RAPM VORP Spreadsheet & Analysis

[Doctor MJ]

Hey folks, here's a projected I've been working on that I want to share and hopefully discuss.

Doc's NBA RAPM VORP Career Leaderboards

The first tab looks to explain a bunch of things so please look there if you're confused.

In a nutshell though:

I've taken Englemann's Career (Regular Season) RAPM, applied a Replacement Player level to it, and with player minutes calculated RAPM-based VORPs for offense, defense & overall.

I'm using a Replacement Level of -2.6 on offense, 0.7 on defense, and -3.3 overall which is close to something I've seen used, but there is no conensus, so I've made the workbook so that if you make a copy, you can just change the RP and then look on the other tabs to see how it changes things.

You'll see other tabs where I've grouped by draft, origin, etc, which I find to be interesting, though I'm also intending to use this thread to give other views that I'll give thoughts on, and hope others give thoughts too.

Some additional things that people should be aware of when looking at this data and comparing to how you see a player's career:

1. This doesn't include the playoffs. Englemann has one of these for the playoffs going through 2024 but I'm hoping we get another one for 2025 before doing a similar project for the post-season. (Note that there are some important points of discussion when looking at Playoff RAPM that unfortunately make it too complicated to say "Playoff RAPM always tells us how good a player is against playoff levels of competition", but it'd still be good to see that VORP calculation).

2. Always keep in mind that when we do a VORP like this, a theoretical player that is only slightly above RP but who is immortal and plays for 500 years could top the list. I've long been on record as believing that a player's accomplishment is much more about what they do at their best than what they do in the long tail of longevity, and that's still how I see it, but it's useful to me to see VORP as a starting point.

3. Parallel to this I've been doing a project that bins player spans by RAPM rank and this gives a much better view on who was truly elite and for how long. In the process of doing this using nbarapm they updated their methods, so that means I have to start some things over. Previously I focused primarily on 4-year RAPM's because that's what that site defaulted too, but I have mixed feelings about this because arguably both the 3-year and 5-year models are more used than 4-year models.

As with replacement level, I'd love to get feedback from folks as to what X-year model people see as most meaningful and why.

Similarly, if people have specific arguments to make for other sites (The Basketball Database for example) doing a better job than nbarapm. Frankly I don't think it should matter too much larger sample models so long as there's no gross incompetence, but the shorter the sample, the more the parsing of the NBA PBP possession data itself swings things (ex: blaming a new substitution for a free throw made which was drawn while he was on the bench).

[tsherkin]

Very interesting. Thanks for sharing, Doc.

[Hoop Hunter]

Nice work. Very interesting. It would be cool to just see playoff rankings.

It validates my feeling about Reggie Miller. The deeper down you go the better he looks. Raw PTs, Rebs, Ast don't tell the whole story.

[iggymcfrack]

Love the data. Gives excellent perspective over just looking at rate numbers and trying to just guess a longevity adjustment.

Makes me feel like maybe Paul Pierce is underrated from an all-time perspective. His consistently positive defense over the course of his career provides a lot more value than you’d guess just from his box score numbers. He should probably be somewhere around #35 all-time.

[NZB2323]

Surprised to see Pierce 8th. I thought generally advanced stats people weren’t high on Pierce.

[Doctor MJ]

Alright folks, it's kinda my intent to work with the data in this spreadsheet within this thread a lot. It's possible no one will be that interested but me, but please do join in.

I'll emphasize again that the data itself is as objective as I could make it and doesn't necessarily say anything simple about any given player.

First thing I'm going to focus on is the "RealGM 100 Perspective" tab where this data (superficially at least) appears to tell a different story that we should consider.

The first 22 spots in the tab's list are all guys who were in the last 100, after that non-100 member are in there above 100 members.

In those first 22 spots, here's what I see as most noteworthy:

1. LeBron is way ahead of everyone else, way ahead of everyone on offense, and ranks 4th on defense in an elite group that is largely made up of DPOYs along with Duncan (should've been DPOY), and some other guys that were at least near DPOY levels (Sheed, Divac, Iguodala).

