Offensive GOAT - Michael Jordan Playoff ON-OFF

[Redmoon]

Don't see the usual Jordan detractors anywhere to be found. Wonder what the game are thinking when they see this thread.

Don't worry they're planning their coordinated counter attack

[RCM88x]

Jordan not being considered a clear Tier 1 offensive GOAT has always been surprising to me. How else would he be in GOAT peak-prime talks if he wasn't? It certainly wouldn't be for his defense (not that he's bad but he's not in the same stratosphere as a KG or Russell).

[jjgp111292]

Jordan not being considered a clear Tier 1 offensive GOAT has always been surprising to me. How else would he be in GOAT peak-prime talks if he wasn't? It certainly wouldn't be for his defense (not that he's bad but he's not in the same stratosphere as a KG or Russell).

Well let's take 1991 for instance. Oh sure the BPMs and PERs of the world will tell you MJ was brilliant but according to my carefully crafted eye test 1991 he was clearly getting worse. Once that Jumper's Knee started flaring up in 1990, he just wasn't creating much any more and frankly, his defense had become an outright negative. The box score will tell you he killed Detroit but that was only because of the free throw blitz at the end of game 2. And Barkley shot 10 points better than him in the second round. Ten. Points. I honestly have a mind to rank him above MJ that year.

Despite what every single person for the last 30 years has said, this was the year MJ stopped being "MJ", yet the Bulls started being champions. ☝️ ☝️ Now why is that?

This message has been sponsored by DIscord.

[lessthanjake]

Jordan not being considered a clear Tier 1 offensive GOAT has always been surprising to me. How else would he be in GOAT peak-prime talks if he wasn't? It certainly wouldn't be for his defense (not that he's bad but he's not in the same stratosphere as a KG or Russell).

Well let's take 1991 for instance. Oh sure the BPMs and PERs of the world will tell you MJ was brilliant but according to my carefully crafted eye test 1991 he was clearly getting worse. Once that Jumper's Knee started flaring up in 1990, he just wasn't creating much any more and frankly, his defense had become an outright negative. The box score will tell you he killed Detroit but that was only because of the free throw blitz at the end of game 2. And Barkley shot 10 points better than him in the second round. Ten. Points. I honestly have a mind to rank him above MJ that year.

Despite what every single person for the last 30 years has said, this was the year MJ stopped being "MJ", yet the Bulls started being champions. ☝️ ☝️ Now why is that?

This message has been sponsored by DIscord.

I once posted a similar parody that I’m pretty proud of:

Spoiler:

Well I for one am now thinking about ranking Jordan outside of the top 100. You see, in my large-sample multi-year extrap, I’ve found that the Bulls won 57 games with Jordan in 1993, and then he got replaced by a below-replacement-level Pete Myers and the team added a rookie Toni Kukoc, and they won 55 games. We can assume that if Pete Myers had just been replacement level, the Bulls would’ve been able to maintain the same record as in 1993—after all, they were very close as it was. And would’ve maintained the same SRS too if they’d been as healthy as in 1993 (fully-healthy SRS in 1994 was already close!). So that has Jordan’s value above replacement level in 1993 as being roughly the same as the value that Toni Kukoc coming off the bench in his rookie season brought—since the effects of Jordan leaving and Kukoc joining seems to have canceled out. Meanwhile, if we try to measure Jordan’s value at his peak in 1991, we see that, without Jordan, the 2015 Hawks won just 1 fewer game than the 1991 Bulls with 3.8 SRS lower. So this large-sample multi-year extrap shows that Jordan only provided about a 3.8 SRS lift in 1991. And Jordan had a better team than the 2015 Hawks—the Hawks didn’t have Pippen or Grant, both better than anyone on those Hawks!—so that 3.8 SRS lift is really the upper bound. And that’s at Jordan’s peak year! Obviously, our other extrap shows that, by 1993, he was only worth as much as a rookie Toni Kukoc. Meanwhile, we know that in the second three-peat, Jordan was past his peak, so we can assume that his value/impact only went down from there. So what these extraps show is a guy who, at his very peak, could provide at most a 3.8 SRS lift (probably less!) and then quickly became only as impactful as a rookie Toni Kukoc, before retiring and then coming back in 1995-1998 in further diminished form. I don’t think this is an impact profile I can put in the top 100. The bottom of the top 100 includes guys like Carmelo Anthony. Carmelo even in his rookie year showed more SRS lift than Jordan ever did, lifting his team from -7.41 SRS the year before he got there, to 1.65 SRS his rookie year. That 9.06 SRS lift is almost 2.5x peak Jordan! And that was Carmelo’s rookie year at age 19, not even close to his peak! And Carmelo also has the longevity edge on Jordan. When large-sample multi-year extraps massively favor rookie Carmelo over peak Jordan AND Carmelo has the longevity edge, there’s really no argument for Jordan over Carmelo. So I don’t see how Jordan can be top 100. In fact, given that almost-peak Jordan was equivalent in impact to rookie Toni Kukoc according to the extraps, I think we’re looking at Michael being roughly a Kukoc-level player at best, with who I’d put ahead depending on whether we think Jordan’s drop from his 1991 peak to 1993 was more or less than Kukoc’s improvement from his rookie year to his peak. I’d have to run some more extraps to figure that out.

