Top 25 peaks of the 2001-25: #1-#2 Spots

[One_and_Done]

How can you say that net rating doesn’t reliably tell us anything right after you just used team record to judge impact? Isn’t net rating just measuring the same thing you valued more accurately? Like you think how well the team plays with a player on the floor and on the bench matters, but not how they actually play with just said player?

I am not basing my evaluation of players on team record, I'm looking at things like how the team performed and the context of that performance. The 00s Pistons had alot of success, but the context was they were an ensemble cast with no true star among them. The context of 02 Duncan is he took a rubbish team to 58 wins. Unfortunately there is no formula for context. It requires some common sense and analysis.

I also note you did not answer my question, despite getting very upset over my characterisation of your views. Do you still hold to the view that you don't need to watch games to rate players? If you've changed your view on this that's fine, but I think it's important you make that clear. If you're not going to reply I will continue to assume you still hold to your previously expressed views.

I do think that watching games helps around the edges, but ultimately I trust hard data much more than my memory. Unless you really study a lot of film on a specific player watching that player only through every play on both sides of the floor, it’s pretty hard to actually come up with a true valuation. If anything I’d say watching a guy play a lot generally helps me contextualize the data more often than it does give me fresh insights, although the latter does sometimes happen.

I guess what I'd note is:
1) It's a false binary to say 'well, we can't only rely on our eyes, so we should rely on stats instead'. As with most things, neither extreme is right. Those stats have lots of reasons they can be wrong, and often are, so we shouldn't be relying on them either.
2) The memory and eyes of one person can be misleading, which is why the wisdom of the crowd can sometimes find insights that stats won't, and rely on alot of different people working together.
3) Basketball is a different type of sport to baseball, it can't be as neatly captured by advanced stats (or any stats, whatever you call them). Baseball is mostly a series of individual encounters, whereas basketball is much more dynamic and chaotic.

[iggymcfrack]

I am not basing my evaluation of players on team record, I'm looking at things like how the team performed and the context of that performance. The 00s Pistons had alot of success, but the context was they were an ensemble cast with no true star among them. The context of 02 Duncan is he took a rubbish team to 58 wins. Unfortunately there is no formula for context. It requires some common sense and analysis.

I also note you did not answer my question, despite getting very upset over my characterisation of your views. Do you still hold to the view that you don't need to watch games to rate players? If you've changed your view on this that's fine, but I think it's important you make that clear. If you're not going to reply I will continue to assume you still hold to your previously expressed views.

I do think that watching games helps around the edges, but ultimately I trust hard data much more than my memory. Unless you really study a lot of film on a specific player watching that player only through every play on both sides of the floor, it’s pretty hard to actually come up with a true valuation. If anything I’d say watching a guy play a lot generally helps me contextualize the data more often than it does give me fresh insights, although the latter does sometimes happen.

I guess what I'd note is:
1) It's a false binary to say 'well, we can't only rely on our eyes, so we should rely on stats instead'. As with most things, neither extreme is right. Those stats have lots of reasons they can be wrong, and often are, so we shouldn't be relying on them either.
2) The memory and eyes of one person can be misleading, which is why the wisdom of the crowd can sometimes find insights that stats won't, and rely on alot of different people working together.
3) Basketball is a different type of sport to baseball, it can't be as neatly captured by advanced stats (or any stats, whatever you call them). Baseball is mostly a series of individual encounters, whereas basketball is much more dynamic and chaotic.

I mean if you watched every single play of a player through an entire season, then maybe you’d have an argument for knowing them as well as their statistical profile although the stats would still probably be more accurate. For instance let’s say after watching an entire season of film how accurately would you be able to guess their shooting percentage with no other data, if you hadn’t seen it on the broadcast.

Now you watch the same player play say 8 regular season games and half their playoff games in a given season while watching the game for entertainment and following the ball without giving particular attention to how they’re making their defensive rotations, how much value does that have? Maybe 5% as much value as the numbers? That sounds about right.

Like I know from watching games that Kobe was extra selfish and didn’t seem to have the team’s best interests at heart due to his jealousies. But that’s not going to make me outright judge him worse than any stats would. It will just make me trust the impact stats a little more that say he was outside the top 30 than the box stats that say he was on the borderline of the top 15. I’ll still weigh both though in ultimately putting him in the top 25. Now if I was grading the 2006 season in particular, the eyes test would be a big factor as I remember for a fact seeing him quit on his team in a Game 7, but that’s a special case.