This really shows how much of an outlier LeBron is with his combination outlier performance and consistency over a record breaking longevity.

2. Garnett has a sizable lead over everyone else, and obviously that includes Duncan who he leads on both sides of the ball. In particular, Garnett's defense VORP is just way ahead of everyone else.

I'll say this to me give some definitive clarity on what RS RAPM says about Garnett vs Duncan and everyone else in the non-LeBron class. All of the early +/- & APM data appeared to favor Garnett above all (LeBron wasn't in prime yet), but the RAPM data that emerged later sometimes appeared to favor Duncan over Garnett.

Now, as I said then (and other times later), one major concern about RAPM (relative to APM) is a tendency to effectively "disbelieve" in certain outliers. Whether using a neutral prior or a box score prior, I'd say there's a tendency to essentially bias RAPM toward on-court +/- over On-Off, which can hurt the case for someone who is putting up massive On-Off with weak teammates - as Garnett had.

To be clear: When I'm doing my later iterations of analysis, I tend to favor on-court over On-Off as well, so I'm not saying that this isn't something that often makes sense, so much as that I'd like to make sure that I'm the one doing that biasing rather than get it from a black box.

Regardless, with greater sample, these issues diminish, and thus over long careers if we see one guy have the edge, to me that's more ground truth. This is particularly true when the guy losing a particular debate is like Duncan and is known for fantastic longevity of impact. When Duncan has a role on a championship team 17 years after he enters the league, we basically know that this isn't a guy having his prime particularly weighted down by the fringe of his career.

(I've said already but I'll emphasize again here: This is regular season data, so anyone favoring Duncan because of how you see his playoff performance, this data isn't arguing against you.)

Last note on KG that is relevant for him, but will be even more relevant for players whose careers began earlier:

Garnett's rookie year isn't in the PBP era, so when we project backwards for VORP, we end up treating his rookie campaign as being the average of the rest of his career. I expect that overrates him a bit - not much because it's just one season out of a very long career - but to the extent all of these projected guys are either overrated or underrated by the method, I expect KG's is a touch over.

3. It's not a shock to me that Paul is very high on this list, but I do have to confess I'm surprised he ranked 3rd (also) ahead of Duncan. With him being ranked #20 in our last 100, it's an immediate point of conversation.

Now, a guy like Duncan has playoff arguments for days over Paul, but how many other guys should? It's a good question I'd love to see some conversation on.

I should also say that sepearate from this 100-based analysis, I am intending to look at other perspectives such as elite pass-first point guards.

4. Let me mention 1998 draft mates Dirk (#5), Pierce (#7) & Carter (#14) together. While Dirk being that high probably doesn't surprise a lot of folks here, if you're like me, you're surprised that Pierce & Carter look so strong. Carter in particular had a rep being something of a failed superstar when spent the rest of his career with his tail between his legs, but as you can see here, the data considers him to be consistently valuable for a long, long time.

5. Stockton having a high place (#6) on this list is something many will expect and there's a lot to talk about there. I do think that the fact he ranks above Jordan (#10) is the thing that immediately warrants conversation. There's long been a major chunk of analysts looking to argue that Stockton was the more impactful Jazzman over the Mailman (who places 21st here), but arguing for him over Jordan is different.

As some probably theorize right away, Jordan's 2nd retirement looms large here. By nbarapm's 2-year RAPM model, Jordan's final 2 years place him as ranking #2 in RAPM in that span, and #1 on offense (Jordan also ranks #1 overall for the 1999 2-year model which only factors in Jordan's final Bull season). Meanwhile, when Jordan returns, he never ranks in the Top 50. Stockton meanwhile ranks 8th in the 1998 2-year model, but ends up climbing up to #1 by 2001.

