Top 25 peaks of the 2001-25: #15-#16 Spots

[eminence]

2 Year RAPM, per nbarapm
Nash 07/08, +7.5, 2nd (MDA)
Nash 09/10, +4.7, 7th (no MDA)
Harden 14/15, +5.3, 7th (best)
Harden 17/18, +4.2 12th (2 stretches including the season I've commonly seen as his peak - '18)
Harden 18/19, +4.4 12th
Russ 13/14, +4.5, 9th (best) - missed half the season
Russ 16/17, +4.4, 11th (two stretches including '17)
Russ 17/18, +3.1, 25th

Nash had his best seasons with MDA, no doubt. But he still measured up just fine to Harden/Russ in non-MDA years.

Not a 2 year thing either, fine by 3/4 years 09-12 as well. Yes, Nash in LA was garbage.

[lessthanjake]

Doesn’t it seem a little suspicious that to put Nash near the top, you need to scrub any box score priors AND scrub any seasons from outside the D’Antoni system in a multi-year analysis?

So this is the crux of what your post is saying. And I think my prior post addressed the box-score-prior thing. In theory, we might think that a guy who is underrated by one box measure might not be underrated by another one. So it does give a bit more pause that Nash does worse than his RAPM in like any box metric, rather than just one box metric. But, ultimately, all box metrics are drawing from really similar information, so it’s not all that unexpected for them to come to similar conclusions about a player. It may just not be very possible to put together a box metric that correlates well with RAPM in the aggregate while *also* not underrating Nash. Indeed, that seems to be the position that BPM’s creator takes on it, writing: “Yes, Steve Nash was ridiculous on offense, and no, the box score still can’t fully capture that fact.”

As for the “D’Antoni system” thing, I’d say two things.

First, I’m quite comfortable with the idea that Nash was simply a far better and more impactful player in Phoenix than he’d been in Dallas. Part of that is surely that he was more impactful when given more offensive primacy. And part of it is probably the stuff Cavsfansince84 talked about (i.e. that we know Nash made serious conditioning/training/diet changes). On the part about him being more impactful with more offensive primacy, that seems really obvious to me, but I don’t ding him for that, because I’m judging him on what happened (which was that he was really impactful in that context), and I don’t even think that that was a relative benefit for him because it’s not like other great players don’t generally have systems designed to optimize them.

Second, I’m always a bit confused by the invocation of D’Antoni when Nash played on the Suns for years after D’Antoni left, and he was still a very impactful player (despite, by that point, being 34-36 years old). The response I often see to this is people arguing that D’Antoni left but the system remained similar. Which, to begin with, doesn’t exactly seem right since they brought in Shaq and that materially changed things. But also, at the point at which we’re saying that a guy is getting his impact from a general style of play rather than from one particular coach, I think we’re really at the point where the argument against him is basically just to say that if he’s put in a system that’s manifestly suboptimal for him then he won’t be as impactful. Which is obviously going to be true for pretty much anyone, and also is something I don’t really think matters a whole lot because the best players will tend to end up in systems that are geared towards them. The fact that Nash didn’t have that for so long probably just means that he was at a material disadvantage over the course of his career compared to most great players, rather than that he was at an advantage when he did have a system geared around him.

The final thing I’d note is that, if the argument is that Nash’s impact on the Suns was in part a result of being ahead of the times in terms of playstyle, I think that’s probably right, but I also don’t ding players for being ahead of the curve, since everyone is trying to get ahead of the curve and the ones that are able to do so should be celebrated. Not only that, but ultimately most of us take a generally era-relative approach, and if a guy is incredible in era-relative terms because they were ahead of their era tactically, then that might potentially be a reason to think that they wouldn’t be quite as good in later eras but it’s not a reason to think they weren’t extremely good in reality in era-relative terms.

It really seems like he’s being given extra credit for a coach that’s ahead of his time. Like, sure you can say “Nash was the system” and he was obviously more important than D’Antoni, but at the same time if he plays for any of the 29 other teams in the league, I don’t see any evidence he could generate better impact stats than the other candidates.

So yeah, this goes to some of what I said above. I think we should expect that an all-time great player will have a system geared around their strengths. It’s not lucky to have that (and is quite arguably unlucky to have not had it for as long as Nash didn’t have it). Nor am I super concerned with hypotheticals about how impactful a guy would be in a different context that’s not geared around his strengths—both because that’s just a speculative hypothetical that did not happen, but also because generally when we compare players we are comparing guys who did have systems geared around them.

[Doctor MJ]

Haven’t seen a lot of discussion on Westbrook yet. He’s someone that’s going to make my ballot here and I think he peaked higher than Nash who’s getting all the credit at the moment. Let’s compare then in xRAPM over their best years:

Nash (‘05 - ‘07): 3.7, 3.8, 4.9

Russ (‘15 - ‘17): 4.9, 5.9, 5.0

Obviously Westbrook has a massive box score edge and is at least a somewhat better prime defender. Why does everyone default to Nash?

I’m not a hater on Nash compared to the general public or anything, but I’m still not sure why he gets SO much love here or with Thinking Basketball. He has the worst box signals of any player under consideration (including Draymond!) and even his impact signals seem mediocre compared to a lot of the other top players.

I know in the past analyses favoring Nash have relied heavily on team ORtg (ignoring what a liability he is on defense), but here I haven’t seen that much of a case period. It’s just kinda like “well, he’s a little better than (other guy) for a reason (different guy still) would do better than both of them”.

So, when you say "massive box score edge", I get why you say that, but there we're just talking about how the various production stuff get added up into arbitrary all-in-one metrics, not talking about Nash failing to do something.