[Djoker]

Sure! Here’s what I’ve got for Nash’s on-court rORTG by series, using PBPstats data:

2001 vs. UTA: +8.57
2001 vs. SAS: -1.84
2002 vs. MIN: +10.33
2002 vs. SAC: +9.73
2003 vs. POR: +11.41
2003 vs. SAC: +16.69
2003 vs. SAS: +4.65
2004 vs. SAC: -2.53
2005 vs. MEM: +22.40
2005 vs. DAL: +14.71
2005 vs. SAS: +17.16
2006 vs. LAL: +10.68
2006 vs. LAC: +15.05
2006 vs. DAL: +7.98
2007 vs. LAL: +6.72
2007 vs. SAS: +8.57
2008 vs. SAS: +4.09
2010 vs. POR: +14.28
2010 vs. SAS: +13.02
2010 vs. LAL: +17.75

It is perhaps a good point that the Suns actually lost series where they had a +17.75 and +17.16 rORTG with Nash on the court. They definitely were an offensively slanted roster.

But I do think that a lot of that was more about the players being bad defensively rather than being incredible offensively. I don’t really look at the 2005-2010 Suns rosters and feel totally bowled over by the offensive talent. Amare was very good, but it’s not like he’s as good offensively as a guy like Kareem or Wade or plenty of other #2 guys on a team. Marion was very limited offensively (and probably was more of a defensively slanted player actually). On the 2010 team, Jason Richardson was definitely offensively slanted, but more because of bad defense than actually being a great offensive player. And they had various 3&D guys like Raja Bell, James Jones, and Jared Dudley, who were fine pieces whose shooting fit well with Nash but they were definitely not uniquely good offensive players. And then you have Nash himself—who was very bad defensively, which contributes to the offensive slant of the team without actually indicating anything about his supporting cast’s offensive talent. So, I guess my point is that a team can be more offensively slanted than another team without actually having a more offensively talented supporting cast—it just requires being worse on defense. And that’s largely how I see that Suns team, when compared to plenty of other teams that other all-time-great offensive players have had. I think plenty of all-time-greats have actually had more offensively talented rosters than Nash did.

The most persuasive argument to me about those Suns is that, even if their supporting cast’s offensive talent really wasn’t more than the talent on other great players’ teams, the *lineups* they put out were offensively slanted, in the sense of running a lot of small Amare-at-center, Tim-Thomas-at-center, and Channing-Frye-at-center lineups. There’s definitely truth to that, and I think it surely contributed. Indeed, in 2007 and 2008, they played Kurt Thomas (an actual traditional defense-first center) a fair bit in 2007 and then got Shaq in 2008, so they had real centers. And the playoff on-court rORTGs for Nash in those years were definitely more down to earth. While he was injured in the 2006 playoffs, Nash played with Kurt Thomas for 53 regular season games in 2006, and Nash’s rORTG in those games was +4.69. So I think we do have some indication that Nash’s incredible offenses on the Suns may have been a product of not playing a traditional center (though they still look very good overall with a real center). I also think there’s no doubt that Nash had a really offensively talented team on the Mavs when he was playing alongside Dirk. So I can definitely buy deflating Nash’s numbers some in our mind on the basis of this. Which, given what you’ve shown in this thread, might be to Jordan’s benefit when it comes to being offensive GOAT.

