New Engelmann Career RAPM (1997–2024)

[Doctor MJ]

Just a methodology question, how does the regression account for different years?
How is a possession in year 1 LeBron affecting his RAPM in year 21? Is it "assuming" he's the same player or there's some correction?
And is this some aggregated total as if it was one unique interval or like the weighted average of each year's RAPM?

My understanding is that whenever you're talking about a direct RAPM, the player is assumed to be the same in all samples where he shows up.

I believe JE's line "- 'season' is the only adjustment" implies there is some sort of factor between individual player seasons, but I do not have any idea how large of driver it'd be.

Interesting. He should really lay out exactly what he means by that.

[Djoker]

Amazing! Would be nice to see the confidence intervals as well.

I don't see RAPM as some holy grail stat as much as some others do but it's a useful metric. One name I didn't expect to see so high is Paul George. What are everyone's thoughts on his impact?

[Doctor MJ]

John Stockton: why does this guy always come out looking better than Karl Malone in impact data? Could we all be underrating just how great of an offensivve/defensive force Stockton was?

So, let's put in perspective this way. I have a list of players who led their team's in raw +/- that includes Harvey Pollack's data as 76er scorekeeper for the regular season for the entire league from '93-94 on. If I combine that with the more complete data we have for later years, here are the Jazz leaders for each season:

'93-94 Malone
'94-95 Malone
'95-96 Stockton
'96-97 Malone
'97-98 Malone
'98-99 Malone
'99-00 Stockton
'00-01 Stockton
<other guys lead from that point onward>

I think what that shows us is that Stockton comes off looking more impressive late each their careers, but if you're just focused on the years where the Jazz were actually contenders (ends with '98-99), Malone has the distinct advantage.

This would suggest that while Malone was in prime he was rightfully seen as the more valuable player, but his impact did not age as gracefully after that as Stockton did. Why might this make sense? Well, it's kinda something we tend to see with volume scorers who continue to be 1st options post-prime. The way to stay impactful as you age is to take on a smaller role that doesn't strain your limitations, and that's basically what the Jazz did with Stockton but not Malone.

Last note: Because we only have the data to do true RAPM back to '96-97, RAPM studies actually end up looking very similar to what we used to see back before we got '90s data. There was a time when the '00s data was all we had, and Stockton looked clearly superior to Malone, raising the question of whether he was the more valuable player all along.

This then led us to guess as to what we might see in earlier years. If we could get data back to the Jazz contending years - like we largely do now - and Stockton still had the advantage, then contemporary viewers would essentially be proven wrong in their conclusion that it was Malone who should get the MVP consideration. But many of us expected that Malone in prime would look considerably better than he did in later years for the reasons I've given above...and I'd say that's how it has played out.

Now, we're still trying to get more data from earlier in their career which was the time when Stockton was really putting up his biggest box score numbers, so we'll see how that ends up looking. Thing is though, that wasn't the best era for the Jazz, so even if we end up concluding that Stockton was more valuable in those years, Malone would still have the advantage when talking about the contending Jazz years.

[Doctor MJ]

Amazing! Would be nice to see the confidence intervals as well.

I don't see RAPM as some holy grail stat as much as some others do but it's a useful metric. One name I didn't expect to see so high is Paul George. What are everyone's thoughts on his impact?

Well it's just plain good, but to try to speak a little bit more about what's unusual for him, it's those early years in Indiana. Young players don't typically look great by +/- data but George did. I do think he was the MVP of that team and deserves a lot of credit, but I also think it became clear that they were catching lightning in a bottle that wouldn't hold up in the long term as a team.

While that dismissive conclusion is more about guys like Roy Hibbert than George, the phenomenon of it has everything to do with why George decided to force his way out of Indiana despite the franchise not really doing anything wrong. They were still a winning team in the last couple George years, and they would literally improve after George's departure.

This then to say that this is the sort of thing I hold against players that not everyone does. George is a great player...but not someone I'd want as my franchise player. Worse, while later he would embrace being a sidekick, early on he really seemed obsessed with presenting himself as this ultra-clutch scorer which he could never really pull off.

[AEnigma]

Amazing! Would be nice to see the confidence intervals as well.

I don't see RAPM as some holy grail stat as much as some others do but it's a useful metric. One name I didn't expect to see so high is Paul George. What are everyone's thoughts on his impact?

I think there can be some misattribution where his defence is higher than it should be but his offence might be lower than it should be. All the same, he is definitely one of the best wing defenders ever. Not “+4” level, but I think he fits naturally in that group of Roberson, TAllen, Iguodala, Artest, Battier, Deng, Lebron, Odom, young Kawhi, etc.*, while being the clear best shooter of the bunch and a significantly more capable first option than all of them save Lebron. That type of archetype is always going to be valuable.

*(Side-note: Marion’s value surprised me, although I feel generally fine about his prime regardless.)