This then to say that RAPM in Jordan's available Bull years definitely favors Jordan over Stockton, but Jordan's Wizard years - which in this career sample count for about 50% of Jordan's "career" drag him down.

Now by contrast, n the case of Stockton & Malone, in those early PBP years Stockton has the edge in overall RAPM, though I will note that in that time frame Malone has the higher Offensive RAPM numbers and is playing considerably more. Stockton's late career impact resilience is to be applauded, but I've never been comfortable arguing that Malone wasn't the team MVP during their elite contending years.

6. We see contemporaries Durant (#8), Curry (#12) & Harden (#13) clustered pretty close together, and we note specifically that Harden has the highest of the Offensive VORP while Curry has the lowest. That's a funny thing given that many of us see Curry as the best offensive player of that cohort, so what's going on there?

Well, the simplest thing here is that Curry has the best Overall & Offensive RAPM of the trio, he's just played less. (Note that Harden was in the same draft class, so that speaks to Curry missing time to injury.)

7. Kidd comes in at an extremely respectable #11, which while it ranks below Paul at 3 and Stockton at 6, puts him above Nash (#18).

I do expect that Kidd's numbers are slightly overestimated by projection because those early Dallas years were brutal, but I expect Kidd's lead over Nash (among others) would hold.

8. Everyone knows I'm a Nash guy and have argued for Nash over Kidd for years, so it makes sense for me to address briefly here why Kidd's edge here doesn't surprise me.

In a nutshell, this is related to Nash seeming to come out of nowhere to win an MVP at 30. For Nash, the first time he has a 2-year RAPM in the Top 25 is in the 2005 model, whereas Kidd first gets there in the 2000 model (though he does drop in 2001 before peaking higher in 2002).

As always, not begrudging those who favor Kidd over Nash due to longevity, but that longevity is definitely the big thing in play here as Nash peaks higher in all their 2-year through 5-year models.

9. We'll note that the top 3 from the 1996 draft class - Allen (#15), Kobe (#16) & Nash (#18) - are all right next to each other which is interesting. In a draft with 3 MVPs (Iverson, Nash, Kidd), it's Allen who tops this list.

Allen's a legend who deserves prominence on these lists, but he's also the guy with the most graceful impact longevity. Nash started slow and ended abruptly, meanwhile Kobe had horrific last few years by these stats.

10. Finally in this elite 22, we have Sheed (#19), Jimmy (#20) & Jokic (#22).

Sheed being so high isn't a surprise to those who are familiar with +/- data. If you just take that data at face value, he should rank easily ahead of his Pistons teammates, but of course as a franchise player, Sheed was problematic.

Jimmy being so high I think would surprise a lot of folks given that he's primarily known for "Playoff Jimmy". Hell of a career he's had.

Jokic of course is at the vanguard of the next generation, and so it's no surprise he's rapidly climbing a list like this.

Okay, I think that's enough for one post, so I'll submit and talk about those not in that top 22 - and particular those not in the 100.

[eminence]

Replacement level - I'd reach out to JE and use whatever values he used to make the dataset. I like a number closer to 0 than he does, but if one is using his data I think you stick with the number he used. First guess is it's probably even a bit lower than -3.3.

I'm not a big fan of the projection stuff, though it's nice to have additional names on the list. Obviously projecting KG is a lot different than Stockton.

Keeps with the general view that gaps get smaller as we go down, eg 15-25 are within 20% of one another, but it's almost a 50% gap up to #5 from #15. Folks should be pretty flexible on specific ordering once we're leaving that very top tier. Longevity generally favored here over what I would expect from most folks criteria, including my own.

I've done some similar things myself so not too shocked by any of the results I've seen scrolling so far, but joining your semi-surprise at Carter. He always looks nice by career long looks, but when going through to look at his individual seasons I can't grade him out that highly.

I'm looking forward to the smaller stint stuff (I'd vote 5 over 3 of common options).

Good effort and well organized Doc, I'll keep following along.

[Doctor MJ]

Alright so now I'm going to focus on the highest guys on this list who didn't make the Top 100.

