John Stockton is underrated here

[penbeast0]

AH, now it makes sense, thank you.

[Special_Puppy]

Is Stockton the player with the biggest discrepancy between his "advanced stats" and reputation?

[Doctor MJ]

A lot here I don’t disagree with, so I’ll just respond on a few things:

Re: better...how? So clearly the big get for the Jazz was Jeff Hornacek, and we actually do seem to have indicators possibly suggesting that Stockton was more impactful for a Jazz lineup than Malone before & after Hornacek, but Malone was the more impactful player when Hornacek was there.

It's obviously oversimplistic to say "It's all because of Hornacek's presence or absence", but I think what we can acknowledge is that a player's irreplaceability from within roster very much depends on who is on said roster, and such changes might have swung and swung back the Jazz hierarchy of irreplaceability.

I do think it is plausible that Stockton’s impact may have gone down when they brought on Hornacek, given that Hornacek was a capable playmaker. I don’t know that we truly have enough data to conclude in the first instance that Stockton’s impact *did* go down in those years, but you’ve brought up some circumstantial evidence to that effect, and I don’t think it’d be all that surprising an outcome.

That said, I don’t think we can really conclude Stockton’s impact tumbled a large amount—after all, if it did then I don’t think we’d expect the Jazz to have improved as much as they did when Hornacek arrived. The Jazz became *very* good! So this is probably more a point at the margins. And I’ll note that, even in 1997-1998, a pretty old Stockton on the Jazz with Hornacek was putting up great RAPM near the top of the league. So if that was his impact nadir, I’d say that’s really good!

I'll start with positives back at you ltj. Indeed, we're really going back & forth here, but it's partly because there's so much we both agree upon that we can even do this. And so, here's to the agreements as I too focus on the disagreements:

Re: if Stockton's impact tumbled with Hornacek, then the Jazz wouldn't have improved so much, would they?

Eh, if we use the term "irreplaceability" I think things are clearer here. If another solid playmaking guard means that you don't actually need Stockton out there to run a great offense around Malone, then clearly Stockton's going to be less irreplaceable, and they'll thus improve in Stockton's off minutes compared to what they were before and become better over all.

If we then substitute "impact" back in for irreplaceability then yeah, we should expect Stockton's impact to go down as the team improves in such a circumstance.

More generally: We tend to anchor our analysis on Stockton as an all-time great level passer and thus expect that he was having amazing passing impact next to Malone as a matter of course, but I think much of the premise of that era of Jazz offense is that the scheme can do much of the heavy lifting when it comes to actually getting the ball to Malone for the offense's chosen primary attack, and it would seem that it worked basically just like it was designed during the years the Jazz were contending.

I'm always going to be sympathetic to the idea that we don't really know what Stockton's ceiling was as a floor general in part because of the way Sloan chose to focus his offense more and more around Malone rather than Stockton, but the fact that it worked means that during the Jazz hey day, Stockton wasn't actually operating as an outlier offensive floor general in the sense we tend to think of it - the head of the snake who the team is hoping will make as many of the decisions out there as possible. Stockton's job became to get Malone the ball, and I'd say unsurprisingly, that meant he wasn't uniquely qualified to do that job.

Re: disagreement about existence of spike in late career Stockton RAPM data.

Well if we look at nbarapm's 2 year RAPM, Stockton goes from a 4.5 in 1998 to a 7.6 in 2001. There are some other related stats, but that's what I'm pointing to.

Is it just noise? I mean, if it is, then why would we use an extreme outlier value for an old man during a time of team irrelevance over the more age-appropriate data from the earlier span? If it's a meaningful improvement, we should feel a need to explain it. If it's noise, then we should not take it very seriously. Either way, we shouldn't be grabbing on to the weird outlier as if it represents the natural trendline for Stockton's career.