I would probably characterise him as an improved Paul Pierce, who similarly has always fared well in long-term samples which if taken at face value could lead people to conclude he was better than a guy like Kobe. Both of them are functional #1s who are ideal #2s — with Paul George especially coming across as a lab-designed #2.

[AEnigma]

George is a great player...but not someone I'd want as my franchise player. Worse, while later he would embrace being a sidekick, early on he really seemed obsessed with presenting himself as this ultra-clutch scorer which he could never really pull off.

I think he was clutch outside of a string of four consecutive playoff series where he very much was not. And four series is going to stick in people’s minds, so that killed the claim, but 2014/16/17/[no Kawhi 2021] pretty much offsets that “not clutch” stretch.

[DraymondGold]

Hey Doc, thanks for the detailed response!

Some Random Observations:
-Jokic #1 makes some sense. Hyper-valuable player with strong peak, #1 in offense, somewhat surprisingly high defense, and no post-peak seasons to drag his average down.

-LeBron #2 also makes sense given how strong his impact is. GOAT-tier stuff from him considering how long he's been doing it. Offensively, he's just in the Top 5.

So taking this section by section:

-While it's true that we haven't seen the tail end of Jokic's career and this may drag him down, it's important to remember that in general older players fair better than younger players in +/-. The reason why Jokic looks so good here is because he was amazingly impactful right from the jump to a degree we haven't seen from any other player in the +/- era.

-LeBron is certainly the gold standard.

I'm interested by your Jokic comment. When you say he's impactful right from the jump to a degree we haven't seen from any other +/- era player, are you saying Jokic was the most valuable rookie since 1997? Or the best rookie since 1997? Or the most valuable first few years, or something else?

Off the top of my head, Duncan would seem like a viable candidate for having a better rookie season than Jokic, or at least he certainly has that reputation. In a cursory statistical glance, Duncan had higher on/off (13.1 > 9.5) and a higher on (8.0 > 2.3), though these would be helped in no small part by sharing a lot of his rotations with David Robinson, who I'd argue was the better player at the time, or at least the more valuable in his costar offensive role come playoffs. Duncan also has the higher BPM (basketball reference, which is far from the best box stat, but still) and a significant minute advantage over Jokic (in a different era, but still). Relative to era, I like Duncan's playoff resilience a little more too -- Jokic's defense was very troubling early on, and still can be in certain playoff matchups, though he's done a fantastic job becoming more versatile and active defensively, and he definitely has some major defensive strengths too (e.g. rebounding).

I'm definitely amenable to the idea that Jokic came out of the gate quite strong, and was underrated, perhaps even among the best +/- rookies. But I'm little hesitant to rank him as being impactful to a degree we haven't seen from any other +/- era rookie. Let me know if I'm missing something!

Re: older players vs younger, I suppose I'm open to the idea older players might more frequently better in plus minus data than younger players than the reverse, but I'm hesitant to say players +/- peak is when they're older compared to, well, their peak Let's say Jokic's impact peak is 2022–today. On/off data certainly things so (he goes from never cracking +12 on/off to +16.4, +21.0, and +18.0 at this point in the season). At the moment, this 2.5 year peak covers ~30% of his total career games, including playoffs. I would bet his +/- and RAPM might still be good when he' older, perhaps a little better than when he was younger. But I might be a bit surprised if he continued to show his current peak level impact metrics (e.g. ~ +18.5 on/off) when he was older (particularly if age starts to make that lateral mobility on defense more of an issue again).
And if the metrics do drop off slightly, his overall career RAPM might look a touch worse, e.g. dropping behind LeBron and maybe some others, once his peak represents closer to ~15–20% of his career games rather than 30% of his career games. Thoughts?

Offensive highlights:
-Curry 8th overall and 2nd offensively (just behind Jokic) is also pretty impressive, consistent with him being GOAT level offensively considering how many more non-peak years he has than Jokic.
-Chris Paul is 3rd overall and 7th offensively, continues to be loved by impact metrics.
-James Harden 4th offensively and 21st overall, solid.
-KAT at 6th offensively is a bit surprising.
-Durant 8th offensively but 16th overall, and Dirk just behind at 9th offensively and 17th overall.
-Nash at #14th offensively and #47th overall, seems a bit low compared to his reputation here.

-Paul has a similar thing going with Jokic where they excel at adding impact to any given possession with smart decisions.

-KAT I think is likely to be now, and continue to be, on a downward trajectory in ORAPM and probably overall RAPM too.

- Main thing with Nash is that he wasn't used maximally for much of his career. His offensive RAPM in the Phoenix years is outlier.
- Curry, Durant, Harden, as you say, solid.

Indeed, there is a similarity re: Paul and Jokic! With Paul, it seems like the data is high on his defense specifically compared to his mean evaluation here (~36th best defense in this career RAPM database). One of the usual knocks against him is playoff performance, but I wonder how much of that comes from injury.