If I’m trying to compare say Joker and SGA last season, the eye test says that both guys are absolutely incredible and amazing and I watched them both a ton, but they were both so good and it was so tight that other than my underlying knowledge of Jokic being an S-tier passer that goes beyond assists, my eye test can’t do much. There’s such a fine line between them, only a comprehensive statistical study of every play they played all year could posssibly begin to separate them.

[lessthanjake]

I am not basing my evaluation of players on team record, I'm looking at things like how the team performed and the context of that performance. The 00s Pistons had alot of success, but the context was they were an ensemble cast with no true star among them. The context of 02 Duncan is he took a rubbish team to 58 wins. Unfortunately there is no formula for context. It requires some common sense and analysis.

I also note you did not answer my question, despite getting very upset over my characterisation of your views. Do you still hold to the view that you don't need to watch games to rate players? If you've changed your view on this that's fine, but I think it's important you make that clear. If you're not going to reply I will continue to assume you still hold to your previously expressed views.

I do think that watching games helps around the edges, but ultimately I trust hard data much more than my memory. Unless you really study a lot of film on a specific player watching that player only through every play on both sides of the floor, it’s pretty hard to actually come up with a true valuation. If anything I’d say watching a guy play a lot generally helps me contextualize the data more often than it does give me fresh insights, although the latter does sometimes happen.

I guess what I'd note is:
1) It's a false binary to say 'well, we can't only rely on our eyes, so we should rely on stats instead'. As with most things, neither extreme is right. Those stats have lots of reasons they can be wrong, and often are, so we shouldn't be relying on them either.
2) The memory and eyes of one person can be misleading, which is why the wisdom of the crowd can sometimes find insights that stats won't, and rely on alot of different people working together.
3) Basketball is a different type of sport to baseball, it can't be as neatly captured by advanced stats (or any stats, whatever you call them). Baseball is mostly a series of individual encounters, whereas basketball is much more dynamic and chaotic.

I feel like point #1 you make above is not consistent with your position that you do not care about “advanced stats” at all. I agree with you that neither extreme is right, but you often seem to espouse a view that genuinely is on one of the extremes.

[One_and_Done]

I do think that watching games helps around the edges, but ultimately I trust hard data much more than my memory. Unless you really study a lot of film on a specific player watching that player only through every play on both sides of the floor, it’s pretty hard to actually come up with a true valuation. If anything I’d say watching a guy play a lot generally helps me contextualize the data more often than it does give me fresh insights, although the latter does sometimes happen.

I guess what I'd note is:
1) It's a false binary to say 'well, we can't only rely on our eyes, so we should rely on stats instead'. As with most things, neither extreme is right. Those stats have lots of reasons they can be wrong, and often are, so we shouldn't be relying on them either.
2) The memory and eyes of one person can be misleading, which is why the wisdom of the crowd can sometimes find insights that stats won't, and rely on alot of different people working together.
3) Basketball is a different type of sport to baseball, it can't be as neatly captured by advanced stats (or any stats, whatever you call them). Baseball is mostly a series of individual encounters, whereas basketball is much more dynamic and chaotic.

I feel like point #1 you make above is not consistent with your position that you do not care about “advanced stats” at all. I agree with you that neither extreme is right, but you often seem to espouse a view that genuinely is on one of the extremes.

I consider it to be a data point, albeit an unreliable one. It doesn't necessarily tell you nothing, but it's not a sensible way to distinguish between great players. I prefer a range of other factors. There are plenty of stats that are more helpful, e.g. team stats, or TS%, etc.

[lessthanjake]

I guess what I'd note is:
1) It's a false binary to say 'well, we can't only rely on our eyes, so we should rely on stats instead'. As with most things, neither extreme is right. Those stats have lots of reasons they can be wrong, and often are, so we shouldn't be relying on them either.
2) The memory and eyes of one person can be misleading, which is why the wisdom of the crowd can sometimes find insights that stats won't, and rely on alot of different people working together.
3) Basketball is a different type of sport to baseball, it can't be as neatly captured by advanced stats (or any stats, whatever you call them). Baseball is mostly a series of individual encounters, whereas basketball is much more dynamic and chaotic.