I think my response to this would be that I’m not really “us[ing] an extreme outlier value” for Stockton. For sake of completeness, I did list his rank in two-year RAPM in my OP, but I’ve focused more on longer-term RAPM than that, and definitely haven’t made arguments centered on the impact data from those specific years by themselves. Once we are looking at 4-year or 5-year RAPM, I don’t really think that probably-noisily-good on-off in two years is much of a problem for the data, since at that point we have enough data that the whole thing is not very noisy anymore.

In other words, I’d say Stockton being #1 in 2000-2001 two-year RAPM with an abnormally high value probably is in part a result of noise going in his favor in those years. But it’s a lot harder to argue that noise is a big factor in Stockton being #2 in 1997-2001 five-year RAPM. I don’t think the arguments for Stockton rely on noise, because we actually have enough great data for him to not be relying on particularly noisy samples. In contrast, I think an argument against Stockton that centers on changes in his impact in different small samples has a good chance of being an argument based around noise.

Re: acknowledging a peak in 2000-01 2-year having something like noise in it, but hard to argue that noise is a big factor in Stockton being #2 in 1997-2001 5-year.

Let me back away from the word "noise" for a second. Stockton is only #2 in that 5-year RAPM because he's #1 in the 2-year. Lower the data in the 2-year, apply to the 5-year, he ranks lower, right?

If your feeling is that a drop to 3-5 from 2 is not that big of a deal, I get that, but to the extent that noise is responsible for the 2-year, that noise is also bolstering the 5-year - albeit to a lesser degree.

Okay so now I'm making use of thebasketballdatabase.com as you suggested, and I'm looking at Stockton's 1-year RAPM where we see him hover between 3.65 & 4.23 before it spikes to 6.35 in '00-01. So what I'm going to do for comparison is to look at the best 1, 3 & 5 year numbers they have that end no later than '99-00. (Note this also means that the 5-year numbers will technically be less than 5 years of Stockton's play, because PBP era only extends 4 years before that.)

Stockton
Top 1 Year: Rank 4 (4.23, 2000)
Top 3 year: Rank 3 (4.07, '98-00)
Top 5 year: Rank 3 (3.89, '97-00)

So, from that, we're still seeing him ranked Top 5, which is undoubtedly elite. I've alluded to other concerns about Stockton's numbers in those later year and I don't want to just drop those, but it's good to get an objective analytic foundation on which to dialogue.

Alright, now let me look at what the site says for the top years for the other point guards I've talked about for comparison (Kidd, Nash, Paul).

I'll preface it by saying I sorted by RAPM rank, followed by RAPM score. If anyone wants to discuss further about why I did it that way, lemme know.

Kidd
Top 1 Year: Rank 3 (4.11, 2005)
Top 3 Year: Rank 5 (3.89, '03-05)
Top 5 Year: Rank 6 (3.45, '01-05)

Nash
Top 1 Year: Rank 3 (4.43, 2008)
Top 3 Year: Rank 3 (4.24, 2005-07)
Top 5 Year: Rank 1 (4.35, 2005-09)

Paul
Top 1 Year: Rank 2 (4.47, 2014)
Top 3 Year: Rank 1 (4.34, 2011-13)
Top 5 Year: Rank 2 (5.06, 2014-18)

So just taking on this lens face value, that Stockton data looks better than Kidd's best, and quite close to Nash/Paul, and that's extremely impressive for any star point guard, let alone an old man!

There's certainly more nuance I could get in there, but my mind is going on a tangent so I'll end it with:

The fact Nash & Paul's best runs don't perfectly align across years, and I think that's something really worth digging into whenever we see it. I was expecting to see this with Nash because it's not the first time this has come up, and I think I have really clear insight to share, so I'll take a minute to do that.

Why isn't Nash's RAPM peak year (2008) a year outside the sample that is his best 3-year?

If we look at his 1-year numbers for the 5 year period that encompasses all of the runs listed, here's what we get:

2005: 5
2006: 22
2007: 6
2008: 3
2009: 28

First, so interesting that numbers like that can combine into the #1 5-year number, isn't it? I wouldn't have expected that.