Certainly fine generally to argue for players based on defense over Nash, but the box score you're referring to here is mostly about offensive production, and again, the idea that Nash is failing to do something offensively is just a non-starter to me.

On the biggest scale for me in this Russ comparison the thing is that Russ guarantees bad decisions that ruin possessions, while Nash is probably the most effective decision maker in the history of the game. Like, these things are not remotely close in my assessment. Fine for others to disagree, but just saying, this isn't like some tiebreaker scenario for me. I literally wouldn't want Westbrook to be the primary decision maker on my contender if I could possibly help it.

[Doctor MJ]

2 Year RAPM, per nbarapm
Nash 07/08, +7.5, 2nd (MDA)
Nash 09/10, +4.7, 7th (no MDA)
Harden 14/15, +5.3, 7th (best)
Harden 17/18, +4.2 12th (2 stretches including the season I've commonly seen as his peak - '18)
Harden 18/19, +4.4 12th
Russ 13/14, +4.5, 9th (best) - missed half the season
Russ 16/17, +4.4, 11th (two stretches including '17)
Russ 17/18, +3.1, 25th

Nash had his best seasons with MDA, no doubt. But he still measured up just fine to Harden/Russ in non-MDA years.

Not a 2 year thing either, fine by 3/4 years 09-12 as well. Yes, Nash in LA was garbage.

So I'll chime in here focused primarily on more data, but I'm also going to just say that I feel like those who are knocking D'Antoni's players because D'Antoni was there coach are really getting themselves confused.

D'Antoni coaches like D'Antoni the player liked to be coached: Set a smart strategy, and let the floor general - which D'Antoni was as a player - figure out the rest. D'Antoni is actively trying to get out of the way of his players, and since that worked with Nash like no one else, D'Antoni deserves a ton of credit for doing that with Nash... but not at the expense of the player he gave control to.

Okay on RAPM, I'm going to emphasize longer spans here because I think that's where not only the stability is, but where the apples-to-apples scaling is more true. You can't use it with a guy who only had a one or two year peak way above the rest of his career, but that's not actually what the data says about Westbrook.

Here are the 4-year numbers for Westbrook during his key impact years:

2012-15: +4.9 (7)
2013-16: +5.3 (7)
2014-17: +5.5 (11)
2015-18: +4.8 (10)
2016-19: +3.7 (21)

This actually a pretty dang gentle arc that I'd say tells a pretty clear story of about how impactful Westbrook was.

And if I grab a similar set of 5 consecutive 4-year spans for Nash, I get:

2004-07: +6.3 (4)
2005-08: +8.8 (1)
2006-09: +7.6 (3)
2007-10: +8.1 (3)
2008-11: +6.9 (3)

So we can see this is a completely different tier.

Now I'll do Harden next, but his is a little bit weirder to look at, because of of his OKC years. If I stick only to years where he actually had 4 years of play, I do get a similar 5 consecutive set of spans topping his list:

2014-17: +6.2 (6)
2015-18: +6.4 (7)
2016-19: +5.1 (12)
2017-20: +4.8 (15)
2018-21: +4.6 (16)

But his 3rd highest span if we include the 2009-12 one where he was only in the league for the last 3 years we see:

2009-12: +5.7 (8)

So this tells a story of him fitting in remarkably well in OKC and this meaning that his impact temporarily went down when he went to Houston and took on an alpha role, from which he rallied and peaked in impact slightly higher than what he did in OKC.

From a peak perspective on tiers, I'd say this largely puts Harden on a tier with Westbrook rather than up there with Nash, while it points out pretty clearly why Harden's career is certainly on a higher tier than Westbrook.

[trelos6]

It's starting to get super close, so I decided to index on a multi year sample, purely as a tie breaker. This will help players with strong 3 year peaks, like Harden 18-20, Ginobili 05-07, Nash 06-08, Draymond 15-17, as they've shown they can sustain the level of excellence over multiple years (less a flash in the pan).

15. Steve Nash 2006-07 (07 > 08 > 06). The best offensive player left on the board. His ability to drive an elite offense is insane and he was an ok scorer in terms of volume with elite efficiency. His net on was +11.4 with a +12 on-off in 2007. Specifically his team were +7.1% rTS% better when he was on the court, with strong rim assist numbers (6.1, +6.5%). I think ranking the playmaking of players left on the board, it's Nash then daylight.

16. Manu Ginobili 2004-05 (05 > 11 > 07). Good scoring volume, great efficiency, good playmaking, great defense, elite impact. +16.4 on, with +17.7 on-off. A lot has been said about Manu in the past thread. Up there with Nash for the best 3 year RAPM of the players left at +8.

17. James Harden 2019-20 (20 > 19 > 18). Hard to pick a year of the 3 elite Harden years. Elite scoring volume with great efficiency also. In 2020 he was a +5.5 net on, with a +10 on-off. His team was +1.6 rTS% better when he was on the court, and they had a crazy high 3PAr (over half their shots), but that makes sense when you look at the roster construction. 2 elite slashers and a bunch of shooters (if you can call 35% "shooters") .

18. Anthony Davis 2019-20 (20 > 18 > 15). +5.2 on, +3.2 on-off (it's hard when the team is built around Lebron). He's one of the best 1B players in history. And at this point, I'd probably take a Harden led team, but definitely not a Dame or Ray Allen led team over AD as the 1B. He has great scoring volume with good efficiency, which in the playoffs became elite efficiency (82% at the rim, 38% from 3). One of the weakest playmakers of the players left, but he brings a strong defensive ability (6 block %, 2 steal %), As well as the best post season of players left, IMO.