I feel like point #1 you make above is not consistent with your position that you do not care about “advanced stats” at all. I agree with you that neither extreme is right, but you often seem to espouse a view that genuinely is on one of the extremes.

I consider it to be a data point, albeit an unreliable one. It doesn't necessarily tell you nothing, but it's not a sensible way to distinguish between great players. I prefer a range of other factors. There are plenty of stats that are more helpful, e.g. team stats, or TS%, etc.

Why would those stats you named be “more helpful”? It is certainly the case that those are “unreliable” ways of distinguishing between great players, because they miss absolute boatloads of context. And, in many cases, the “advanced stats” that you barely consider actually aim to account for the context missed by the stats you’re saying are more helpful.

Also worth noting that those “advanced stats” are also often adjusted in order to be consistent with team stats (that’s inherently true of RAPM, and meanwhile box-metrics like BPM include adjustments to make sure that the overall values across the team are consistent with how well the team actually did). So the team stats you say are “more helpful” are basically baked into the pie of the “advanced stats” you say are not helpful.

In any event, how is your position any different than the opposite extreme you’re criticizing? Your view seems to be that “advanced stats” could theoretically be useful but are not a “sensible way to distinguish between great players” and therefore shouldn’t be relied on in discussions like this. How is that different than what you’re criticizing—which is someone holding essentially the exact same position about the eye test?

[tsherkin]

Hmm.

My early thoughts for top-4 would be 09 Lebron, 00 Shaq, 23 Jokic and some version of Duncan, 02 or 03. I need a minute to think and I'm mid-set, so I don't have time for a thoughtful post, but I'll drop something over the weekend.

Just a quick note before you make your final decision: 2000 Shaq wouldn't qualify for this project, only 2001 Shaq.

Right! Literacy!

This is what happens when I get back to shifts, lol. Let's say 2001 as an amendment. Really anything from that 00-02 stretch works for me.

[Doctor MJ]

Just in a nutshell: It's unrealistic to have positive +/-, and thus I believe high RAPM, when you're a big minute guy on the losing team in the series. As such, I believe there's likely a kind of "winning vs losing bias" thing wherein players lead teams with first & second round exits are going to see their RAPM hurt on average even if they don't actually underperform relative to perception.

Alright so heading over to Cheema's study from 1997-2021 career Postseason RAPM, the top guys from #1 to Shaq:

1. LeBron James +5.875 (don't assume this scale is the same as nbarapm study)
2. Draymond Green +5.483
3. Manu Ginobili +5.169
4. Kevin Garnett +4.767
5. Tim Duncan +4.289
6. Steph Curry +4.117
7. James Harden +4.106
8. Shaquille O'Neal +3.932

Shouldn't stuff like this force people to be higher on #19 peak harden? He's basically tied with Steph and duncan while being the only non-KG guy who wasn't on a dynasty, which you say above tends to make it tough to look good in these measures, like we see with Manu and green.

Well, I can only speak for myself, and for myself I'm not coming in with any assumption of where Harden should be.

But I would also note that you just singled out the measure I posted where Harden looks best - so there are other perspectives that look worse - and we are talking about a career measure here rather than a peak/prime measure. We should remember Harden's best Playoff RAPM span almost certainly has a significant OKC component, and while I'm not looking to knock his time in OKC at all - I thought he was severely underrated - this likely means that around the time of his MVP he probably wouldn't rank quite as high as this study suggests.

On this note, I'm going to take data from a source I'm really cautious about using because a) playoff RAPM is fraught with issues, b) single post-season RAPM data all the more so, and c) I don't know anything about who the source actually is, but it's what we have and it's relevant here:

Single-season RS & PS RAPM studies on Github

So for Harden, in the playoffs, this is where this source rates Harden, fwiw:

OKC
'09-10: +0.3665 (50th & below Durant, Westbrook)
'10-11: +1.9783 (6th & above them)
'11-12: +2.7003 (3rd & above them)

Houston
'12-13: -.5004 (negative)
'13-14: -.0815 (negative)
'14-15: +0.2547 (67th)
'15-16: -1.4063 (negative)
'16-17: +0.6537 (37th)
'17-18: +1.6445 (8th)
'18-19: +1.1475 (26th)

That's as far as their studies go, and we should note that the Cheema study goes two more years which could have helped some. In neither case is it factoring in the last 4 years.