Second, clearly what's going on is that Nash's RAPM stumbled a bit in 2006 & 2009. The '08-09 season is a pretty straight forward thing: In the wake of trading away Marion for old man Shaq, the Suns were experimenting Terry Porter's offensive scheme, and it didn't work as well, which is why they'd end up going back to a more free D'Antoni-type scheme under Gentry and make another run afterward.

'05-06 though is very interesting because it's the year Nash won his 2nd MVP, and did so based on a narrative of doing more with less given the injury of Amar'e Stoudemire and the senseless loss of Joe Johnson. But if Nash's RAPM goes down that year, was he getting too much credit for the resilience of that Suns team?

Quite possibly, but the key objective thing to note is the Suns struggled early in the season figuring out who was working out, and Nash had a negative +/- in 8 of the first 9 games. The team righted the ship abruptly with the most obvious change being Boris Diaw emerging as a major force. This then to say that if you want to think like the voters did, you're thinking not so much about season-long numbers, but how good the Suns looked during the meat of the season.

How impactful do I think Nash was once things clicked for the Suns that year? Well, on peer with what he was doing in those other years with the higher ranks.

Okay, before moving on I'll just note again that Paul's numbers deserve such an analysis too.

[lessthanjake]

Re: acknowledging a peak in 2000-01 2-year having something like noise in it, but hard to argue that noise is a big factor in Stockton being #2 in 1997-2001 5-year.

Let me back away from the word "noise" for a second. Stockton is only #2 in that 5-year RAPM because he's #1 in the 2-year. Lower the data in the 2-year, apply to the 5-year, he ranks lower, right?

If your feeling is that a drop to 3-5 from 2 is not that big of a deal, I get that, but to the extent that noise is responsible for the 2-year, that noise is also bolstering the 5-year - albeit to a lesser degree.

I’ll just limit what I respond to to this (though some of what I say below relates to what you said after this precise passage). I don’t really disagree with the first section of your post, so am not responding to that.

I’d say a few things on this:

1. Lowering the data in the 2-year and applying that decrease to the 5-year assumes that there wasn’t negative variance in the other years of the 5-year. But the fact that 5 years is typically a large enough sample to reduce noise suggests that that’s exactly what we would expect! If we identify years we think there was positive variance and try to adjust the longer-term picture down to account for that, then we’re probably just creating a picture in our brain that gives the player negative variance overall.

2. Even if we accepted the premise, I think what we’d drop Stockton down to would still be really great. A lot of the rest of your post basically goes to that point. You’re basically looking at old Stockton, taking out his best years in the data, and still getting him looking similar to the peak years of the top PGs in the play-by-play era. I think that’s actually pretty incredible!

3. To add to this giving-the-player-negative-variance point, the methodology you’re using with Stockton would be the equivalent of trying to take Nash’s best RAPM data but not including 2007 and 2008, since those are his highest two RAPM years. If we do that, he’s peaking out 4th in three-year RAPM, and 7th in five-year RAPM. I get that your position probably is that Stockton’s best data years aren’t really real and Nash’s are, but I think that’s an inherently biased approach to the data, where the decisions on how to apply the data are presupposing the answer to the question the data is trying to get to (i.e. Nash was better than Stockton so we should count his best impact numbers but not Stockton’s, and when we do that Stockton doesn’t look better than Nash).

[Doctor MJ]

Re: acknowledging a peak in 2000-01 2-year having something like noise in it, but hard to argue that noise is a big factor in Stockton being #2 in 1997-2001 5-year.

Let me back away from the word "noise" for a second. Stockton is only #2 in that 5-year RAPM because he's #1 in the 2-year. Lower the data in the 2-year, apply to the 5-year, he ranks lower, right?

If your feeling is that a drop to 3-5 from 2 is not that big of a deal, I get that, but to the extent that noise is responsible for the 2-year, that noise is also bolstering the 5-year - albeit to a lesser degree.