Let me first just place the 10 top guys along these lines:

23. Mike Conley
24. Jrue Holiday
26. Vlade Divac (debut in 1989, significantly projected - also had made a past 100)
28. Andre Iguodala
35. Detlef Schrempf (debuted in 1985, heavily projected)
40. LaMarcus Aldridge
41. Rashard Lewis
43. Lamar Odom (highest ranked non-all-star)
44. Baron Davis
45. Eddie Jones (debuted in 1994, slightly projected)

Now, these are the kind of guys that I definitely like to evaluate in smaller cohorts (like draft classes) because they aren't the ultra-outliers, but obviously they stand out, and I think most wouldn't expect they stand out like they do.

The question I'm looking to focus on here is: Which of these guys should get serious consideration for the 100 next time we do it?

- I'll start with Divac who has made it in the past, and who I'd say generally gets some discussion in those projects. Here I'd double down and say "Yup, take Divac very seriously." Here I'll also note that Divac is more known for his time with the Kings than anywhere else (partially because of that ignominious GMing run but still), but within that core was not seen as either "the star" or even "the other star". Yet he towers over those teammates on this list (I want to do a core-based comparison of great teams in the near future, and this will illustrate this further).

- I think I'd be remiss not to make the connection between Divac, Schrempf, and Arvydas Sabonis, all of whom were part of that early group of European players to make their way over to the US & NBA. All of them were just generally more impactful than you'd expect given their box score.

- While Conley is the top of that list, I have a hard time seeing him as having a better career than Jrue because of the playoffs. Conley's been a quite solid payoff performer, but Jrue's become something of a legend there. With Jrue playing a critical role joining a second would-be contender and elevating them to the chip since our last 100, honestly, he feels to me like he should make the next 100.

- Similarly, Iggy deserves some serious respect. While teammate Klay Thompson got the all-star love as "the other star" for that Warrior team along with Steph & Dray (and eventually KD), I'd say Iggy should be seen as the greater player.

- I'm not an Aldridge guy, and many here know that - I felt like the way he changed from being a guy claiming to be "Blazer for life" to wanting out when Dame Lillard emerged as the superior offensive engine really rubbed me the wrong way given that Dame never seemed to have any negative toward him, and while he later embraced a secondary role next to Kawhi, there was drama in that process too, and so in the end I see him as a guy too attached to his scoring volume and not attached enough to his teammates . Nevertheless, this data certainly says good things about him, and maybe I hold some things against him too much.

- I would say Lewis & Odom represent something similar. While Lewis became an all-star and Odom didn't, both were guys where were outstanding secondary players for contenders.

- Baron's a guy I've long tried to get people's attention on, though I've never actually voted for him in the 100 that I recall. He's just a guy who was consistently valuable even when his box score seemed to say "chucker", and of course there's also the matter of him being known for raising his play (and putting his body on the line for free throws) come the post-season.

- Eddie is a guy with a pretty bad playoff rep which I'd say had everything to do with why he stopped being seen as an all-star level guy while still in prime. I do think his RS RAPM data probably overrates him relative to what he can do in playoff competition, but I don't think he deserve the status of "the guy you trade so you can win a title". If you know about his excellent defense and you note his 3-point shooting, I think he could have been something of a super-role player for a contender with the prescience to see how the league would change.

[Doctor MJ]

Last shot, in this volley, here are the guys who are or have been in our 100 previously who score the lowest by this metric (though note, I didn't try to project anyone pre-Drexler who came in in 1983). In this case, I'm going to include the VORP number I have for them, and just remember when you look at the tab in question, I didn't include any non-100 guys who had a VORP below 1000.