So yeah, we at least have a study that seems to indicate that Harden's playoff RAPM peaked in OKC, and while that study may be wrong - we should not trust it blindly - it's not a surprise given what Harden's On-Off looked like toward the end of his time on the Thunder.

Now, should Harden get "knocked" because he was awesome in OKC? Not from any kind of career perspective obviously, but if we're trying to get a sense of what a player's Playoff RAPM was around his MVP candidate years, it does indicate that his arc of post-season impact might not be what we'd tend to expect based on typical player careers.

I think Curry & Green being next to each other is fortuitous. As I always say, I welcome people seeking to argue Green > Curry, I just tend to push back against people who want to use Green to tear down Curry. Did Green help Curry get the chips? Absolutely. Could that mean their On & On-Off are each inflated? Yes. Does inflation explain why Curry an all-timer on ORAPM while Green is an all-timer by DRAPM? No, it really doesn't. Both of these guys are worth discussion among the all-timers.

It's not necessarily about tearing Steph down as seeing him being the most kids gloves treated superstar ever (see below). The same people who will say we really need to focus on RAPM will then say we can't look at green as being Steph's equal, essentially because it just doesn't make sense. But of course the whole point of RAPM is supposed to be that it sees beyond what makes sense and tells us the truth. If it only does that when it makes sense, then it's tough to use. I mean KD, who supposedly doesnt impact much of anything (even in non warriors contexts), beats Steph in playoff RAPM in the engelmann data and Draymond beats Steph fairly regularly and people will acknowledge this (well, not the KD part), but then just go back to it was all Steph. Or jump on the one year with crazy outlier impact in 2017 as if it is essentially the validating data for what everyone sees in the regular season numbers for Steph through his prime and not the outlier.

Or we have people describing steph as like an outlier among outliers in offensive impact and then he's not even 1st in playoff ORAPM or even second behind LeBron.

I mean i wouldnt put green over curry but if I was mostly sticking to RAPM then it would seem like a serious convo.

So, I'd say what bugs you is a perception of goalpost moving by people who claim to be rational but continue to side with the same guys no matter what. I get that and I'm not asserting that I have any kind of immunity to that. I try to be rational, but I'm also sure that I have emotions warping my reasoning just like everyone else, and so part of trying to be rational for me is trying to identify the existing knots of inconsistency in what I think I know.

So I'm game to discuss this stuff because I want to get better, but it's also only going to be fruitful if I'm talking to someone who really is talking to me rather than a larger group that I'm taken to be part of.

And one of the things I really need to emphasize is that there are concerns about using RAPM in the playoffs as if that data will give you something apples-to-apples with what we think of it as meaning in the regular season. Playing a small number of games against a far smaller number of opponents just means there's going to be a huge amount of noise. I'm happy to discuss PS RAPM, but from a perspective of "You value RAPM so much in the RS but ignore the PS, what's up with that?", well, noise is what's up.

To give another example on this that I've said for a long-time: I really like using regression-based models for the modern NBA, but that doesn't mean I'm itching to get that data for the deep past because I don't actually think it would give us meaningful information for guys like Wilt & Russell who played basically all meaningful minutes, and thus their Off sample would perpetually be noisy even after years and years of sample.

So I'll probably always talk more about RS RAPM data than PS RAPM data even as I acknowledge that legacy is determined primarily from the post-season. And this definitely leaves room for the possibility of a player being much less or much more effective in the playoffs relative to rivals than RS RAPM would indicate, and once again drives us back to the final step of analysis not being following stats but explaining how we come to differ in assessment from them.

Considering all of that plus the fact that we've seen Curry leader quality seasons without Green playing major minutes, but when the reverse situation happened the team fell off a cliff (maybe due to apathy, but a missed opportunity for proof),

I mean the team lost its top 3 offensive weapons. It would effectively be like if Steph had to play for a year as the best defender on the team (and not just relative to position, like actual best) and also that team didn't really have good offensive players either. And the team tanked. And the team also sucked even when Draymond played (like it did when steph played). And Steph only played like 60% of the season.

Sure, but there's levels to this. No one was expecting Draymond to lead them to a chip sans Splash Bros, but that doesn't they had to be as bad as they were, and it certainly doesn't mean we should treat it as proof of Draymond being able to do something we didn't see him do.