Here, it's interesting to consider how much playing through injury at a worse level vs missing games entirely effect things. If you play through injury (but not playing so poorly that you're a negative), you're probably adding more value than if you miss the game entirely, but playing through injury would drop your average in a database like this. Not sure from memory how often Paul played injured vs was out compared to other injury-prone players, but if he did sit out more rather than playing injured, that might boost his RAPM slightly.

Agreed about KAT.

For Nash, he definitely looks like one of the best in his single-season Phoenix-specific years, and in some RAPM samples his peak looks outlier level / better than all others, though it's worth mentioning there are some versions of RAPM where Nash doesn't look clearly better than the competition (e.g. Goldstein scaled offensive RAPM: Shaq has the top season, Nash has 3 of the top 8, LeBron and Curry both have 2 of the top 8 too and all are pretty close together; in this RAPM https://www.cryptbeam.com/rapm/, you get similar top seasons in a slightly different order, and again Nash looks right near the top but doesn't really get separation over the other top seasons).

Defensive highlights:
-Garnett's defense is #1, enough to boost him to #4 overall. Impact metrics alway love him, but seeing his defense that far above the crowd even with his non-prime years is a bit new for me. His offense is proportionally lower at #173, keeping him from looking best overall in this new stat.
-Old Mutombo is #2 defensively, with some separation after. I wonder if I've been underrating his defense.
-Caruso a bit surprising at #3 defensively, and Paul George at number 10 defensively... the top defenders are obviously dominated by big men, but a few perimeter players do sneak in. Andre Roberson, Tony Allen, and Iguodala are some of the other top perimeter defenders, though they're a tier down between #20–#30. Some of these players are lower-minute defensive specialists, but Paul George specifically is pretty high minutes player. I wonder if his defensive reputation compared to the other recent defensive wings underrates him.
-Draymond Green (5th overall, 5th defensively), continues to be loved by impact metrics.
-Duncan is 6th defensively and 9th overall. Great stuff. A bit below Garnett (as often is the case in impact metrics), but unlike some of the guys I mention below, both Garnett and Duncan don't seem to have their career value cratered that much by their post-prime years (note for DocMJ: Manu is 11th overall, so he's right behind Duncan, and supports your argument that he's underrated)
-Gobert, Ben Wallace, and old Robinson are all also in the top 15, as expected

-The thing about Garnett and "non-prime years" that's so scarry is that these is his DRAPM without him being fully utilized ever in his prime. Had the NBA removed illegal defense sooner and a team decided to play Thibs-style from the beginning, he would tower over his contemporaries all the more. Truly a player ahead of his time that the basketball world didn't know what to do with despite recognizing his revolutionary nature.

-Mutombo warrants the shout out, but I feel a need to also shoutout his fellow Hoya Zo who ranks nearly as high here. It's crazy that these guys were on a college team together.

-With Caruso the thing that drives me nuts as an analyst is us not seeing him for more minutes. Is he incapable of playing more with this impact...or not?

-George and the other non-bigs. Yeah, I do think we tend to overstate how big the gap is between defensive bigs and mediums in the 3-point age. Obviously you'd always love your arms to stretch to be as long as a bigs, but team defenses need serious horizontal game in the modern NBA.

-Yup, Draymond's a stud.
-Duncan looks great for a non-Garnett!

-I appreciate the Manu shout out. I do want to emphasize that the thing I was always bolder on was not the idea that Ginobili was more valuable than Duncan...but that his offense was more valuable than Duncan's. This might seem like a petty thing to emphasize, but it's significant because of the way people tend to slot people in as #1's, #2's, etc. No, it's not a given that the Spurs would have been better had they based their offense around Ginoblii more than Duncan, we simply know that Ginobili was more valuable on offense than Duncan in an offense catering to Duncan. And that's no small thing.

-Gobert, Ben, Robinson, yup, solid.

For Garnett, do you have any thoughts on what specific differences there were under Thibs/rule changes that enabled Garnett to have higher impact performance than Flip Saunders? Obviously the better defensive supporting cast could help, and allowing zone defenses might help a strong rim protector with great versatility, horizontal foot speed, communication, etc.

Didn't mean to mis-represent your Manu stance, it was meant more as a shoutout. But yes, that makes sense, and indeed this RAPM data actually supports your argument! Manu has the greater offensive value than Duncan in this dataset (+4.8 or 19th, vs +2.6 or 139th), while Duncan is the overall more valuable player in this dataset (+7.7 or 9th, vs +7.5 or 11th).