Just missed the cut: Embiid, T-Mac, Butler, Draymond

[One_and_Done]

It's starting to get super close, so I decided to index on a multi year sample, purely as a tie breaker. This will help players with strong 3 year peaks, like Harden 18-20, Ginobili 05-07, Nash 06-08, Draymond 15-17, as they've shown they can sustain the level of excellence over multiple years (less a flash in the pan).

15. Steve Nash 2006-07 (07 > 08 > 06). The best offensive player left on the board. His ability to drive an elite offense is insane and he was an ok scorer in terms of volume with elite efficiency. His net on was +11.4 with a +12 on-off in 2007. Specifically his team were +7.1% rTS% better when he was on the court, with strong rim assist numbers (6.1, +6.5%). I think ranking the playmaking of players left on the board, it's Nash then daylight.

16. Manu Ginobili 2004-05 (05 > 11 > 07). Good scoring volume, great efficiency, good playmaking, great defense, elite impact. +16.4 on, with +17.7 on-off. A lot has been said about Manu in the past thread. Up there with Nash for the best 3 year RAPM of the players left at +8.

17. James Harden 2019-20 (20 > 19 > 18). Hard to pick a year of the 3 elite Harden years. Elite scoring volume with great efficiency also. In 2020 he was a +5.5 net on, with a +10 on-off. His team was +1.6 rTS% better when he was on the court, and they had a crazy high 3PAr (over half their shots), but that makes sense when you look at the roster construction. 2 elite slashers and a bunch of shooters (if you can call 35% "shooters") .

18. Anthony Davis 2019-20 (20 > 18 > 15). +5.2 on, +3.2 on-off (it's hard when the team is built around Lebron). He's one of the best 1B players in history. And at this point, I'd probably take a Harden led team, but definitely not a Dame or Ray Allen led team over AD as the 1B. He has great scoring volume with good efficiency, which in the playoffs became elite efficiency (82% at the rim, 38% from 3). One of the weakest playmakers of the players left, but he brings a strong defensive ability (6 block %, 2 steal %), As well as the best post season of players left, IMO.


Just missed the cut: Embiid, T-Mac, Butler, Draymond

Come on. Manu over Luka, who took his team to the finals in 24 and the WCFs in 22? I love Manu, but he could never have done that.

[homecourtloss]

Best individual PO series is an interesting thought. Dray choking out the Rockets in '16 is a personal favorite among guys still eligible. I've never looked at a best relD series list before, but I'd bet that one would rank pretty well.

-19.2 rDRtg
+7.7 rORtg

Massive stuff.

By the way, it's one of Draymond's ten different series in which he was -10 rDRtg or better. Six of these were -13 or better.

[One_and_Done]

A team built around Luka can make the finals. A team built around Draymond is lucky to win 30 games. Draymond is a great player to have, but he's not a franchise player, and as such it's way too early for him.

[f4p]

You don’t have to lecture me on the value of impact stats vs. box stats. I’m as pro-impact stats as anyone. In fact I was recently accused of focusing too much on them myopically at the expense of everything else.

What, you don't love a little "you look at the box score, idiot?" scolding? I only get about 2 a week around here. Of course weirdly when I try to use hardens impact stats from the playoffs, I'm told they are too noisy so we should look at more stable box score stuff. But, even though I haven't actually said a box score stat in this thread or the last, apparently I also can't look at the box score stats in the playoffs when talking about Nash. Also cant use the team being better. Or regular season box score stats. All we can look at for playoff level of play is some regular season RAPM data where Steve Nash on longer scales doesn't measure up with basically anyone in this project but does apparently spike for a few years. Also regular season RAPM seems to be a stat which seems to be as low on harden as is possible so it's basically Nash's best vs Harden's worst (I realize I'm "agreeing" with someone who is lower on harden than the people I'm arguing with).

I mean 2 yr RAPM for harden in 2017/2018 is +4.2, 12th in the league? Am i supposed to take that seriously? Feels like when Kobe is 100th or something in some impact stat, and I'm not even a big Kobe guy. I guess the rockets won 55 and 65 (with bad health) by literally everybody on the rockets except james harden being better than we think they are. What an amazing coincidence.

Eric Gordon and Ryan Anderson were on the same team in 2016 the year before they joined the rockets and were even playing with Anthony Davis so its not like they had a scrub as the alpha (edit: and 1800 minutes of Jrue). What did they win, 55, 60 games, I guess? Oh, 30 wins (I think something like 15-30 when all 3 played so it wasn't injuries). They must have both practiced a lot in the off-season.

The 2018 rockets with Chris paul like 4 to 9 years past his peak (per this project) play perfectly even with apparently 3 guys who might make this project over harden (who were all at their peak or 98% of it). Apparently perfectly expected and no one needs to rethink anything because a loss is a loss, whether by 1 ppg or 20 ppg or if injuries are involved. I guess if harden had been the 10th best player in the league (or as good as RAPM says he was in 2011), the Warriors would have been in real trouble, lol.

2019 James harden sees his 11-14, luxury tax savings-decimated and injuries-decimated team struggling and basically creates through sheer will power a 61 win pace (43-15 finish to season, same win% with and without cp3 so it wasn't cp3 getting healthy or anything), +15 first round playoff team that plays the Steph/KD warriors 6 points per 100 closer than any other team ever, but I guess RAPM and McKayla are unimpressed. Just another crazy coincidence that you can nerf Chris Paul and make Iggy healthy for the whole series and it's still a super close, -1.7 series like the year before when the stars were healthy. Maybe the rockets had a super deep bench. Oh wait, 3 of the 4 guys never played in the nba again and the 4th, Austin rivers, finished the rest of his playoff career as a 6.6 PER, 0.012 WS48, -4 BPM player (-13.1 playoff on/off after leaving the rockets). What a quartet.