I’ll just limit what I respond to to this (though some of what I say below relates to what you said after this precise passage). I don’t really disagree with the first section of your post, so am not responding to that.

I’d say a few things on this:

1. Lowering the data in the 2-year and applying that decrease to the 5-year assumes that there wasn’t negative variance in the other years of the 5-year. But the fact that 5 years is typically a large enough sample to reduce noise suggests that that’s exactly what we would expect! If we identify years we think there was positive variance and try to adjust the longer-term picture down to account for that, then we’re probably just creating a picture in our brain that gives the player negative variance overall.

2. Even if we accepted the premise, I think what we’d drop Stockton down to would still be really great. A lot of the rest of your post basically goes to that point. You’re basically looking at old Stockton, taking out his best years in the data, and still getting him looking similar to the peak years of the top PGs in the play-by-play era. I think that’s actually pretty incredible!

3. To add to this giving-the-player-negative-variance point, the methodology you’re using with Stockton would be the equivalent of trying to take Nash’s best RAPM data but not including 2007 and 2008, since those are his highest two RAPM years. If we do that, he’s peaking out 4th in three-year RAPM, and 7th in five-year RAPM. I get that your position probably is that Stockton’s best data years aren’t really real and Nash’s are, but I think that’s an inherently biased approach to the data, where the decisions on how to apply the data are presupposing the answer to the question the data is trying to get to (i.e. Nash was better than Stockton so we should count his best impact numbers but not Stockton’s, and when we do that Stockton doesn’t look better than Nash).

Hmm, so I think the specifics of why I focused on what I did were clearly not well-communicated. That might damp the energy of conversation to the point of goodbyes, but let me try to communicate better on this:

The reason why I've been focused on the spike near the end of Stockton's RAPM numbers is not for reasons I'd want to classify as "noise". I have no particular reason to believe that it was micro-event randomness that led to Stockton looking better there. Such randomness is real, but I don't think that the best instinct when we encounter data deviating from expectation is that it must have been randomness. I think more likely that the continued evolution of the Jazz roster, including aging, shifted that irreplaceability factor of Stockton some in the direction that the RAPM appears to say.

And let's note as you presume my Nash bias is crippling my objectivity, me going into depth as to what was happening year by year with Nash was in part me asking others to do the same explicitly for Paul, and certainly implicitly for Stockton.

I just went through and explained what Nash's numbers have some unexpected ups & downs, and what I've been doing in pointing to that late career Stockton RAPM spike is the need for us to really understand what was going on on the court there.

If that data isn't noise - which is what I find to be most likely - what is it?
If it is noise, then we might as well remove it from the calculation...which is why I did just that with the RAPM on your preferred site.

[sp6r=underrated]

Is Stockton the player with the biggest discrepancy between his "advanced stats" and reputation?

Stockton has a gap between his advanced metrics and his place all-time. By the former he is a very credible player for top-ten all-time while few regards him that way. That said during his career he didn't have the biggest reputation gap to his statistical output. Stockton was selected for the Dream Team and regularly made All-NBA teams. His ranking was lower than his statistical footprint but it wasn't minor.

In real time the player from his generation with the biggest statistical-reputation gap was Reggie Miller who received very little recognition. Oddly fans identified Reggie Miller as a superstar while the basketball press and teams did not. I say oddly because normally iron men, who are their team's best player, get recognized when they have playoffs heroics deep in the playoffs. In retrospect, the fans got this one right.

[lessthanjake]

Re: acknowledging a peak in 2000-01 2-year having something like noise in it, but hard to argue that noise is a big factor in Stockton being #2 in 1997-2001 5-year.

Let me back away from the word "noise" for a second. Stockton is only #2 in that 5-year RAPM because he's #1 in the 2-year. Lower the data in the 2-year, apply to the 5-year, he ranks lower, right?

If your feeling is that a drop to 3-5 from 2 is not that big of a deal, I get that, but to the extent that noise is responsible for the 2-year, that noise is also bolstering the 5-year - albeit to a lesser degree.