-Some surprisingly lower ranks:
-Kawhi Leonard at #19 overall and #22 offensively is still pretty good, but a defensive rank of 173 is way lower than his reputation.
-Giannis Antetokounmpo at 38th overall is definitely lower than expected, particularly since his impact metrics in 19/20 were so good and we're also mid peak/prime for him too. Both his defense and offense are out of the top 50.
-Anthony Davis and Luka Doncic at 88th and 95th overall are also a bit low. Doncic's reputation of having worse plus/minus numbers continues.
-Dwyane Wade and Kobe Bryant are out of the top 100 overall. Wow! Kobe's 22nd offensively, right in line with some of the other Top 20 overall players, Wade's a bit worse at 42nd offensively. It's their defense that's disappointing... Wade's neutral and Kobe's a negative defender. One wonders how much non-prime years are lowering their ranking here. Kobe's reputation as having slightly worse impact metrics continues. For the Kobe fans: Kobe has an unusual number of non-prime games in his career (e.g. 266 RS games pre-age 22 when Wade was drafted, 325 RS games at age 32+ in 2011–2016, 50 + 22 PS games in those age ranges, for 663 total. For comparison, Wade has 0 games pre-age 22, 389 games after age 32+ in 2014–2019 RS, 45 PS games in those age ranges, for 434 total. Based on this, and considering their overall rank is right near each other, there is an argument that Kobe might end up looking better than Wade in say ~10 year prime RAPM, though neither are probably valuable enough in RAPM to crack the top tier based on how low their total career is).

-The way I tend to think about Kawhi, is that he's an extreme example of an individual sportsman playing a team game. Kawhi's combination of iso scoring and man defense might be the greatest we've ever seen, but when it comes to the team dynamics, he's limited, and surprisingly, this might actually be a bigger limitation on defense than on offense. Because on offense, his individual threat with the ball warps the defense to make team play easier. By contrast, since the offense dictates play, they can effectively quarantine the dude.

-Yeah, Giannis is weird in this family of stats. I don't know if I've ever seen a player become THE dominant player in the league both by consensus and +/- stats, and then really just lose the latter while not becoming a fundamentally worse player, and in fact further rounding out his game. Holistically, I tend to cut Giannis slack here because he did enough to prove his capabilities, and I don't think he's failed his team by changing his approach, but I do tend to get in fights when it comes to MVP debates nowadays when people just want to ignore the fact the data isn't what it used to be.

-AD's interesting. I think the fundamental issue here is that the way AD really stands out is the way he seems to be a match up nightmare to so many teams come playoff time. You can argue that's a fluke, but I don't actually think it is. I think there's something about playing in these series that lets AD really focus his talents.

-Doncic, honestly, this is going to sound damning with feint praise, but making the Top 100 on this list, and doing so having to rely on your own shoulders as the foundation for your team's context, is a pretty significant accomplishment by general standards. Doing so while still being under 25 in a sport where heady players can peak at 30+? All the more so.

-Wade & Kobe. It's clear that the nastiest thing haunting these guys is their non-prime, and so whatever happened in that time, their prime performance would definitely be higher. It's worth noting how much their non-prime hurts them compared to someone like Duncan, but we should acknowledge that's a phenomenon distinct from the level of prime.

Interesting stuff! I like a lot of these takes.

Older players: These are the guys with a major portion of their career missing (pre-1997).
-Jordan 14th overall, and 12th offensively is GOAT-tier stuff, considering ~40% of his sample comes from his Wizards years at age 38–39, and there were nagging injuries in the 1998 regular season. His defense is 159, which is definitely isn't high, but at the same time it's right around the level of Marcus Smart, Jaren Jackson Jr., Kawhi, Jarrett Allen, so not exactly terrible.
-Shaq 13 overall is a touch town from the other modern bigs in Duncan/Garnett, but we're missing 93–96 which includes several prime years. His offense at #23 is strong but not quite as high as expected, but his defense is better than his reputation at 72.
-Old Stockton (12 overall) >> Old Malone (224 overall)
-Old Ewing and old Hakeem are disappointing, both out of the Top 400 overall (69th/70th defensively), at least compared to old Mutombo and Mourning and Barkley and Robinson.

-Jordan's +/- data in the 2nd 3-peat really, really impress me because I always felt that those years were far from his very best. I figured it likely that Jordan had massive +/- numbers at his best, but I kinda thought we'd see something lesser in the last couple years there. Nope, resolute in his domination that one. (Sigh, and other things that really held him back in Washington.)

-Shaq's earlier years are a very interesting thing to consider because of teammates Penny & Grant. Their data doesn't look that great here, but they would look a lot different if we were regressing on those same years. I'm not sure if Shaq's numbers here would see a decisive shift upward.

-Re: O(S >>M) True if focused on effective value-add, but I think it's worth noting how much more the team was relying on the M rather than the S.

-Old 1984/1985 Draft guys who played for a Really Long Time look worse than Guys Drafted Later or Retiring Sooner. I'm afraid this is exactly what I'd expect given that we're starting with data from 1997.

Re: Shaq, agreed it wouldn't be a big bump, but I do think it might be a slight shift up to have a few more true prime seasons and a slightly smaller percentage of post-prime seasons. Nothing hugely transformative, just a slight bump. Penny and Grant would definitely be better!