Also, the fact that you're alluding to fit is something I very much approve of. Being able to say that Green specifically needs to be playing with scorers to do his offensive thing while Curry needs other things to round out his play is helpful.

Here's where I'll also say that I'm quite cautious about trying to normalize for fit when doing player evaluation. I want to note the degree of synergy between the teammates, and I do recognize that some of that is out of the player's hands, but I also want to credit players for actively seeking to learn to fit around the teammates they have, and doing that historically well as has happened with Curry & Green is something I see as a major positive.

Next I want to point out to Curry's RS being in '15-16 but PS in '16-17. This is something I've alluded to already on this thread as a think that's being used as a cudgel against Curry, but I think we should be very careful about this.

I understand saying "they're different seasons", but do I think it's a coincidence that Curry's marks in the respective RS & PS came only a year apart? No, can't say I do.

To me Steph's 2015 to 2018 regular seasons look more like each other than his 2017 playoffs do to any of the other playoffs.

Okay. Do you understand what I was pointing to when I said what I said? Can you theorize as to where in our respective processes we diverge?

Fine to knock a season in this context for health, but from a perspective of trying to dismiss the '16-17 PS run as something unearned because of Durant's presence, I think we know we wouldn't be so prone to do this if Curry had simply been able to be enough better to have the team end up 4-3 in the finals rather than 3-4, and while the bragging rights are huge there in the context of putting LeBron over Steph... why exactly should we put Step below other guys for losing to LeBron, when we expect everyone would lose to LeBron?

This is what I mean by kids gloves. Yes, LeBron is amazing and maybe had his best series ever, or maybe anybody's best series ever. But the warriors lost because steph was real bad. Full stop. The scoring champion averaged 22 ppg and had more turnovers than assists for the series and went 6/19 in a nail biter game 7 against a team with no elite perimeter defenders that regularly had 35 year old richard jefferson and kevin love playing forward minutes.

It doesn't get talked about like 2011 LeBron, but it's arguably even more of a thrown away title. Like the Heat needed to win 2 more games, the warriors just needed to score more than 0 points in the final 5 minutes of the series.

So, let me acknowledge that the mere fact that it was LeBron's team that Steph's team lost to doesn't mean Steph should be above criticism in non-LeBron debates.

But let me also point out that you're zeroing in on what you perceive as Steph's greatest failure to evaluate him as a player, and this is something that I try to avoid, just as I try to avoid getting overly infatuated with runs that seem "perfect" as if the player's lack of perfection in any other years is about that player getting a tier worse.

So in the '15-16 Finals, Steph has worse numbers than in the '14-15 Finals. Does that mean he generally got worse at basketball from one year to the next? I think all would say "No", and would also note that Curry clearly improved from '14-15 to '15-16 looking at the entirety of the season. So what does it mean that Curry was worse in the '15-16 series?

From a pure only-looking-at-this-season perspective, easy enough to just aggregate over the year with a heavy weight toward the playoffs and come to a conclusion, and the reasons for why a guy was inconsistent are just ignored.

But if all of this can get explained from a perspective of Curry being banged up over the course of a season where they killed themselves to get to 73-9, then why are indexing so hard on that particular moment?

Not saying we should ignore the moment either, but when we see a close to unprecedented level of RS & PS success from a team over a half-decade run, why would we want to try to evaluate the peaks of the players by looking at their nadir?

Now, I do understand having less confidence in Curry's peak because of inconsistencies and I'm not really looking to imply he "has to be at least" any particular rank, but I would just urge caution when seeking to approach this project by finding nits to pick on particular guys because

a) I believe that tends to lead to fixation on narrative and notions of perfection/imperfection rather than basketball.

b) I think people tend to apply this lens inconsistently. Some players end up getting anchored to positive moments while others get anchored to negative moments, and while this doesn't necessarily start out through any personal emotional bias, it tends to go in that direction when we get into intense debates on the subject over time.

[One_and_Done]

Why would those stats you named be “more helpful”? It is certainly the case that those are “unreliable” ways of distinguishing between great players, because they miss absolute boatloads of context.

It's not that looking at TS% tells you how good a player is, it's more that it is more reliable at showing what it is intended to measure accurately.

I'm not rating guys on things like TS%, it's just one piece of context. Every means of assessing players needs alot of context. I'm more interested in the argument presented, and the evidence underlying it, than I am in people hurling a wall of unreliable math at me.