And of course 2020 harden has an MVP caliber regular season and then puts up box score playoff numbers (27.5 PER, 0.253 WS48, 9.4 BPM) that are basically identical to 1986 bird, 2005 ginobili, 2009 Kobe, 2017 KD and 2017 steph for basically all of their peak stats (seriously, all 6 of them have weirdly similar peak stats). But RAPM says no.


Not exactly apples.to apples admittedly, but when Nash left the Mavs seemingly in his prime, they improved and made the finals 2 years later. When harden left the rockets, everyone from ownership on down assured us they were in it to win it and didn't need James Harden. To that point, they immediately went on a 6 game winning streak and showed everyone they were right. They finished with the worst record in the league.

[f4p]

Also, a lot of this Nash stuff goes back to the previous peaks project and then definitely the top 100. People keep saying I'm arguing against Nash based on my preferences (over simplistically apparently) but I'm really trying to use the same methodology as his proponents and asking why it doesn't seem to apply to Nash.

Like in the Top 100, for the first 30 picks I was tracking what best modeled the project's picks. I managed to get an R^2 of 0.8. Not my formula, a formula of how the rest of the PC board was voting. The 0.8 seemed a rather remarkable fit. I figured I would be lucky to get 0.5. And other than Moses getting picked 23rd when the best fit predicted he should be 36th, no one was off by more than 10 spots. Plenty of guys were within 1 or 2 spots of the prediction and most everyone was within 5 spots. Steve Nash was 33 spots ahead of the prediction! 24th vs 57th. Again, not 33 spots above where I would put him, but where the project preferences seemed to indicate they would put him. He's just in his own category.

[Cavsfansince84]

Re Nash it seems like he might be the guy who gets the most discussion in this thread and I'm likely about to add him to my ballot so I want to talk a bit on why I would do that. It's not like he does that badly on box score/bpm type stuff. He peaks at 5.9 which is usually going to be like top 5-10 in the league most years. Usually with a top pg though you expect to see pretty good ORtg's. Oscar basically led #1 offenses most every year, Magic likewise, Stockton led some #1 offenses in Utah and obviously I think most of us know that Nash did the same in Phx. Here's a bit more on that though as it relates to Nash:
In the 3 years before he went to Phx(In Dallas) he led the #1 team by ORtg all 3 years. Then he goes to Phx and replaces Marbury and Penny as the pg and they go from 21st in ORtg to 1st and add 33 wins with A'mare breaking out as a big time scorer and with a young Joe Johnson. So you could say it was largely due to A'mare coming into his prime except that then the next year A'mare gets injured and only plays in 3 games. Yet, the Suns are still #2 in ORtg with Nash getting another mvp and still won 54 games. Then they keep having top 2 offenses until Steve leaves.
There's an amazing proof of concept there regarding Nash's ability to lead an offense and then on top of that he's one of the best shooters of all time with 3 straight 200+ ts add seasons on semi low volume. Which just sort of leaves the playoffs where they keep losing to the Spurs who won titles and to the Mavs. So that's the rationale with Steve as I see it. I wasn't that high on Nash back in the day either and thought maybe D'Antoni deserved more of the credit(which honestly the media sort of pushed as well despite the mvps), but it's funny to me that no one does this with with regard to Steph and Kerr. Kerr showed up and the Warriors immediately win 67 and Steph is b2b mvp just like Nash was but people don't think Steph's success is mainly due to his coach. We understand that Kerr just ran offense which helped his players play up to their potential.
Nash was already an all nba guy before he went to Phx and then became a fitness nut which is what pushed him to another level with the Suns.

[70sFan]

Also, a lot of this Nash stuff goes back to the previous peaks project and then definitely the top 100. People keep saying I'm arguing against Nash based on my preferences (over simplistically apparently) but I'm really trying to use the same methodology as his proponents and asking why it doesn't seem to apply to Nash.

Like in the Top 100, for the first 30 picks I was tracking what best modeled the project's picks. I managed to get an R^2 of 0.8. Not my formula, a formula of how the rest of the PC board was voting. The 0.8 seemed a rather remarkable fit. I figured I would be lucky to get 0.5. And other than Moses getting picked 23rd when the best fit predicted he should be 36th, no one was off by more than 10 spots. Plenty of guys were within 1 or 2 spots of the prediction and most everyone was within 5 spots. Steve Nash was 33 spots ahead of the prediction! 24th vs 57th. Again, not 33 spots above where I would put him, but where the project preferences seemed to indicate they would put him. He's just in his own category.

What model did you come up with?

[f4p]

Also, a lot of this Nash stuff goes back to the previous peaks project and then definitely the top 100. People keep saying I'm arguing against Nash based on my preferences (over simplistically apparently) but I'm really trying to use the same methodology as his proponents and asking why it doesn't seem to apply to Nash.

Like in the Top 100, for the first 30 picks I was tracking what best modeled the project's picks. I managed to get an R^2 of 0.8. Not my formula, a formula of how the rest of the PC board was voting. The 0.8 seemed a rather remarkable fit. I figured I would be lucky to get 0.5. And other than Moses getting picked 23rd when the best fit predicted he should be 36th, no one was off by more than 10 spots. Plenty of guys were within 1 or 2 spots of the prediction and most everyone was within 5 spots. Steve Nash was 33 spots ahead of the prediction! 24th vs 57th. Again, not 33 spots above where I would put him, but where the project preferences seemed to indicate they would put him. He's just in his own category.