For Stockton and Malone, I might draw a connection to Garnett, or old Robinson, or Manu. Some players could play a 1b or 2 option on offense (or defense for that matter), but still end up being more valuable than their 1a teammate, on one or both ends. And we're left with a question of how to evaluate that. Some people say the 1a option must always be better than a 2nd option -- people like these might be lower all-time on Garnett for example. On the other end of the spectrum, one might say it's just about the value you add to the team, regardless of whether you're a 1st or 2nd offensive or defensive option or whatever. And there's plenty of room in the middle.

Me personally, I think I might be a little more forgiving to Stockton -- I'm okay with him being a 1b or 2 option, if his passing, decision making, spacing, and perimeter defense really was that much more valuable than Malone. Now there is of course the question of whether it was (this dataset covers many post-prime years, Malone played his final year in the Lakers with Kobe/Shaq which could lower his 1997–2004 average value, etc.)... maybe more prime to prime favors Malone. Playoff resilience / scalability could also be a factor, and if Stockton does become extra passive (particularly as a scorer) in the playoffs, that could also change our evaluation of him. But on the other hand, if there does start to be a collection of plus minus stats that are more friendly to Stockton > Malone than our previous data (e.g. this new sample, most of the newer Squared2020 RAPM), that does start to make me wonder whether Stockton's helping the team more than we're giving him credit for. Maybe not enough to raise him over Malone for peak/career, but maybe the gap between them shrinks, with Stockton a little higher and Malone a little lower.

Edit: just saw your longer post about Stockton and Malone. Good points! One thing to add: we also have single-season RAPM since 1997. in 97, Malone's definitely ahead of Stockton (4.42 > 3.88), but Stockton barely sneaks ahead in 98 (5.32 >5.31) and stays ahead in 99 (5.14 > 4.58) and 00 (6.18 > 4.12), and the Jazz were still Top 5 SRS teams in 98–00. Not saying RAPM is proof of anything here, Stockton's 97 RAPM is pretty disappointing, and the Jazz's odd rotations could add all sorts of issues to both on/off and RAPM. Just pointing out that there are some impact metrics that favor Stockton even when the Jazz were in contention.
source: Goldstein RAPM, one of the standard RAPM sources.

Other Current players: these are some other guys for whom we only have young/prime seasons. We might expect their ranks to drop over time if they're mid-peak now and haven't had any post-prime seasons to drag them down.
-Tatum at 7th overall is obviously great, propped up by being Top 15 offensively
-Embiid 10th overall is also great. He seems much more valuable defensively.

-So first I'll say that Embiid looking that great just seems right. The man's a monster out there.

-Tatum's placement though calls attention to something I feel like the analytical world hasn't really grappled with like it needs to with his career. That being with the exception of Jokic, Tatum might be the guy who jumped in with impact faster than anyone else in recent generations, and he's maintained elite standing as a matter of course with remarkable consistency. And this has happened while seeming to convince absolutely no one - and not me either - that he's truly a top tier MVP-type player. I think we've reached a point where we can't dismiss what we see as noise. Either there's something we're missing about Tatum, or there's something RAPM is missing about him (or both, of course), and wherever there's a gap in what's being accounted for, there's something for us to learn.

Agreed about Embiid!

Re: Tatum, it's interesting. He's ranked +5.1 in offensive RAPM here, 13th in the sample, slightly ahead of Nash, Luka, Kobe, Shaq, which just doesn't feel right. The Celtics have been a top 10 offense since Tatum's sophomore year, so perhaps there's a lot of offensive value to go around among the Celtics players. But this data definitely doesn't quite match people's impression. One place of disagreement might be resilience (e.g. people may have resilience concerns for Tatum's offense when facing pressure as a ball handler and passer against the top defenses, e.g. 2022 finals).

Another issue might be how to evaluate having a strong collection of shooters. Even outside of Tatum, the Celtics have had a really strong cast of shooters recently. Maybe Tatum is the one to 'stir the pot', e.g. his scoring threat could be what gets his teammates more open shots, which is what RAPM picks up on, whereas maybe people give more credit to the supporting cast for being such a strong shooting group in the first place and think the scoring threat of Tatum is more replaceable than RAPM. I'm not confident on any of this, just brainstorming ideas.

[MrVorp]

Broadly speaking, players are less impactful as they age, but it is not a hard rule, and I think it is erroneous for people to look at high values in the late 1990s and assume that they therefore must be even higher earlier. Especially when you are specifically looking at years where an equally old Karl Malone was similarly top of the heap.

John Stockton did not peak in 2001, nor would his 1997-2001 values demand that we treat him as some Magic level player in his actual prime. 2008-12 Garnett was not on balance a better player than 2005-07 Garnett even though I am pretty sure that is what pure RAPM suggests. RAPM is a reflection of lineups, league, role… and the 1996-98 Bulls were a bigger outlier relative to their league than the 1991-93 Bulls were.