[Caneman786]

The Question of Curry's Superior Impact, and the importance of individual Offense > Defense

...

Hey DraymondGold,

Great post. I have a question.

Team Overall SRS
- 17 Curry’s Warriors 16.15 >> 01 Shaq’s Lakers 12.20 > 03 Duncan’s Spurs 9.01 > 23 Jokic’s Nuggets +8.50 > 04 Garnett’s Timberwolves +5 (decimal values weren’t given)
So while Curry obviously had better teammates, there is significantly more value to go around for the peak-Curry Warriors.

What is your source for these SRS numbers? I looked on basketball-reference and it says the 2017 Warriors had a +11.56 SRS.

[70sFan]

Hmm.

My early thoughts for top-4 would be 09 Lebron, 00 Shaq, 23 Jokic and some version of Duncan, 02 or 03. I need a minute to think and I'm mid-set, so I don't have time for a thoughtful post, but I'll drop something over the weekend.

Will you make the voting post today? I collected the votes and made the calculations, but I can wait another few hours.

[70sFan]

Voting results

Votes:

Spoiler:

Code: Select all

    "Cavsfansince84": ["LeBron James", "Tim Duncan", "Nikola Jokic", "Shaquille O'Neal"],
    "One_and_Done": ["LeBron James", "Tim Duncan", "Shaquille O'Neal", "Kawhi Leonard"],
    "trelos6": ["LeBron James", "Stephen Curry", "Nikola Jokic", "Tim Duncan"],
    "babyjax13": ["LeBron James", "Nikola Jokic", "Stephen Curry", "Tim Duncan"],
    "TrueLAfan": ["LeBron James", "Tim Duncan", "Shaquille O'Neal", "Nikola Jokic"],
    "Joao Saraiva": ["LeBron James", "Nikola Jokic", "Shaquille O'Neal", "Tim Duncan"],
    "homecourtloss": ["LeBron James", "Tim Duncan", "Kevin Garnett", "Stephen Curry"],
    "-Luke-": ["LeBron James", "Tim Duncan", "Nikola Jokic", "Shaquille O'Neal"],
    "lessthanjake": ["Nikola Jokic", "LeBron James", "Tim Duncan", "Stephen Curry"],
    "LA Bird": ["LeBron James", "Tim Duncan", "Stephen Curry", "Shaquille O'Neal"],
    "TheGOATRises007": ["LeBron James", "Nikola Jokic", "Tim Duncan", "Stephen Curry"],
    "iggymcfrack": ["LeBron James", "Nikola Jokic", "Kevin Garnett", "Tim Duncan"],
    "Djoker": ["LeBron James", "Tim Duncan", "Shaquille O'Neal", "Stephen Curry"],
    "70sFan": ["LeBron James", "Tim Duncan", "Nikola Jokic", "Shaquille O'Neal"],
    "DraymondGold": ["LeBron James", "Stephen Curry", "Nikola Jokic", "Kevin Garnett"],
    "Doctor MJ": ["LeBron James", "Stephen Curry", "Nikola Jokic", "Kevin Garnett"],
    "eminence": ["Tim Duncan", "LeBron James", "Stephen Curry", "Shaquille O'Neal"],

Number of voters: 17 (+2 unofficial ones from new members)

Top 3 lists by Kemeny score:

Spoiler:

1. LeBron James > Tim Duncan > Nikola Jokic > Stephen Curry > Shaquille O'Neal > Kevin Garnett > Kawhi Leonard : Kemeny score = 308
2. LeBron James > Nikola Jokic > Tim Duncan > Stephen Curry > Shaquille O'Neal > Kevin Garnett > Kawhi Leonard : Kemeny score = 307
3. LeBron James > Tim Duncan > Stephen Curry > Nikola Jokic > Shaquille O'Neal > Kevin Garnett > Kawhi Leonard : Kemeny score = 307

Top 25 peaks of the 2001-25: #1-#2 Spots:

#1. 2008/09 LeBron James



#2. 2002/03 Tim Duncan

[DraymondGold]

The Question of Curry's Superior Impact, and the importance of individual Offense > Defense

...

Hey DraymondGold,

Great post. I have a question.