_____________

I’ll note that these lineup factors can be applied to other players as well. Steph spent many years with his team playing a traditional center *and* Draymond at the PF position. The lineups tended to veer smaller in the years with Durant (which also were the highest rORTG years for him in the playoffs), but he has a lot of years in which his team played lineups very geared towards defense, in the sense of playing two defense-first big men. Jordan’s second-three-peat Bulls were similar, often playing a traditional center alongside Dennis Rodman (though they did occasionally go small with Kukoc and Rodman at PF and C, and the results were generally incredible). In Jordan’s earlier years, they did this with guys like Corzine and Oakley and then later Grant and Cartwright, though I wouldn’t say having two bruising big men was particularly abnormal back then, and prime Grant at least wasn’t bad offensively (though he was a defense-slanted player, for sure). Overall, in the context of his era, I wouldn’t say lineup factors materially disadvantaged Jordan, but they definitely didn’t operate to his advantage either.

In contrast, LeBron’s teams kind of pioneered running very offense-slanted lineups, after he was unable to pilot consistently great playoff offenses without such lineups. He tended to have pretty traditional lineups in his first stint in Cleveland, and the rORTGs were not very good overall in those years (he averaged something like a +2.5 on-court rORTG in those years). But then he went to Miami. From the beginning, they really experimented with offensively slanted lineups, putting Bosh at center. They only did it some in 2011, and the playoff offenses were not particularly impressive that year. So they leaned into it even more afterwards, putting Bosh virtually exclusively at center. Predictably, the offensive results became better. He then went to Cleveland. In the 2015 playoffs, due to injuries they ended up with a real traditional frontcourt with Mozgov and Tristan Thompson. The offensive results were not great. After that, Love + Thompson was a bit offensively slanted (since Thompson was more of a PF than a C), but not overly so. However, between running Love sometimes at center (and most of the time by 2018), and taking a page out of the Nash Suns book by putting Channing Frye at center, they started running a lot of very offensively slanted lineups. The results, again, were much better offensively. LeBron then goes to Los Angeles. They did play AD at center some with a smaller PF, but there was a good bit of AD at PF with a traditional center (generally Dwight or McGee). And, again, the offensive results were good but not particularly great (he averaged something like a +6 on-court rORTG in those 2020 playoffs).

So yeah, I think if we look at lineup factors, it should tend to make us curve up Jordan and Steph and/or curve down LeBron and Nash, since the latter two benefited a good deal from their teams playing innovative, offensively-slanted lineups and their on-court rORTGs in years where they did not have that tended to be noticeably lower.

Thanks for posting. Also a fantastic breakdown of the lineups that I pretty much find all agreeable as well.

Just want to add that the offensive slant of Nash's team is way more pronounced. For instance, in the 2005 playoffs, the team had a +16.2 rORtg but +5.7 on rDRtg. The latter number is really really bad over a sample size of 3 series because they made the WCF. In 2006, a +8.9 rORtg and +4.2 rDRtg over 3 series. In 2007, a +6.7 rORtg and -3.6 rDRtg over 2 series. Nice for a change. In 2008, a +2.3 rORtg and -1.1 rDRtg in 1 series. In 2009, missed the playoffs with a great offense but horrendous defense. In 2010, a +12.6 rORtg and +2.9 rORtg over 3 series. From 2005-2010, the Suns posted a horrible average of about +2.5 rDRtg over 12 series. Obviously those aren't numbers with just Nash on the court but I imagine they are similar.

[lessthanjake]

Thanks for posting. Also a fantastic breakdown of the lineups that I pretty much find all agreeable as well.