1. Mark Price 129 (heavily projected)
2. Dennis Rodman 692 (heavily projected)
3. Arvydas Sabonis 876 (slightly projected)
4. Anfernee Hardaway 948 (significantly projected)
5. Kevin Johnson 952 (significantly projected)
6. Shawn Kemp 1113 (significantly projected)
7. Luka Doncic 1131
8. Chris Webber 1172 (significantly projected)
9. Elton Brand 1187
10. Chris Mullin 1222 (heavily projected)

Obviously this list is dominated by guys who are projected, and so that's probably the #1 thing to focus on, but to just jump in on these guys:

- I would say that guys like Price, Penny & KJ are guys whose numbers are severely damaged because we're having to project their pre-injury years. While we know they all of have longevity issues, I would be careful taking this data as if it speaks to them at their best.

- Rodman's an interesting one because by age, it makes sense that this data didn't see him at his best...but by fame and acclaim, Rodman being the 3rd in the Big 3 that resulted in the Bulls' 2nd 3-peat would seem to indicate that we should see huge impact if he really was all that.

I'll note that prior to us getting this +/- data, there was a contingent of folks arguing that maybe Rodman was actually more valuable than Pippen & Jordan because his rebounding specialization gave so much value add. We really don't see anything like that though. Rodman may well have been super impactful for the Pistons, but it seems likely that he eventually got overrated amidst the chips and the shock value.

- Sabonis's numbers aren't projected much because he only came to the NBA in 1995, but of course, his prime was in the '80s. Relative to how good people though he was in the NBA at the time (a non-all-star), his impact was really, really solid though. (Note that when he made the 100, we didn't have a hard rule against used non-NBA play to count toward achievement. Ever since that rule was instated, Sabonis hasn't made the 100, so I wouldn't say he's necessarily a strong candidate to make it going forward under these rules, but man, what a player!)

- Kemp, Webber & Brand are guys who I think are landing in about the right spot, and being in that 1000 club is no small thing...but it is a smaller thing than those who have championed them had tended to think. In the case of all 3, I'd say we're talking about guys who became seen as stars for their PPG while playing positions that expected the player to be defense-oriented. Doesn't mean they were terrible defenders, but their offense generally wasn't as impactful as people though, and meanwhile contemporary bigs with great defense gained major advantages over these guys on that front.

- Finally we've got Luka here, and of course Luka's here for reasons relating to him being the youngest guy in the last 100, and I believe the youngest guy to make the 100 since LeBron.

Okay, I'll refrain from posting more of these for a bit, and see what conversation comes. Cheers!

[Doctor MJ]

Replacement level - I'd reach out to JE and use whatever values he used to make the dataset. I like a number closer to 0 than he does, but if one is using his data I think you stick with the number he used. First guess is it's probably even a bit lower than -3.3.

I'm not a big fan of the projection stuff, though it's nice to have additional names on the list. Obviously projecting KG is a lot different than Stockton.

Keeps with the general view that gaps get smaller as we go down, eg 15-25 are within 20% of one another, but it's almost a 50% gap up to #5 from #15. Folks should be pretty flexible on specific ordering once we're leaving that very top tier. Longevity generally favored here over what I would expect from most folks criteria, including my own.

I've done some similar things myself so not too shocked by any of the results I've seen scrolling so far, but joining your semi-surprise at Carter. He always looks nice by career long looks, but when going through to look at his individual seasons I can't grade him out that highly.

I'm looking forward to the smaller stint stuff (I'd vote 5 over 3 of common options).

Good effort and well organized Doc, I'll keep following along.

Thanks for responding eminence!

Well with regards to RP level, to me the key thing is that there isn't a consensus that I've ever found, and so I wanted to make a resource that would let people set their own levels. I'm not opposed to arguments that one particular RP level distinct from what I used is best, but I probably won't tinker with what I'm doing unless I'm convinced of something.

Re: What Englemann uses when making the dataset. Well, for RAPM itself, there is no replacement level factored in (so far as I've ever undestood). You just run the data with whatever prior, and the zero level basically represents the average quality of play in the league by minute played, and those minutes are made up disproportionately of the minority of players who are above that average quality.