I’ll just limit what I respond to to this (though some of what I say below relates to what you said after this precise passage). I don’t really disagree with the first section of your post, so am not responding to that.

I’d say a few things on this:

1. Lowering the data in the 2-year and applying that decrease to the 5-year assumes that there wasn’t negative variance in the other years of the 5-year. But the fact that 5 years is typically a large enough sample to reduce noise suggests that that’s exactly what we would expect! If we identify years we think there was positive variance and try to adjust the longer-term picture down to account for that, then we’re probably just creating a picture in our brain that gives the player negative variance overall.

2. Even if we accepted the premise, I think what we’d drop Stockton down to would still be really great. A lot of the rest of your post basically goes to that point. You’re basically looking at old Stockton, taking out his best years in the data, and still getting him looking similar to the peak years of the top PGs in the play-by-play era. I think that’s actually pretty incredible!

3. To add to this giving-the-player-negative-variance point, the methodology you’re using with Stockton would be the equivalent of trying to take Nash’s best RAPM data but not including 2007 and 2008, since those are his highest two RAPM years. If we do that, he’s peaking out 4th in three-year RAPM, and 7th in five-year RAPM. I get that your position probably is that Stockton’s best data years aren’t really real and Nash’s are, but I think that’s an inherently biased approach to the data, where the decisions on how to apply the data are presupposing the answer to the question the data is trying to get to (i.e. Nash was better than Stockton so we should count his best impact numbers but not Stockton’s, and when we do that Stockton doesn’t look better than Nash).

Hmm, so I think the specifics of why I focused on what I did were clearly not well-communicated. That might damp the energy of conversation to the point of goodbyes, but let me try to communicate better on this:

The reason why I've been focused on the spike near the end of Stockton's RAPM numbers is not for reasons I'd want to classify as "noise". I have no particular reason to believe that it was micro-event randomness that led to Stockton looking better there. Such randomness is real, but I don't think that the best instinct when we encounter data deviating from expectation is that it must have been randomness. I think more likely that the continued evolution of the Jazz roster, including aging, shifted that irreplaceability factor of Stockton some in the direction that the RAPM appears to say.

And let's note as you presume my Nash bias is crippling my objectivity, me going into depth as to what was happening year by year with Nash was in part me asking others to do the same explicitly for Paul, and certainly implicitly for Stockton.

I just went through and explained what Nash's numbers have some unexpected ups & downs, and what I've been doing in pointing to that late career Stockton RAPM spike is the need for us to really understand what was going on on the court there.

If that data isn't noise - which is what I find to be most likely - what is it?
If it is noise, then we might as well remove it from the calculation...which is why I did just that with the RAPM on your preferred site.

I agree that the idea of irreplaceability shifting could be another explanation for Stockton’s impact looking better in certain years. I didn’t mean to suggest noise is the only possible explanation and to just shut the door on anything else. Another thing is that a brief glance at the lineup data I’ve mentioned indicates to me that he wasn’t quite being staggered with the bench units as much in those years as he had before that, so perhaps that helped his impact too (though I don’t think that’s a factor that inherently must go one way or the other in terms of its effect on RAPM, so that’s just a theory). These are all possibilities. But I’d say two things about this:

1. I’m not sure if those explanations would justify removing the better years from the equation for Stockton (or at least focusing on the other years). For instance, let’s say him no longer being staggered as much with bench units helped his RAPM in those years. Does that mean that his RAPM in 2001 was inflated compared to other star players, or that his RAPM in the earlier years was deflated? Arguably it’s the latter, since the Jazz rotation strategy in the late 1990s was actually pretty abnormal. And to the extent that irreplaceability shifting had an effect, was Stockton actually more irreplaceable in 2001 than other superstars are, or was having Hornacek in those earlier years actually cutting down on his irreplaceability more than is the case for most stars? I might lean towards the latter, though it’s hard to really say anything for sure on a non-concrete subject like irreplaceability. Anyways, overall I wouldn’t say the answer to this stuff is clearly that the highest-impact years for Stockton were him being particularly advantaged contextually, rather than potentially no longer being disadvantaged contextually! In which case, if we accept these explanations as being the reason for the shift, it’s not clear to me what we should regard as the “real” value, so to speak.