Just want to add that the offensive slant of Nash's team is way more pronounced. For instance, in the 2005 playoffs, the team had a +16.2 rORtg but +5.7 on rDRtg. The latter number is really really bad over a sample size of 3 series because they made the WCF. In 2006, a +8.9 rORtg and +4.2 rDRtg over 3 series. In 2007, a +6.7 rORtg and -3.6 rDRtg over 2 series. Nice for a change. In 2008, a +2.3 rORtg and -1.1 rDRtg in 1 series. In 2009, missed the playoffs with a great offense but horrendous defense. In 2010, a +12.6 rORtg and +2.9 rORtg over 3 series. From 2005-2010, the Suns posted a horrible average of about +2.5 rDRtg over 12 series. Obviously those aren't numbers with just Nash on the court but I imagine they are similar.

Yeah, the Nash Suns were definitely offensively slanted. I’d just note a couple things that go to the fact that a team having a large offensive slant doesn’t necessarily mean that the supporting cast was uniquely good offensively:

1. A significant portion of that slant is a result of Nash himself being incredible offensively and awful defensively. Nash creates a significant slant, independent of his supporting cast.

2. His supporting cast was in significant part just bad defensively, rather than being uniquely great offensively. Of particular note, Amare was a genuinely terrible defensive player, but not actually some uniquely great offensive player (and quickly perusing his RAPM values on NBArapm bears that out). I think he was simultaneously a more offensively-slanted player than the vast majority of #2 players on top-tier teams, while not actually being particularly better offensively. (Which logically suggests he wasn’t as good a player overall as most other #2 players on top-tier teams, but I think that’s accurate).

As previously discussed though, I think it’s appropriate to recognize lineup effects here and curve down Nash’s offensively numbers some because of that. To take the Amare example: However good on offense and bad on defense he was, those two things got exaggerated even further by him playing at C instead of PF. Similarly, while Marion was not an offensively slanted player in general, he might well have been a bit offensively slanted while slotted in at PF.

I have long been a Nash-is-offensive-GOAT guy, but the information in this thread does have me wavering on this. While Nash’s case is at least in part an eye-test thing for me, his case really does hinge a lot on ridiculous team offensive results. Those team numbers are undoubtedly partially a product of playing really offense-first lineups, but his team results were so far in the stratosphere that it still feels like really good evidence. But if Jordan actually achieved somewhat close team offensive results when he was on the court, then Nash’s argument against Jordan actually gets pretty dicey, because it’s hard to argue in this comparison that Nash’s team numbers don’t need to be curved down a significant amount for lineup effects.

[tsherkin]

1. A significant portion of that slant is a result of Nash himself being incredible offensively and awful defensively. Nash creates a significant slant, independent of his supporting cast.

I think their 5-hole was a lot more of a problem than Nash himself, personally.

As you note in your second point, Amare was a waste of skin defensively and not a stunner on the boards, either.

Why were they bad defensively? Well, look at 2005.

Second-worst defensive rebounding team in the league, 3rd-worst at generating turnovers. Actually 10th in defensive eFG%. Same problems but marginally better the year after, and again in 07.

Their frontcourt impact was pretty brutal.

[Shanghai Kid]

Wait wait wait, are you suggesting that MJ might be the goat??

We won't have any of that talk around here mister! The player comparison board proved that wasn't the case long ago!

[tsherkin]

Wait wait wait, are you suggesting that MJ might be the goat??

We won't have any of that talk around here mister! The player comparison board proved that wasn't the case long ago!

I know you guys are joking around and such, but at some point, it's pretty clear that MJ's blend of efficiency and low turnovers with decent passing did very good things for his team's O in-era. And then even as his efficiency went down in the later rounds in the second three-peat, his ability not to turn the ball over at high usage was clearly quite valuable in that environment. I mean it always is, but I suspect it was that much more relevant when possession quantity was at such a premium. Paired with an insane offensive rebounder, the Bulls just crushed it with possession control even when MJ's shot wasn't going. That specific strategy wouldn't work quite as well forward twenty years, but it was pretty gloriously effective at the time.