What model did you come up with?

Model would be a strong word. But it turned out if you just took my age 22-31 (i.e. prime) box score metric, which I've posted somewhere here before, which was weighted 75/25 to the playoffs, multiplied by longevity of prime (admittedly subjectively determined by me but I don't think many would argue that LeBron and Kareem had long primes and i doubt there would be more than a year or two of disagreement on anybody) and then added championships multiplied by some factor and FMVPs by another factor, you got a 0.8 R^2.

And if you take box score multiplied by length of prime, it's literally just cumulative career box score production. And championships and FMVPs is basically just "ringz, erneh". So weirdly, it turned out doing things in the box score and winning lots of titles was doing most of the heavy lifting, despite the fact most voters would have probably said they were looking beyond those things and would consider those more "general board" things.

FMVPs didn't actually help much more than just using championships because most titles won by the top 30 were FMVPs.

Note 1: Nash, Karl Malone, and John Stockton box score were based on age 26-35 because of their much later primes.

Note 2: Nash's prime length was considered 12 years as I believe I included everything from 2001 to 2012.

Note 3: Bill Russell got 9 FMVPs because, uhh, winning 9 out of 10 before they had the award seemed about right.

[iggymcfrack]

A team built around Luka can make the finals. A team built around Draymond is lucky to win 30 games. Draymond is a great player to have, but he's not a franchise player, and as such it's way too early for him.

A team built around Draymond without Steph has been better in the playoffs than a team built around Steph without Draymond for their entire primes together. You can’t use the break season where everyone’s hurt and no one’s trying after 5 straight Finals to judge him.

[One_and_Done]

A team built around Luka can make the finals. A team built around Draymond is lucky to win 30 games. Draymond is a great player to have, but he's not a franchise player, and as such it's way too early for him.

A team built around Draymond without Steph has been better in the playoffs than a team built around Steph without Draymond for their entire primes together. You can’t use the break season where everyone’s hurt and no one’s trying after 5 straight Finals to judge him.

I'm not bading my evaluation on advanced stats. I'm looking at Draymond's complementary skillset and using logic and inference to draw an obvious conclusion. A team built around him is going nowhere.

A handful of games against 500. type teams, while Klay is playing amazing, is not a good argument for Draymond being able to take a team to the WCFs, never mind the finals.

[One_and_Done]

1. Luka (2022)
2. Nash (2007, 05, 06)
3. Harden (2019)
4. Butler (2023)

HM: T-Mac, AD, Tatum, Dwight

In terms of Luka, the argument is pretty clear. The guy basically carried a fairly solid-ish support cast to the finals, and only lost because his team was totally outgunned (and because he was hurt). If Luka had been healthier in 2024 I’d have that as the year to vote for him, but he was sufficiently banged up in the playoffs for it to be a demerit. I’ll take the younger and springier Luka from 2022, who “only” got them to the WCFs.

In terms of skillset, Luka is way above guys like Kobe. He runs an offense, which Kobe can’t, and he can score and set up guys in a way that is levels above Kobe, a guy who bizarrely got in already. Ok, Luka’s defence isn’t good, but you can get away with that when you’re point guard. Kobe’s D was overrated and average after 2004 anyway.

Looking at numbers, it’s not even close.

Luka 22 RS: 40/13/12 per 100, on 571 TS%
Kobe RS 09: 38/7/7 per 100, on 561 TS%

Luka 22 PS: 45/14/9 per 100, on 577 TS%
Kobe PS 09: 39/7/7 per 100, on 564 TS%

It’s not even close. Luka is a guy who will elevate a bad team more, and raise the ceiling of a good team higher as well.

The other 3 are fairly straightforward. Like Luka, Nash and Harden are heliocentric offensive players, who elevate your offense to an elite level, which is one of the most important skills in basketball (alongside anchoring an elite defence). It probably sounds strange to call Nash heliocentric, given he didn’t “score” in volume, and he is different to Harden and Luka, but he still controlled the Suns offense, and their horrific record in his absence shows how much he was elevating them.

Harden’s stats drop in the playoffs, but they merely drop from his absurd level. His playoff numbers are still better than guys who already got in, like Kobe. Take Harden’s insane 2019 season, which won’t be a common choice because the team around him wasn’t as good, but that wasn’t his choice.

Harden 2019 RS: 48/9/10 per 100, on 616 TS%
Harden 2019 PS: 41/9/9 per 100, on 567 TS%

Kobe wishes he had numbers like that. The Rockets were often carried by Harden with some pretty meh support casts too. His impact on their success was obvious. Take the 2015 Rockets. Their starters next to Harden were Ariza, Montiejunas, Dwight (only played 41 games), and Pat Bev (only played 56 games), with Terrence Jones sucking up the missed time from Dwight. That is a team that had no business winning 56 games.

As for Jimmy, he is giving you enough of Kobe’s scoring, on better efficiency, but is supplementing that with elite D, and a vastly better floor game, that just leads to more winning. I’m not entirely sure why we’d focus on things like “ppg” in 1 series is a sensible response.

Sometimes stats aren’t everything, because volume stats can’t always capture the impact a guy is having. That said, when you look at say Jimmy Butler’s 2022 playoffs it looks better than Kobe’s 09 playoffs.

Butler 22 PS per 100: 38/10/6, on 604 TS%
Kobe 09 PS per 100: 39/7/7, on 564 TS%

Even raw numbers, which don’t capture Butler’s D and floor game, seem to favour him.