Team Overall SRS
- 17 Curry’s Warriors 16.15 >> 01 Shaq’s Lakers 12.20 > 03 Duncan’s Spurs 9.01 > 23 Jokic’s Nuggets +8.50 > 04 Garnett’s Timberwolves +5 (decimal values weren’t given)
So while Curry obviously had better teammates, there is significantly more value to go around for the peak-Curry Warriors.

What is your source for these SRS numbers? I looked on basketball-reference and it says the 2017 Warriors had a +11.56 SRS.

Thanks! The source is here: https://forums.realgm.com/boards/viewtopic.php?t=2012241 . Highly recommend giving the articles on some of your favorite teams a read-through. Really well written stuff by Sansterre. No stat's perfect, but this one does a fairly good job at ranking teams (with occasional exceptions like any stat).

As to why it's different, basketball-reference provides regular season only SRS. Overall SRS is both regular season and playoffs (with 7x extra weighting to the playoff games, calculated each round of the playoffs, per the Sansterre article I linked).

[tsherkin]

Hmm.

My early thoughts for top-4 would be 09 Lebron, 00 Shaq, 23 Jokic and some version of Duncan, 02 or 03. I need a minute to think and I'm mid-set, so I don't have time for a thoughtful post, but I'll drop something over the weekend.

Will you make the voting post today? I collected the votes and made the calculations, but I can wait another few hours.

Sorry, I didn't get to it in time. Will try for the next slot. It's been my first week back to work, so the timing was a little less than ideal, heh.

[lessthanjake]

[
On this note, I'm going to take data from a source I'm really cautious about using because a) playoff RAPM is fraught with issues, b) single post-season RAPM data all the more so, and c) I don't know anything about who the source actually is, but it's what we have and it's relevant here:

Single-season RS & PS RAPM studies on Github

FYI, we do actually know who did that GitHub RAPM, and even have an overview of the code: https://basketball-analytics.gitlab.io/rapm-data/page/about/

So, I'd say what bugs you is a perception of goalpost moving by people who claim to be rational but continue to side with the same guys no matter what. I get that and I'm not asserting that I have any kind of immunity to that. I try to be rational, but I'm also sure that I have emotions warping my reasoning just like everyone else, and so part of trying to be rational for me is trying to identify the existing knots of inconsistency in what I think I know.

So I'm game to discuss this stuff because I want to get better, but it's also only going to be fruitful if I'm talking to someone who really is talking to me rather than a larger group that I'm taken to be part of.

And one of the things I really need to emphasize is that there are concerns about using RAPM in the playoffs as if that data will give you something apples-to-apples with what we think of it as meaning in the regular season. Playing a small number of games against a far smaller number of opponents just means there's going to be a huge amount of noise. I'm happy to discuss PS RAPM, but from a perspective of "You value RAPM so much in the RS but ignore the PS, what's up with that?", well, noise is what's up.

To give another example on this that I've said for a long-time: I really like using regression-based models for the modern NBA, but that doesn't mean I'm itching to get that data for the deep past because I don't actually think it would give us meaningful information for guys like Wilt & Russell who played basically all meaningful minutes, and thus their Off sample would perpetually be noisy even after years and years of sample.

So I'll probably always talk more about RS RAPM data than PS RAPM data even as I acknowledge that legacy is determined primarily from the post-season. And this definitely leaves room for the possibility of a player being much less or much more effective in the playoffs relative to rivals than RS RAPM would indicate, and once again drives us back to the final step of analysis not being following stats but explaining how we come to differ in assessment from them.

Yeah, you make some really good points here, which I’ll just take some time to second here.

I think it’s important to understand the limitations and issues with different stats and to not blindly consider one type of stat to always be the best. In some contexts, the limitations of a particular type of data will be crippling, such that it’s not very useful, while in other contexts that exact same type of data might be great.

And that’s in essence why it seems very reasonable to treat regular season RAPM and playoff RAPM differently. RAPM is a great type of stat, but it requires a large sample size in order to not be super noisy. Which makes it great for multi-season analysis, not very good for single-season or multi-playoff analysis, and essentially useless for single-playoff analysis.

Box stats are worse than RAPM for multi-season analysis, because they are not directly getting at impact, even if we are talking about box metrics that are designed to correlate with RAPM. But in smaller samples, they’re not nearly as noisy as RAPM. So you likely get a better sense of a player’s impact over small samples by looking at box data rather than RAPM (or potentially by looking at box-impact hybrids). It’s still not great—after all, the fact that there’s a small sample doesn’t erase the problems with box stats. But it’s very likely better than RAPM in that context.