[lessthanjake]

We’ve previously seen a lot of analysis regarding how players’ offenses fared specifically against great defensive teams and/or good teams in general.

On that front, I want to just use this Jordan data to make a comparison with LeBron this front. I hope to have the time to do this for Steph and Nash too, but this took me a while to compile and I’m not sure I’ll have the time to do the same for them anytime soon.

First, I’ll look at these players’ on-court rORTGs against opponents with various regular season rDRTG thresholds. Please note that these numbers are possession-weighted. For Jordan, that means taking his minutes played, dividing by 48, and multiplying by the series pace. For LeBron, we get possession numbers from PBPstats. Once again, I took Jordan’s entire career (i.e. 1985-1998) and compared to LeBron’s timeframe that is most comparable in terms of age (i.e. 2007-2020). LeBron’s numbers look worse if you add any more years.

Playoff On-Court rORTGs vs. Good Defenses: 1985-1998 Jordan & 2007-2020 LeBron

First, I’ll look at how these two players’ offenses fared with them on the court against good defenses. I will use every possible opponent rDRTG threshold we could use to define what is a good defense.

Playoff On-Court rORTG vs. -8.0 or better rDRTG Opponents

Jordan: +18.20
LeBron: +3.23

Playoff On-Court rORTG vs. -7.0 or better rDRTG Opponents

Jordan: +18.20
LeBron: +8.08

Playoff On-Court rORTG vs. -6.0 or better rDRTG Opponents

Jordan: +12.71
LeBron: +8.48

Playoff On-Court rORTG vs. -5.0 or better rDRTG Opponents

Jordan: +12.85
LeBron: +9.51

Playoff On-Court rORTG vs. -4.0 or better rDRTG Opponents

Jordan: +8.33
LeBron: +8.07

Playoff On-Court rORTG vs. -3.0 or better rDRTG Opponents

Jordan: +9.43
LeBron: +7.69

Playoff On-Court rORTG vs. -2.0 or better rDRTG Opponents

Jordan: +8.50
LeBron: +7.83

So Jordan’s playoff on-court rORTG was better than LeBron’s against good defenses, using any threshold we might use to define good defenses.

Playoff On-Court rORTG vs. Good Opponents: 1985-1998 Jordan & 2007-2020 LeBron

It is difficult to define what is a good opponent, and any formulation of this does run the risk of being overinclusive or underinclusive in some way. I will provide a couple versions of this below.

For the first one, I will just use 5 SRS as the threshold to define what is a good opponent, while adding any team that actually won the title that year. This would be a very standard way of looking at good opponents. However, that can be pretty underinclusive, particularly when it comes to teams that demonstrated they were really good by winning a title with the same core in a different year.

So the second version I will provide will include all 5+ SRS teams and any team that won a title with the same core, even if it was not that year.

Note that both of these versions are still only in the 2007-2020 timeframe for LeBron, so we are excluding several poor data points outside that timeframe, including a -4 on-court rORTG against the 2006 Pistons.

Playoff On-Court rORTG vs. 5+ SRS Opponents & That Year’s NBA Champion

Jordan: +7.08
LeBron: +6.68

Playoff On-Court rORTG vs. 5+ SRS Opponents & Teams that Won Title w/ Same Core

Jordan: +7.56
LeBron: +6.22 (or +6.06 if we include 2020 Nuggets)

I haven’t tested this across all other SRS thresholds, but to test out a higher SRS threshold, I did also look at series just against 7+ SRS opponents & teams that won the title that year. I looked at this specifically because it is narrowly tailored to keep in the 2018 Raptors data point while keeping out a bunch of really good Jordan numbers in the 6-7 opponent-SRS range and keeps out LeBron’s relatively poor numbers against teams that won the title in other years with the same core. Despite that, Jordan’s numbers are still higher:

Playoff On-Court rORTG vs. 7+ SRS Opponents & That Year’s NBA Champion

Jordan: +6.36
LeBron: +5.87

In case anyone is curious, here’s the same thing but with a 6+ SRS threshold. This still advantages LeBron by not including series against opponents below the SRS threshold that won the title in other years with the same core, but Jordan is significantly ahead:

Playoff On-Court rORTG vs. 6+ SRS Opponents & That Year’s NBA Champion

Jordan: +8.40
LeBron: +6.68

So yeah, while we could potentially look at other thresholds too, the data really really does indicate that Jordan’s playoff offenses were better against good teams with him on the court than LeBron’s offenses were.
____________________

I think the overall takeaway here is that, when Jordan and LeBron were on the court, Jordan’s playoff offenses were better against good teams and good defenses than LeBron’s playoff offenses were. This is a really strong data point for Jordan, particularly when we keep in mind the previously discussed fact that LeBron’s teams innovated with smaller, more offense-slanted lineups, while Jordan’s teams did not.

[jalengreen]

Wait wait wait, are you suggesting that MJ might be the goat??

We won't have any of that talk around here mister! The player comparison board proved that wasn't the case long ago!

I know you guys are joking around and such, but at some point, it's pretty clear that MJ's blend of efficiency and low turnovers with decent passing did very good things for his team's O in-era. And then even as his efficiency went down in the later rounds in the second three-peat, his ability not to turn the ball over at high usage was clearly quite valuable in that environment. I mean it always is, but I suspect it was that much more relevant when possession quantity was at such a premium. Paired with an insane offensive rebounder, the Bulls just crushed it with possession control even when MJ's shot wasn't going. That specific strategy wouldn't work quite as well forward twenty years, but it was pretty gloriously effective at the time.

I think this is an interesting discussion

For the sake of it, I would give the obvious counter which would be that the “possession game” is a big reason why the current NBA champions were as good as they are.

https://xrapm.com/table_pages/tf.html

The leader this year in “Team TO xRAPM” (so impact on minimizing their offense’s turnovers committed) was Shai Gilgeous-Alexander by a wide margin. The leader this year in “Opp TO xRAPM” (impact on maximizing their defense’s turnovers forced) was Alex Caruso by a wide margin. No doubt, possession control was key to Oklahoma City’s dominance.

Counter to this might be “sure, but OKC’s offense wasn’t as good as the Bulls’, and Shai did not produce Jordan’s statistical offensive impact markers”. And yeah that’s obviously pertinent, though I suppose the question is how much of that to attribute to Shai being inferior at basketball versus a difference in the times.

[tsherkin]

For the sake of it, I would give the obvious counter which would be that the “possession game” is a big reason why the current NBA champions were as good as they are.

I mean, possession control is always going to be valuable, I just meant that in the late 90s, it was that much MORE important because you had less efficient transition offense

Counter to this might be “sure, but OKC’s offense wasn’t as good as the Bulls’, and Shai did not produce Jordan’s statistical offensive impact markers”. And yeah that’s obviously pertinent, though I suppose the question is how much of that to attribute to Shai being inferior at basketball versus a difference in the times.

Strictly Shai versus Jordan is an interesting discussion. SGA is also a super low-turnover guy, even in the playoffs. And generally hyper-efficient. And then obviously still pretty good in the playoffs. And it's not like Old Jordan was crushing efficiency like a monster in the second three-peat, after all. He was actually worse than playoff league average in I believe all three of his second-three-peat titles during the Finals series. No, sorry, just versus Utah, but spot-on playoff league-average versus the Sonics. He was all about high volume/usage with low turnover percentage in those series, with crap efficiency because he was mostly shooting jumpers in stupid volume.

Now, as you start comparing him to pre-retirement Jordan and his impact relative to that league, even in the playoffs, the story changes some. Though less so against the Knicks.

[lessthanjake]

.

I am curious how you got your overall playoff rORTG averages for Jordan over different timespans. I calculated possession-weighted averages for Jordan and got some different numbers.

Just to be clear on my methodology, here’s what I did:

To estimate the number of possessions Jordan played in each series, I looked at your “Legends Playoff Plus Minus” spreadsheet, and divided the Minutes ON by 48 and then multiplied by the series Pace. I then used the on-court rORTG for each series that you reported in your OP, and did a possession-weighted average of them (i.e. I multiplied the rORTG in a series by the estimate of the number of possessions Jordan played in that series, added all those together for each series, and then divided by the total number of possessions across all the series).