Butler’s absurd carry jobs are being too quickly forgotten. All three of his playoff runs, in 20, 22, and 23, were absurd. He took a starting group of Gabe Vincent, Max Struss, old Kevin Love, and Bam, to the finals. That’s ridiculous.

Butler’s impact on those teams is borne out by the win loss record also. From 20-23 the Heat were 144-81 with Butler, and only 37-41 without him. I’m going with 2023 Butler to account for injuries, even though he was a bit banged up by the finals, just because 64/82 games is enough, given all he did in the playoffs. His 2020 health is better than you think too, given there was a shorter season, he actually played 58/73 games.

Butler’s stats don’t jump off the page as some do, although as I noted above he sure looks comparable to someone like Kobe per 100, but his impact does. Stats don’t always capture stuff like your floor game, or defence, or the little things like intangibles, but it was very clear Jimmy had all those.

[Doctor MJ]

Also, a lot of this Nash stuff goes back to the previous peaks project and then definitely the top 100. People keep saying I'm arguing against Nash based on my preferences (over simplistically apparently) but I'm really trying to use the same methodology as his proponents and asking why it doesn't seem to apply to Nash.

Like in the Top 100, for the first 30 picks I was tracking what best modeled the project's picks. I managed to get an R^2 of 0.8. Not my formula, a formula of how the rest of the PC board was voting. The 0.8 seemed a rather remarkable fit. I figured I would be lucky to get 0.5. And other than Moses getting picked 23rd when the best fit predicted he should be 36th, no one was off by more than 10 spots. Plenty of guys were within 1 or 2 spots of the prediction and most everyone was within 5 spots. Steve Nash was 33 spots ahead of the prediction! 24th vs 57th. Again, not 33 spots above where I would put him, but where the project preferences seemed to indicate they would put him. He's just in his own category.

What model did you come up with?

Model would be a strong word. But it turned out if you just took my age 22-31 (i.e. prime) box score metric, which I've posted somewhere here before, which was weighted 75/25 to the playoffs, multiplied by longevity of prime (admittedly subjectively determined by me but I don't think many would argue that LeBron and Kareem had long primes and i doubt there would be more than a year or two of disagreement on anybody) and then added championships multiplied by some factor and FMVPs by another factor, you got a 0.8 R^2.

And if you take box score multiplied by length of prime, it's literally just cumulative career box score production. And championships and FMVPs is basically just "ringz, erneh". So weirdly, it turned out doing things in the box score and winning lots of titles was doing most of the heavy lifting, despite the fact most voters would have probably said they were looking beyond those things and would consider those more "general board" things.

FMVPs didn't actually help much more than just using championships because most titles won by the top 30 were FMVPs.

I actually "model" is very much an appropriate word and applaud the use of such a technique, but I'd note the saying,

"All models are wrong, but some are useful"

As in, it's not a question of if there's something wrong with the model, but what is wrong with it, because all models have something wrong.

In this case, if I'm understanding your explanation, you were able to take a bunch of factors and train them - which might not have been a formal process - to correlate with one set of voter preferences pretty well, but then when tested on a new sample, it doesn't correlate well.

What conclusion would I generally draw from that? Sounds like classic overfitting. Correlation is not causation. You used a set of factors that voters were not actually using exclusively, or even directly, and it worked except when it didn't.

But I will say, it's certainly not random that the guy who didn't correlate that you noticed first was Nash, because Nash's box score isn't why he won the MVPs, perceived impact was. And that was incredibly controversial at the time and led to all sorts of debates which included dismissals of what Phoenix was doing as a gimmick.

But as we started getting access to more and more precise data we could associate with impact (+/- with regression), it largely supported the perception of those who advocated for Nash, and thus the question of "How is the box score biased relative to actual impact?" came front and center.

In other words, your model isn't pointing something out that's unexpected and hard to explain, it's just re-affirming the divergence between box-based analysis and impact indicators and my guess would be that wherever we'd see that divergence, if we had a specific debate about those players, it would break down similarly.

[ShotCreator]

A team built around Luka can make the finals. A team built around Draymond is lucky to win 30 games. Draymond is a great player to have, but he's not a franchise player, and as such it's way too early for him.

What do these words mean? What constitutes ‘built’ around Draymond? What constitutes ‘franchise’ player? If you could point to some benchmarks for those words?

Because I can tell you now, if it’s just scoring numbers, there’s a lot, and I mean a ton of examples that can be used to show scoring isn’t the highest impact method of playing basketball. Especially compared to a guy who does everything at a high level.

[lessthanjake]

Haven’t seen a lot of discussion on Westbrook yet. He’s someone that’s going to make my ballot here and I think he peaked higher than Nash who’s getting all the credit at the moment. Let’s compare then in xRAPM over their best years:

Nash (‘05 - ‘07): 3.7, 3.8, 4.9

Russ (‘15 - ‘17): 4.9, 5.9, 5.0

Obviously Westbrook has a massive box score edge and is at least a somewhat better prime defender. Why does everyone default to Nash?

I’m not a hater on Nash compared to the general public or anything, but I’m still not sure why he gets SO much love here or with Thinking Basketball. He has the worst box signals of any player under consideration (including Draymond!) and even his impact signals seem mediocre compared to a lot of the other top players.

I know in the past analyses favoring Nash have relied heavily on team ORtg (ignoring what a liability he is on defense), but here I haven’t seen that much of a case period. It’s just kinda like “well, he’s a little better than (other guy) for a reason (different guy still) would do better than both of them”.

So, when you say "massive box score edge", I get why you say that, but there we're just talking about how the various production stuff get added up into arbitrary all-in-one metrics, not talking about Nash failing to do something.