I would say that it would have been pretty trivial for Engleman to release VORP for his RAPM if he wanted, and the fact he hasn't done so speaks to him being reluctant to commit to an RP level. Which to be clear, I get if you're a public figure who doesn't want your work tied to anything people see as arbitrary.

Worth noting that while people might argue that there's a host of arbitrary things in any given RAPM, I'd say Engelmann has been pretty clear about him using prediction/retrodiction to optimize for things run the risk of arbitrariness, which is part of why I have a tendency to use his long-term RAPM data where it's available over other sources.

(On the other hand, I've long been critical of his preferred xRAPM approach which maxes out the prediction/retrodiction further, but factors in a grab bag of stuff that renders the methods a black box.)

Re: not a fan of the projection. Understandable and I'll note that the Raw tab has PBP Era VORP to go along with the Projected VORP.

ftr, here are the guys in that original list of 22 who are projected, and how their ranking changes if we use PBP ERa:

2. Kevin Garnett (stays at 2)
6. John Stockton (drops to 81st)
9. Shaquille O'Neal (drops to 16th)
10. Michael Jordan (drops to 125th)
11. Jason Kidd (stays at 11)
17. Reggie Miller (drops to 74th)
19. Rasheed Wallace (rises to 17th)
21. Karl Malone (drops to 121st)

So yeah, these guys from the '80s just fall off a cliff if I don't project which isn't necessarily a problem depending what you're trying to do, but I do think that the Projected values better align with how they are seen.

[SHAQ32]

If I'm reading this right, which I'm probably not, Kidd looks too high on offense and David Robinson too low. Those are just the first two I noticed.

[eminence]

Re: What Englemann uses when making the dataset. Well, for RAPM itself, there is no replacement level factored in (so far as I've ever undestood). You just run the data with whatever prior, and the zero level basically represents the average quality of play in the league by minute played, and those minutes are made up disproportionately of the minority of players who are above that average quality.

You appear to be right, JE really has left everyone in. Hmm, not sure how I feel about that, seeing the Ade Murkeys of the world included is non-traditional. Probably of marginal overall effect at that sample size level.

I think it would still be worth reaching out to JE and seeing where he'd set it based on that data, he'll always have more info than us on why/how he set various parameters.

[GeorgeMarcus]

Fantastic work

[TheShow2021]

Why does Hakeem have a -2 average offensive RAPM? Sorry if you already mentioned this, but what year does it start?

[GeorgeMarcus]

Side note I think Vince is VASTLY overrated and suffers from "not having any particular fan base supporting him" syndrome. He lost that with Toronto the way they left things and never stuck around long enough elsewhere to build that sort of rapport

[eminence]

Why does Hakeem have a -2 average offensive RAPM? Sorry if you already mentioned this, but what year does it start?

'97-'25 I believe.

I wouldn't worry too much about guys who started in the 80s. It's nice if they look nice in late 90s apm, but not a big deal if they don't.

[Verticality]

This is quite interesting. Incredible work Doctor MJ.

[ReggiesKnicks]

Where is Jeff Foster?

[Doctor MJ]

Replacement level - I'd reach out to JE and use whatever values he used to make the dataset. I like a number closer to 0 than he does, but if one is using his data I think you stick with the number he used. First guess is it's probably even a bit lower than -3.3.

I'm not a big fan of the projection stuff, though it's nice to have additional names on the list. Obviously projecting KG is a lot different than Stockton.

Keeps with the general view that gaps get smaller as we go down, eg 15-25 are within 20% of one another, but it's almost a 50% gap up to #5 from #15. Folks should be pretty flexible on specific ordering once we're leaving that very top tier. Longevity generally favored here over what I would expect from most folks criteria, including my own.

I've done some similar things myself so not too shocked by any of the results I've seen scrolling so far, but joining your semi-surprise at Carter. He always looks nice by career long looks, but when going through to look at his individual seasons I can't grade him out that highly.

I'm looking forward to the smaller stint stuff (I'd vote 5 over 3 of common options).