This is true but I think we can have more confidence with some players than others. RAPM generally correlates with On/Off and from the sampled games Jordan’s On/Off is much more impressive than 97-98, so for him specifically I don’t think that it’s a stretch to assume his RAPM would have been higher during his prime years.

[AEnigma]

Broadly speaking, players are less impactful as they age, but it is not a hard rule, and I think it is erroneous for people to look at high values in the late 1990s and assume that they therefore must be even higher earlier. Especially when you are specifically looking at years where an equally old Karl Malone was similarly top of the heap.

John Stockton did not peak in 2001, nor would his 1997-2001 values demand that we treat him as some Magic level player in his actual prime. 2008-12 Garnett was not on balance a better player than 2005-07 Garnett even though I am pretty sure that is what pure RAPM suggests. RAPM is a reflection of lineups, league, role… and the 1996-98 Bulls were a bigger outlier relative to their league than the 1991-93 Bulls were.

This is true but I think we can have more confidence with some players than others. RAPM generally correlates with On/Off and from the sampled games Jordan’s On/Off is much more impressive than 97-98, so for him specifically I don’t think that it’s a stretch to assume his RAPM would have been higher during his prime years.

His on-court was not though, and RAPM cares about both. His highest on-court values are likely 1996, 1992, and 1997. His highest on/offs would be 1988-90.

[MrVorp]

Broadly speaking, players are less impactful as they age, but it is not a hard rule, and I think it is erroneous for people to look at high values in the late 1990s and assume that they therefore must be even higher earlier. Especially when you are specifically looking at years where an equally old Karl Malone was similarly top of the heap.

John Stockton did not peak in 2001, nor would his 1997-2001 values demand that we treat him as some Magic level player in his actual prime. 2008-12 Garnett was not on balance a better player than 2005-07 Garnett even though I am pretty sure that is what pure RAPM suggests. RAPM is a reflection of lineups, league, role… and the 1996-98 Bulls were a bigger outlier relative to their league than the 1991-93 Bulls were.

This is true but I think we can have more confidence with some players than others. RAPM generally correlates with On/Off and from the sampled games Jordan’s On/Off is much more impressive than 97-98, so for him specifically I don’t think that it’s a stretch to assume his RAPM would have been higher during his prime years.

His on-court was not though, and RAPM cares about both. His highest on-court values are likely 1996, 1992, and 1997. His highest on/offs would be 1988-90.

His 97-98 on-court was +11. In 91 he was +10 on court without a 20+ game sample of his team playing at a +15 overall. In the 93 sample he was also roughly +10. I don’t really see much of a difference.

[AEnigma]

This is true but I think we can have more confidence with some players than others. RAPM generally correlates with On/Off and from the sampled games Jordan’s On/Off is much more impressive than 97-98, so for him specifically I don’t think that it’s a stretch to assume his RAPM would have been higher during his prime years.

His on-court was not though, and RAPM cares about both. His highest on-court values are likely 1996, 1992, and 1997. His highest on/offs would be 1988-90.

His 97-98 on-court was +11. In 91 he was +10 on court without a 20+ game sample of his team playing at a +15 overall. In the 93 sample he was also roughly +10. I don’t really see much of a difference.

To whatever extent we want to take those samples at face value, it sounds more like your main issue is 1998 being a clearer step down.

[eminence]

This is true but I think we can have more confidence with some players than others. RAPM generally correlates with On/Off and from the sampled games Jordan’s On/Off is much more impressive than 97-98, so for him specifically I don’t think that it’s a stretch to assume his RAPM would have been higher during his prime years.

His on-court was not though, and RAPM cares about both. His highest on-court values are likely 1996, 1992, and 1997. His highest on/offs would be 1988-90.

His 97-98 on-court was +11. In 91 he was +10 on court without a 20+ game sample of his team playing at a +15 overall. In the 93 sample he was also roughly +10. I don’t really see much of a difference.

I've found simply adding the two (on + on/off) correlates reasonably well with RAPM type numbers, certainly more clearly than one or the other.

From the MJ thread, RS samples only (/48 minutes)
'85: +2.20 On, +13.85 On/Off, +16.05 (27 game sample)
'88: +5.57 On, +21.57 On/Off, +27.14 (42 games)
'91: +10.03 On, +17.12 On/Off, +27.15 (56 games)
'92: +12.56 On, +23.34 On/Off, +35.90 (55 games)
'93: +9.42 On, +14.66 On/Off, +24.08 (63 games)
'96: +15.22 On, +13.87 On/Off, +29.09 (full)
'97: +12.64 On, +8.75 On/Off, +21.39 (full)
'98: +9.02 On, +9.93 On/Off, +18.95 (full)

Caveats about partial samples in mind.