Certainly fine generally to argue for players based on defense over Nash, but the box score you're referring to here is mostly about offensive production, and again, the idea that Nash is failing to do something offensively is just a non-starter to me.

On the biggest scale for me in this Russ comparison the thing is that Russ guarantees bad decisions that ruin possessions, while Nash is probably the most effective decision maker in the history of the game. Like, these things are not remotely close in my assessment. Fine for others to disagree, but just saying, this isn't like some tiebreaker scenario for me. I literally wouldn't want Westbrook to be the primary decision maker on my contender if I could possibly help it.

Yeah, I actually find it to be a bit of a mystery/puzzle why box-score metrics are so low on Nash. It’s not something that you’d expect from looking at the raw box score numbers themselves. For instance, in the 2005 playoffs, Nash put up 23.9 PPG, 11.3 APG, and 4.8 RPG with +9.0 rTS%. Those look like great numbers (especially in a low-pace and low-efficiency era where even the Suns only had a 94.1 pace in those playoffs)! It’s really not immediately obvious at all that someone with those numbers would have relatively weak box-metric numbers (note: his BPM in those playoffs was a relatively modest 4.7). Indeed, those box numbers would fit right in with peak CP3 years, which grade out super highly by box metrics. So how could Nash put up those numbers and have box metrics like BPM grade him out at only like half of what we see when other guys put up similar numbers? I think there’s two things happening here, as well as a third issue that makes even those raw box numbers I just listed understate Nash’s impact:

1. If I had to guess, I’d say one thing I think the models are seeing is Nash having abnormally low number of steals. I think they’re punishing him a lot for that. Which, on one hand, is actually getting at something negative about his game. It is actually reflective of his relatively weak defense. But I’ll note that he took quite a lot of charges—like amongst the highest in the league, and around 0.5 more per game than a normal player—which I don’t think ends up factored into the vast majority of box-score metrics (someone can feel free to correct me on that, to the extent I’m wrong about some particular box metric or box component, but I know it’s not factored into things like BPM, WS, and PER). The result is that RAPM parsed out by factor actually has him as a neutral or only very slight negative (and more like neutral in his peak years) in terms of his impact on opponent’s turnovers—similar to someone like Tyrese Haliburton. I think box metrics will generally assume he’s a bigger negative there than he is. This steals thing also ends up bleeding into offensive box measures—for instance, steals are factored into OBPM. Which makes some sense, since steals lead to easy transition opportunities the other way (which is something the charges he took didn’t do, so that doesn’t mitigate this issue). But Nash had no problem producing a ton of transition for his team regardless. Anyways, yeah, overall, I think this is one of the factors the models are seeing, and it’s a legitimate one to ding him for, but he’s probably dinged a little bit too much for it when we realize that few of these models are accounting for taking charges and Nash was genuinely doing that way more than virtually any other player.

2. Another big thing I think these models are seeing is that Nash has a fairly large amount of turnovers himself. And limiting turnovers is important, so this does matter! But Nash had the ball a huge amount and the offensive system his team ran basically had him taking all the team’s risks (this is especially so given how uniquely often he kept his dribble after driving to the basket). Consequently, if we look at RAPM that parses impact out by different factors, Nash comes out as a genuinely positive-impact player when it comes to limiting his team’s turnovers. He’s not quite Chris Paul in that regard, but he’s actually legitimately in a similar ballpark, despite having a lot more turnovers. This is a factor where I think there’s just going to be a huge delta between what a box metric sees and what the actual reality is in terms of impact. Box metrics are going to think Nash is a negative in terms of turnovers, when he was actually a significant positive.

3. Finally, given the disparity between box metrics and RAPM, one major thing worth thinking about is what area we think RAPM is higher than box-score data on Nash and whether we think the box score can adequately understand Nash’s impact in that regard. If it can’t, then that’d give us good reason to think box metrics (and even the raw stats I listed at the beginning of this post) are underrating him. Given that we have six-factor RAPM, we can actually be pretty sure where RAPM has Nash deriving his huge impact. It is in his team’s TS%. Nash has massive RAPM impact in that regard. Indeed, I can’t find anyone with RAPM impact as high as Nash in terms of impact on the team’s TS%. Steph and LeBron get close, but even they don’t get as high as Nash. And guys like Harden, Luka, Wade, and Jokic are a step below that. Meanwhile, other high-assist PGs like CP3 and Stockton are even below that. Part of this impact comes in the player’s own TS%—which obviously affects the team’s overall number. Box data does know that Nash was a really efficient scorer. But so much of this is about shot quality produced for teammates (especially for a relatively low-volume scorer like Nash), which is really just far too complex for box-score data to account for. The box score does know Nash had a lot of assists, but it doesn’t know that he’s impacting his team’s TS% more than anyone else—including guys with similar assist numbers to him. More complicated box-score measures may be able to pick up on the fact that Nash has outlier-levels of rim assists, but even that is a bit of a blunt instrument that wouldn’t really allow the model to be able to truly account for Nash’s TS% impact. After all, guys like CP3 and Stockton—who at least approach Nash’s level of rim assists—really don’t approach his impact on TS%, so if you put a lot of weight on rim assists such that you’re actually getting at Nash’s TS% impact, you’d way overrate those other guys. So basically, TS% impact is a massive factor for Nash that I really don’t think box metrics can account for without throwing the model out of whack for like everyone else.