Unfortunately, the upshot is basically that we don’t have a *great* option in terms of data for the portion of the year that is most important (i.e. the playoffs). We just have a least bad option, basically. And that also simply makes data not as important to the analysis of playoffs as it is when looking at larger samples IMO. Which is somewhat okay, I think, because it’s actually possible to watch all of various players’ playoff games, whereas watching everyone’s regular season games over years is impractical. Data is less helpful in the playoffs but is also less necessary.

Of course, regular season vs. playoffs isn’t the only contextual difference that might affect how useful different types of stats are. You mention a good point about whether RAPM data would even be helpful if it went back to much earlier years of the NBA. I’d not thought about this before, but it’s an interesting point. You’re absolutely right that guys like Russell and Wilt played so many minutes that the “off” sample would be extremely small and consumed by garbage time. This would probably make RAPM a bad measure to assess them, even if it were multi-season data that we might regard as a good type of data in almost all situations.

Basically, I would push back on anyone acting like preferring different types of data in different contexts must be inconsistent. Different types of data are genuinely significantly more reliable in some situations than others, and that really *should* affect what types of data we focus most on in different contexts. That’s not inconsistent, but rather rationally applying an understanding of the limitations of different data types.

[Caneman786]

yeah, but if we say those steph underperformances didn't happen, then it's just all good games from him. i mean, it doesn't feel like a coincidence that steph fell off in the 2015 playoffs, and especially the finals, with a lot of pressure. then fell off a cliff in 2016 with a lot of pressure. then fell off a nearly equally high cliff in 2018 with a lot of pressure. and then fell off in 2019 at about a 2015 level with a lot of pressure. and then soared and had his best playoffs ever with no pressure and a cruise title (apparently it was only KD who had it easy). but apparently it actually was just a huge coincidence.

Hey f4p -- we've been round this before, so I'll keep it brief, but I think this misses a lot.

so i appreciate the engagement but you're going to have to miss me with a lot of this.

-2015: This was his first deep playoff run, where he faced Tony Allen (who many consider on the short list of best point guard defenders ever, with Kobe and Durant both saying he was one of the best to ever guard them) and the first time he faced the 'Curry rules' against the Cavs. Many players struggle more in their first deep playoff run, learn a ton, then come back better.

it was still a fairly significant drop. 307th out of 416 playoff runs in my resiliency spreadsheet. and right after tony allen, he faced 37 year old jason terry backed up by 37 year old pablo prigioni. and tore them up. that feels like that balances out. and "curry rules"? he was getting guarded by delly. part of the issue (and this comes up when people pretend steph is doubled so much) is that the warriors won't just iso curry. they keep telling us the offense can't work any other way and it's better. but like, curry could have just iso'd delly all day and destroyed him (or at least should have destroyed him). instead, either because they don't think curry can do that or whatever, they always bring other defenders to the ball for screens, either on-ball screens or off ball screens because curry can't physically separate from defenders by himself. which leads to "curry rules" because now the defense has to choose what to do with the extra defenders. but curry could just do what lebron or harden or luka would do and massacre someone like delly and there would be no curry rules to worry about. i've rarely if ever seen a team just run an extra defender at curry when he's iso'ing. even usually letting him go at people like kevin love one on one being preferred to an open court double. .....
........
........

Great post f4p!

I'm pretty curious about the resiliency metric you speak of. I get the image that it shows how well a player's great performance in a regular season transferred to the playoffs.

Can you link the place you talk about it or say where I can find these numbers? I'd like to read more about this and see how a bunch of seasons rank, including LeBron 2007–18, Paul 2013–23, Paul '08–'09, Butler 2020–23, Jordan '91–'98, and many, many more.

Thanks,

[Caneman786]

Should be fun to see how this list turns out.
'03 Duncan was truly one of the best players ever. People forget what a great athlete he was. He was agile and powerful for a man of his size. He had 111 dunks this year which is 2nd only to his rookie season of 149. He was 3rd in the league in Restricted Area FGs made trailing only Shaq and Gary Payton. He was fantastic on both ends of the court in the regular season. But the backbone of his argument for #2 here is how unbelievable his playoff run was.

You aren't talking about the 6' 4", 180 pound point guard named Gary Payton, are you?

That's insane!