I'd say it's likely that '88-'96 would be a step up from '97/'98 in an APM style measurement, likely peaking in '92.

[OhayoKD]

His on-court was not though, and RAPM cares about both. His highest on-court values are likely 1996, 1992, and 1997. His highest on/offs would be 1988-90.

His 97-98 on-court was +11. In 91 he was +10 on court without a 20+ game sample of his team playing at a +15 overall. In the 93 sample he was also roughly +10. I don’t really see much of a difference.

I've found simply adding the two (on + on/off) correlates reasonably well with RAPM type numbers, certainly more clearly than one or the other.

From the MJ thread, RS samples only (/48 minutes)
'85: +2.20 On, +13.85 On/Off, +16.05 (27 game sample)
'88: +5.57 On, +21.57 On/Off, +27.14 (42 games)
'91: +10.03 On, +17.12 On/Off, +27.15 (56 games)
'92: +12.56 On, +23.34 On/Off, +35.90 (55 games)
'93: +9.42 On, +14.66 On/Off, +24.08 (63 games)
'96: +15.22 On, +13.87 On/Off, +29.09 (full)
'97: +12.64 On, +8.75 On/Off, +21.39 (full)
'98: +9.02 On, +9.93 On/Off, +18.95 (full)

Caveats about partial samples in mind.

I'd say it's likely that '88-'96 would be a step up from '97/'98 in an APM style measurement, likely peaking in '92.

Eh...

There is no way that +45 on/off net rating is remotely accurate.

1. Dipper had similarly high on/off estimates for Barkley (+36.3) and Hakeem (+34.5) but those estimates were disproven with Harvey Pollack's full season data.

Barkley net on/off
1987: +7.9
1988: +2.7
1989: +11.0
1990: +8.3
1991: +8.8
1992: +6.0
1993: ?
1994: +6.8
1995: +6.8
1996: +7.8

Hakeem net on/off
1994: +14.5
1995: +9.0
1996: +10.3

In none of the 12 full seasons did they come anywhere near the +35 on/off estimate over the same period.

2. There has only been a handful of +20 on/off net seasons in the last 25 years, none above +25. A +35 or + 45 on/off was never realistic in the first place, especially not by 3 players at the same time in just the few seasons prior. This was most likely just a calculation error somewhere in Dipper's spreadsheet which was repeated when he plugged in new numbers for different players.

[eminence]

His 97-98 on-court was +11. In 91 he was +10 on court without a 20+ game sample of his team playing at a +15 overall. In the 93 sample he was also roughly +10. I don’t really see much of a difference.

I've found simply adding the two (on + on/off) correlates reasonably well with RAPM type numbers, certainly more clearly than one or the other.

From the MJ thread, RS samples only (/48 minutes)
'85: +2.20 On, +13.85 On/Off, +16.05 (27 game sample)
'88: +5.57 On, +21.57 On/Off, +27.14 (42 games)
'91: +10.03 On, +17.12 On/Off, +27.15 (56 games)
'92: +12.56 On, +23.34 On/Off, +35.90 (55 games)
'93: +9.42 On, +14.66 On/Off, +24.08 (63 games)
'96: +15.22 On, +13.87 On/Off, +29.09 (full)
'97: +12.64 On, +8.75 On/Off, +21.39 (full)
'98: +9.02 On, +9.93 On/Off, +18.95 (full)

Caveats about partial samples in mind.

I'd say it's likely that '88-'96 would be a step up from '97/'98 in an APM style measurement, likely peaking in '92.

Eh...

There is no way that +45 on/off net rating is remotely accurate.

1. Dipper had similarly high on/off estimates for Barkley (+36.3) and Hakeem (+34.5) but those estimates were disproven with Harvey Pollack's full season data.

Barkley net on/off
1987: +7.9
1988: +2.7
1989: +11.0
1990: +8.3
1991: +8.8
1992: +6.0
1993: ?
1994: +6.8
1995: +6.8
1996: +7.8

Hakeem net on/off
1994: +14.5
1995: +9.0
1996: +10.3

In none of the 12 full seasons did they come anywhere near the +35 on/off estimate over the same period.

2. There has only been a handful of +20 on/off net seasons in the last 25 years, none above +25. A +35 or + 45 on/off was never realistic in the first place, especially not by 3 players at the same time in just the few seasons prior. This was most likely just a calculation error somewhere in Dipper's spreadsheet which was repeated when he plugged in new numbers for different players.

As the person who replied similarly just ahead of LA Bird in that thread. Yes I'm aware of questions with Dipper's work and highlight sampling bias generally.

'92 is still most likely his high end result. MJ being individually better prior to his first retirement is not some fringe position, it was the team high water mark in that iteration, and the limited data we do have supports it. If you believe differently I'd love to hear something more detailed than 'ehh'.