_________

I think issue #3 is probably the biggest issue for box data when it comes to assessing Nash. Box data can’t really get at shot quality created. It can try to approximate it by using data like assists, but if you scaled up the value that a model put on assists such that it was giving Nash credit for how much he was improving his team’s TS%, you’d end up wildly overrating like every other high-assist guy. So the models don’t do that, because they’re designed to provide a line-of-best-fit in general. And that leaves Nash underrated. This is significantly exacerbated by the fact that box score models can’t quite understand that Nash was taking an abnormally high percent of his team’s risks and therefore he had a notably positive impact on his team’s turnovers despite having a lot of turnovers himself. And it’s also probably at least slightly exacerbated by the fact that box metrics generally aren’t seeing that he was taking like 0.5 charges a game more than a normal player.

Of course, one response to this might be for someone to say that they think the box models *are* adequately accounting for Nash’s TS% impact and that RAPM is simply overrating it. And I guess my response to that is that I just really don’t think you’d come to that conclusion if you watched a lot of him at the time. He was just an absolute alien in terms of producing good quality shots for his team.

[homecourtloss]

I took a quick look at players who have been voted in and/or are currently being discussed right now

Spoiler:

LeBron, Duncan, Jokic, Curry, KG, Giannis, Shaq, Wade, Kawhi, Kobe, Dirk, SGA, CP3, KD, Embiid, Draymond, Manu, Tatum, AD

to see how they did without the other best player on the team and as you would imagine, they all did well with the exception of one player. I also took at look at each player’s best 2 yr, 3 yr, and 5 yr RAPM stretches, and every player, except one, had at least a top 5 finish finish somewhere, with most at least a top 5 in a 2 yr stretch. Now, this wouldn’t necessarily exclude AD from a top 20 single year peak, but he’s a very glaring exception here.

AD’s rank of 15th in his best 2 yr RAPM among this group of players, is a massive outlier statistical outlier by the IQR method, the standard Z-score method, and the modified z-score method. 16 players discussed so far have a 2yr RAPM stretch that was either the highest or 2nd highest in the league.

He’s also the only player discussed so far who was a negative on court without the team’s other best player in a season discussed as a peak, i.e., 2020 (and 2018 for that matter).

Highest 2 yr RAPM rank Occurrences



Peaks without second best player and best 2yr/3yr/5yr RAPM ranks

2009 LeBron
+17.1 without Andy V.
+13.5 without Mo
+12.0 without Big Ben
2012 LeBron
+5.6 without Wade
2013 LeBron
+9.3 without Wade
2016 LeBron
+13.3 without Kyrie
*2020 LeBron (for reference)
+10.3 without AD

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 1st

2003 Duncan
+6.4 without DRob
+9.3 without Manu

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 1st

2023 Jokic
+13.3 without Murray

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 1st

2017 Curry
+14.6 without KD

2016 Curry
+3.4 without Draymond

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 1st

2001 Shaq
+4.7 without Kobe

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 2nd

2004 KG
+3.7 without Cassel

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 1st

2021 Giannis
+5.8 without JRue
+8.3 without Middleton

Best two year RAPM stretch: 2nd
Best three year RAPM stretch: 3rd
Best five year RAPM stretch: 2nd

2006 Wade
+6.5 without Shaq
2009 Wade
All teammates trash

Best two year RAPM stretch: 2nd
Best three year RAPM stretch: 3rd
Best five year RAPM stretch: 3rd

2017 Kawhi
+6.6 without Green
+13.6 without LMA
2019 Kawhi
+6 without Lowry (Lowry without Kawhi was a wild +13.6)
-7 without Green

Best two year RAPM stretch: 2nd
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 1st

2025 SGA
+17.6 without Chet

Best two year RAPM stretch: 2nd
Best three year RAPM stretch: 2nd
Best five year RAPM stretch: 9th (this will very, very likely change)

2011 Dirk
+14 without Kidd
+9 without Chandler

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 3rd

2009 Kobe
+2.1 without Odom
2008 Kobe
+7.1 without Gasol

Best two year RAPM stretch: 6th
Best three year RAPM stretch: 5th
Best five year RAPM stretch: 5th

2015 CP3
+9.9 without Griffin

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 2nd
Best five year RAPM stretch: 1st

2018 Harden
+8.6 without Paul

Best two year RAPM stretch: 7th
Best three year RAPM stretch: 9th
Best five year RAPM stretch: 5th

2014 KD
+8.7 without Russ
2016 KD
+8 without Russ

Best two year RAPM stretch: 5th
Best three year RAPM stretch: 3rd
Best five year RAPM stretch: 5th

2016 Draymond
+10.8 without Curry

Best two year RAPM stretch: 2nd
Best three year RAPM stretch: 3rd
Best five year RAPM stretch: 4th

2005 Manu
+9.9 without Duncan

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 1st
Best five year RAPM stretch: 2nd

2005 Nash
+7.1 without Amar’e
-.2 without Marion in 300 minutes
2007 Nash
+12.5 without Amar’e

Best two year RAPM stretch: 2nd
Best three year RAPM stretch: 2nd
Best five year RAPM stretch: 3rd

2019 Embiid
+7.6 without Butler
+8.5 without RoCo
+7.7 without Simmons

Best two year RAPM stretch: 1st
Best three year RAPM stretch: 2nd
Best five year RAPM stretch: 3rd

2020 AD
-2.4 without LeBron
2018 AD
-.6 without Jrue

Best two year RAPM stretch: 15th
Best three year RAPM stretch: 26th
Best five year RAPM stretch: 24th

2024 Tatum
+15.4 without Jrue
+15.7 without Brown

Best two year RAPM stretch: 5th
Best three year RAPM stretch: 2nd
Best five year RAPM stretch: 5th