors just forfeit and say "hey, can we play this game in 4 years"?
I feel like the Nets being fairly easily the best offense in the league (and something crazy when the big 3 played), and harden having a top 5 offense for 5 franchises and 6 coaches, should reduce scalability concerns.
Well, what exactly do you mean by 'one of the best teams ever'? Do you have them top 10? I personally think that's beyond uncertainty range. Do you have them top 20? I could maybe see the argument, but it doesn't make me comfortable. Do you have them top 30? Personally, that's where I start to think things are more reasonable, though I could see them lower depending on criteria, and top 40 wouldn’t be crazy to me at all.
They're
-tied 19th-22nd in regular season wins (I think over full NBA history),
~33rd in regular season SRS (full NBA history),
-way down at 94th in overall SRS (RS+PS; although as we've discussed this likely underrates their opponent Warriors since it was calculated on a rolling basis; they were on pace for getting just inside the top 30 from 1955-2020 before they faced the warriors, i.e. on pace to be ranked in the mid/low 30s all time from 1947-2025),
-they're 29th in full-season ELO (from 1947-2023 and likely just outside of 30 including the last two seasons).
Which is genuinely great stuff! Very respectable ceiling raising.
But at the same time, we're talking about two top ~30 peak players of all time (both playing in their prime and one playing during their ~3 year peak), + a respectable 3rd best player who fits well, + solid depth that fits well, + a fantastic / all-time coach, + playing a strategy ahead of their time. Has there ever been a team in NBA history that had even 4/5 of these and not been in the top 30/40 ever?
They may be slightly higher if we filter for health. The oft-cited record or MOV with Chris Paul (+ optionally Harden/Capella) healthy which you mention looks even better! Although it starts getting into smaller and smaller samples, doesn't correct for opponents, doesn’t correct for home court, and doesn't compare apples to apples (the fair thing would be to filter every team's SRS when their top 3 were playing and compare, rather than just compare the healthy rockets to the potentially unhealthy competition). Add in that we're talking about regular-season only with two players who don't exactly have a reputation for playoff improvement (fair or not) and two players who play a lot worse just one year later even with the same big 3... and it takes some of the shine off that stat.
The 2018 Rockets won 65 games. Then in 2019, with the same big 3 of Harden Paul and Capella, with Harden nominally getting a smidge better (and a young Capella being a year older which should hopefully partially balance out the aging of Paul), with the same coach and scheme, just with a little loss of depth (they lose Ariza, admittedly one of their better starts but far from a star, and Luc who's their 7th man)... they only manage 53 wins. They get a blazing hot 50-8 in games Paul played in 2018, but a much cooler (though still good) 39-19 (55 win pace) when Paul played in 2019. The team roster’s a bit worse in 2019, but it still does give credence to the idea that small sample noise/luck boosted their numbers a bit in the healthy-only regular-season-only 2018 team numbers. And if the team drops so much due to a few changes outside of Harden, that too gives credence to the idea that the apparent ceiling-raising was helped by factors outside of Harden, potentially moreso than some of the other players we’re ranking in this tier.
In terms of scalability concerns re: heliocentrism, well, yeah Harden upping his load to game-breaking levels and not really seeing any increase in his pure impact metrics is a concern. It may have been the best way for that team to win, but as D'Antoni has said, he designs his schemes to maximize the impact and winning of his players and teams... so Harden having lower impact numbers than some of the the competition (in this thread or in the ones just voted in ahead of him) in this scenario doesn't give a ton of confidence that Harden's the better player. He's close enough we can argue it and people may differ, but it is a legit concern.
It’s not that a younger Harden would have done better. It’s that as the team got better, as we added more talent around Harden, Harden’s impact dropped — which is exactly what bad scalability means!
Then we look at the chemistry and fit with costar Paul and it does leave a little to be desired. In 2018, they had a net rating of + 12.84 with Paul + Harden on, but a net rating of +12.37 with only Paul on, and a net rating of +8.64 with only Harden on. Really great stuff to reach +12, but a far cry from the +14 +16 or +18 that some pairings with great chemistry + all the things Harden had (depth, coaching, scheme) can reach. Compare the on-rating with them both vs just Paul, and it's a disappointingly small improvement adding Harden to Paul. Then comparing the Paul-only vs Harden-only lineups and Harden's disappointingly lower (against potentially more starting lineups, but still…). Then we look at the film, and they take a your-turn my-turn style of offense that doesn't suggest much chemistry as some costars, and we see that their rotations are separated by an unusual amount for costars, again suggesting a lack of additive chemistry... which gets harder in the playoffs, when the expectation is you'll be playing both your best players for most of the game so it gets harder to get dominance if they aren't additive on the floor together.
Adding in 2018-2019 and the trend continues. +9.14 Harden+Paul on, +10.41 Paul-only, +7.16 Harden-only. So they're actually worse with the two of them playing than Paul-only, and Paul-only lineups continue to look better than Harden-only (again potentially against more bench lineups which might shrink the concerns, but still).
What about chemistry with Cappella? Taking a larger sample over Harden's full 3-year peak now that we have it available, 18-20 Harden + Capella on are +7.25, Harden-only 6.97, Capella-only 1.94. So it's a tiny bit more additive this time, but still a pretty small improvement pairing the stars, and now we're far from the region of diminishing returns. Compared to AD... 20-22 LeBron + AD on are + 6.83, LeBron only 3.93, AD only -2.83 (oof). So AD looks less impactful, but more is a little more scalable (additive relative to his normal value), pretty much exactly like I predicted in my voting post.
Again, Harden's a great player, and those 2018 Rockets were great. I see him as a Tier ~2 best peak in this century, with a high-end argument that could get him voted a few threads earlier. But there are limitations too, and it's a matter of balancing the pros with the cons, and reasonable people might disagree on how that ends up looking for Harden.