Good effort and well organized Doc, I'll keep following along.

So let me say what I see as the big elephant in the room:

One point in APM is a basketball point. One point in RAPM generally isn't (it's not held still and the optimization process alters it).

So what that means is that for any RAPM study, the author should be providing Replacement Player level as a matter of course...but they don't, despite the fact such calculations are surely trivial for them. I have to assume they don't calculate (or share) them because they lack confidence as to what they mean, but maybe I'm wrong.

This then to say I'd really like super-long term studies to abandon RAPM for APM...which of course I've been saying ever since RAPM came out.

Of course as I say this: I don't really think super-long term studies are the best way to calculate RP. I think the right way to do it is probably to focus on 1-year data, because "replacement level" players are guys who are popping in over the course of a given season, and that doesn't necessarily reflect what they are over the course of their career.

So then what I'd probably do - if I had more patience scraping nba.com data (an embarrassing character flaw that would be good to get passed) - what I'd do is set players below a certain MPG level (not MP) for a given season, run the APM regression, find the RP, then repeat this for all years we have the data.

At that point, it would depend on what the data said. If the numbers are consistent without evidence of trend, I'd just average them to get my RP level.

If there's anything more interesting than that though, well, I'd want to talk to folks (like you) about it to try to figure out what we think it means.

It's possible RP is getting more (or less) negative with time, or that we can specifically note era shifts based on changes to rule & strategy.

Anywho, let me ask this:

If I made another model based on a different RP for side-by-side analysis, is there anything you would want? If memory serves when we talked about this before we talked about a couple possibilities:

a) -2.0
b) One-fifth of the SRS for a historical team that seems maximally a stop-gap team filled with guys not there the prior year or the year after. And if memory serves, the team that stood out to me like this was the '97-98 Nuggets who had an SRS of -11.74. So that's going to give us something a bit more negative than -2.0, but not nearly as negative as the -3.3 I'm currently using.
c) Let me also just emphasize that because I feel pretty confident that the number I'm using is on the high end of what's been calculated by folks in the past, I like the idea of specifically choosing a low-end estimate for the 2nd study. Doing so won't just show us more possibilities, but will show us how much longevity boost a high-end estimate (like my extremely negative -3.3) is "bringing in", and from there we can each decide what to do with it.

A couple other things on this I'd note:

If I simply scaled to -2.0, then I'd probably keep the proportions between offense & defense the same.

If I use the Nuggets, I could use something based on their rORtg & rDRtg. It seems honestly like what I should do if I'm calling the Nuggets the "replacement team"...but the thing is, their rDRtg was worse than the rORtg. To say I'm skeptical that replacement players on defense are generally more negative than on offense in the modern game is an understatement. To me those Nugget numbers are canaries in the coal mind that when we go based on an individual team, we're getting something unique based on them regardless of whether they are good or bad.

[Doctor MJ]

If I'm reading this right, which I'm probably not, Kidd looks too high on offense and David Robinson too low. Those are just the first two I noticed.

Oh I wouldn't say you're reading it wrong. You're clearly seeing Kidd rank above Robinson in VORP, and Offensive VORP, by quite a lot.

My thoughts:

a) Keep in mind that this metric projects based on these guys' PBP Era ('96-97 onward) career back to their rookie campaign. For Robinson, I'd expect this is going to cause his Offensive VORP to underrate him, and quite possibly by a good bit. For Kidd, much less is projected but I'd note that I expect his projection overrates him. Those first two seasons in Dallas were rough.

b) Remember that the metric I'm highlighting here - VORP - is going to accumulate over a longer career. Kidd played a lot more minutes than Robinson.

Worth noting that if you look at the original data, which was RAPM, Kidd does have a lead still, but very slight (5.0 vs 4.8). I would actually expect that if we did have full career PBP data for these guys, Robinson would come out ahead on RAPM...but honestly, I'm not so sure about VORP. Kidd played a lot.