2. I do also think an always-looming explanation for shifts in on-off or RAPM over small samples is just noise. And that’s especially true when the ON value stays very stable (which it did here). The difference was in the relatively small sample of OFF minutes (though, I will note that old Stockton’s MPG was lower than most stars, so the OFF sample is at least a little bigger, making his RAPM a little less noisy). To the extent the explanation is noise, I don’t think the correct approach is to remove it from the calculation, because we can generally assume that the noise evens out over larger samples, and so if we remove the years we think have positive noise then we should expect we will be left with years with negative noise. I think the correct approach regarding noise is just to look at a big enough sample that noise is unlikely to be a concern. And we do have big enough samples for Stockton to do that (and he looks great still). Of course, if there’s a non-noise explanation such as those discussed above, then “just increase the sample size” isn’t necessarily a cure-all (though it would still help, since the larger sample would include multiple different contexts and so the overall effect of contextual factors is more likely to have evened out).

Regarding having a “Nash bias,” to be clear I was meaning to use the term “biased” in my last post in the statistical sense, not in an impugning-your-motives way. I actually have a significant “Nash bias” myself, and have spent a whole lot of time arguing in his favor on things, particularly his status as an offensive great. To me, he is the GOAT offensive player, and I’ve spilled a lot of ink here and elsewhere going back well over a decade on that. So, even if I thought you had a Nash bias, we’d just be in the same club. It actually brings me no pleasure at all to say I think Stockton was a more impactful player. I like Nash far more. In fact, he may be my favorite basketball player ever—certainly I’ve never rooted for any team as much as I did the Nash Suns, in large part because I felt strongly that I wanted them to succeed so that the NBA would adopt their style of play. I just look at the overall impact data and find it hard not to come to the conclusion that Stockton was probably more impactful—albeit with a healthy degree of uncertainty, as appropriate when assessing the impact of any player that played all or most of their career prior to 1997.

[Sailor Haumea]

Is Stockton the player with the biggest discrepancy between his "advanced stats" and reputation?

There's a fair number of players who suffer this but in the opposite direction - look at some of the great defenders of the 50s and 60s like Nate Thurmond or K. C. Jones. Without steals and blocks data, and no All-Defensive Teams until 1969, these players wound up with few accolades and stats like win shares don't tell the full story.

(This is also why I maintain that Dave DeBusschere is one of the most underrated players.)

[Doctor MJ]

Is Stockton the player with the biggest discrepancy between his "advanced stats" and reputation?

Well that all depends on what stats you're referring to, but the root of the disconnect isn't about anything advanced, but about assists and steals. These were stats during Stockton's prime that everyone paid attention to even more so than today, and yet they didn't lead to Stockton being championed as a top tier player like Malone, so what gives?

Complicated but:

1. Stockton having a slow start to his career didn't help.

2. Stockton looking like he did didn't help and of course contributed to his slow start.

3. The fact that the Jazz were a winning team with the younger Malone starring while the older Stockton was still on the bench didn't help.

4. The fact that the team didn't that big of a leap forward when Stockton became the co-star (44 wins the prior year, 47 that year) and didn't make the leap to elite 60 number until Stockton's 8th year on the roster and his volume stats was decreasing didn't help.

5. The fact that when the Jazz were good enough to make people start asking seriously about an MVP candidate on the team, it seemed pretty clear that that was Malone, didn't help.

[penbeast0]

. . . 6. The fact that he played in Utah, rather than LA or NY.

[One_and_Done]

Is Stockton the player with the biggest discrepancy between his "advanced stats" and reputation?

Nah, there are all manner of odd role players with crazy advanced stats, but for some reason nobody brings them up.