[AEnigma]

The question was not whether he was individually better, the question was how much better he would look by an equally applied measure of impact. I think with how inflated top players were in the late 1990s, it would not be much — and 1992 is not an especially convincing counter-point when I do not see any real case that he was better in 1992 than in 1990 and 1991 (and even 1989 and 1988 would need to be specifically using a lens of postseason experience that applies more to old Jordan anyway).

[MrVorp]

His on-court was not though, and RAPM cares about both. His highest on-court values are likely 1996, 1992, and 1997. His highest on/offs would be 1988-90.

His 97-98 on-court was +11. In 91 he was +10 on court without a 20+ game sample of his team playing at a +15 overall. In the 93 sample he was also roughly +10. I don’t really see much of a difference.

To whatever extent we want to take those samples at face value, it sounds more like your main issue is 1998 being a clearer step down.

My point is that people saying Jordan’s late 90’s RAPM looks so great, he must have looked even better in his prime have reasonable justification to do so.

[AEnigma]

His 97-98 on-court was +11. In 91 he was +10 on court without a 20+ game sample of his team playing at a +15 overall. In the 93 sample he was also roughly +10. I don’t really see much of a difference.

To whatever extent we want to take those samples at face value, it sounds more like your main issue is 1998 being a clearer step down.

My point is that people saying Jordan’s late 90’s RAPM looks so great, he must have looked even better in his prime have reasonable justification to do so.

Sure, but my original point was that you cannot really assume the degree. He was better when he was younger, but not because of some hypothetical RAPM output.

[eminence]

The question was not whether he was individually better, the question was how much better he would look by an equally applied measure of impact. I think with how inflated top players were in the late 1990s, it would not be much — and 1992 is not an especially convincing counter-point when I do not see any real case that he was better in 1992 than in 1990 and 1991 (and even 1989 and 1988 would need to be specifically using a lens of postseason experience that applies more to old Jordan anyway).

Fair enough.

The '92 Bulls are very slighly behind the '97 team in team measures of goodness (despite the stronger league). They pretty clearly surpass the '98 version.

Hence, any combined '97/'98 measure will favor the '92 version in on-court (+11 > +10 and all that). The other half (on/off) we don't have the data outside of samples given, but those very clearly favor '92 versions. (similar thoughts all apply to '96 likely outpacing '97/'98 as well)

How large the margin would be is very up in the air, but evidence points towards '92, and imo, pretty clearly. What do you find unclear about it?

Semi aside - I have '91 on a similar level to '92, but I have '90 a half step down. I find it difficult to believe a team jumped from +3.3 to +9.5 while keeping the same top 7 guys and that the #1 guy didn't improve at least a bit.

[AEnigma]

I do not find it unclear which is better. Again, my point is that RAPM is not really giving me the “answer”.

As for 1992’s value distinct from whatever the RAPM measure is, I feel like you would basically need to assess that as Jordan’s true defensive peak, or at least narrowly behind 1988, to make up for the offensive drop-off from 1989-91 (or 1989/90 if we want to exclude 1991 from this frame as a reasonable top season choice either way). Maybe you would see that in some hypothetical 1992 RAPM, but if it does not come through on film then I am not particularly interested in it.

[falcolombardi]

The question was not whether he was individually better, the question was how much better he would look by an equally applied measure of impact. I think with how inflated top players were in the late 1990s, it would not be much — and 1992 is not an especially convincing counter-point when I do not see any real case that he was better in 1992 than in 1990 and 1991 (and even 1989 and 1988 would need to be specifically using a lens of postseason experience that applies more to old Jordan anyway).

Fair enough.

The '92 Bulls are very slighly behind the '97 team in team measures of goodness (despite the stronger league). They pretty clearly surpass the '98 version.

Hence, any combined '97/'98 measure will favor the '92 version in on-court (+11 > +10 and all that). The other half (on/off) we don't have the data outside of samples given, but those very clearly favor '92 versions. (similar thoughts all apply to '96 likely outpacing '97/'98 as well)

How large the margin would be is very up in the air, but evidence points towards '92, and imo, pretty clearly. What do you find unclear about it?

Semi aside - I have '91 on a similar level to '92, but I have '90 a half step down. I find it difficult to believe a team jumped from +3.3 to +9.5 while keeping the same top 7 guys and that the #1 guy didn't improve at least a bit.

Semi aside - I have '91 on a similar level to '92, but I have '90 a half step down. I find it difficult to believe a team jumped from +3.3 to +9.5 while keeping the same top 7 guys and that the #1 guy didn't improve at least a bit.

Is this a valid argument? Team improved so jordan -must- have too even if neither boxcore nor eye test suggest it?

is a rather heavy assumption specially because i doubt you think jordan hypothetical improvement from 90 to 91 was worth 6 points so i assume you are already attributing the lion share lf the leap to non direct jordan element

Even more so when the 90 team that finished the season was already rather stronger than the 90 team than started it as shown in their margin lf victory across either half of their year, making the overall 90 team net rating non all that representative