Here’s what I got for various timeframes:

Jordan Playoff On-Court rORTG

- Full Career: +8.90
- Full Career minus 1995: +9.16
- 1989-1998: +9.49
- 1989-1998 minus 1995: +9.82
- Title Years Only: +10.28
- 1985-1990: +6.92

Since we’ve discussed these years in particular, I’ll also note that I have the 1991 playoffs at +14.41 overall, and the 1993 playoffs at +14.85 overall.

For reference, I used the same technique for LeBron, Steph, and Nash, except I got the number of possessions they played each series from PBPstats, and calculated their on-court rORTG in those series using PBPstats data for their on-court ORTG and the opponents’ regular-season DRTG. Here’s the numbers I got for them (I did not calculate across all possible timeframes—so this is just what I already have):

LeBron Playoff On-Court rORTG

- 2006-2020: +6.41
- 2007-2020: +8.06
- 2009-2018: +8.78
- 2006-2010: +2.50

Steph Playoff On-Court rORTG

- 2014-2023: +7.41

Nash Playoff On-Court rORTG

- 2001-2013: +10.39
- 2001-2010: +10.65
- 2005-2010: +12.58

____________

It looks to me like your numbers for Jordan are different, though. Are you using a different method to calculate Jordan’s averages? I don’t think I have any data input errors, but of course it’s definitely possible, so I could double-check that again. But figured I’d run by you to see your thoughts.

[Djoker]

.

I am curious how you got your overall playoff rORTG averages for Jordan over different timespans. I calculated possession-weighted averages for Jordan and got some different numbers.

Just to be clear on my methodology, here’s what I did:

To estimate the number of possessions Jordan played in each series, I looked at your “Legends Playoff Plus Minus” spreadsheet, and divided the Minutes ON by 48 and then multiplied by the series Pace. I then used the on-court rORTG for each series that you reported in your OP, and did a possession-weighted average of them (i.e. I multiplied the rORTG in a series by the estimate of the number of possessions Jordan played in that series, added all those together for each series, and then divided by the total number of possessions across all the series).

Here’s what I got for various timeframes:

Jordan Playoff On-Court rORTG

- Full Career: +8.90
- Full Career minus 1995: +9.16
- 1989-1998: +9.49
- 1989-1998 minus 1995: +9.82
- Title Years Only: +10.28
- 1985-1990: +6.92

Since we’ve discussed these years in particular, I’ll also note that I have the 1991 playoffs at +14.41 overall, and the 1993 playoffs at +14.85 overall.

For reference, I used the same technique for LeBron, Steph, and Nash, except I got the number of possessions they played each series from PBPstats, and calculated their on-court rORTG in those series using PBPstats data for their on-court ORTG and the opponents’ regular-season DRTG. Here’s the numbers I got for them (I did not calculate across all possible timeframes—so this is just what I already have):

LeBron Playoff On-Court rORTG

- 2006-2020: +6.41
- 2007-2020: +8.06
- 2009-2018: +8.78
- 2006-2010: +2.50

Steph Playoff On-Court rORTG

- 2014-2023: +7.41

Nash Playoff On-Court rORTG

- 2001-2013: +10.39
- 2001-2010: +10.65
- 2005-2010: +12.58

____________

It looks to me like your numbers for Jordan are different, though. Are you using a different method to calculate Jordan’s averages? I don’t think I have any data input errors, but of course it’s definitely possible, so I could double-check that again. But figured I’d run by you to see your thoughts.

Why would you do a possession-weighted average when you can calculate directly?

ON Net = Raw ON * 48/Minutes ON * 100/Pace
ON ORtg = ON Team * 48/Minutes ON * 100/Pace
ON DRtg = ON Opp * 48/Minutes ON * 100/Pace
OFF Net = Raw OFF * 48/Minutes OFF * 100/Pace
OFF ORtg = OFF Team * 48/Minutes OFF * 100/Pace
OFF DRtg = OFF Opp * 48/Minutes OFF * 100/Pace

So for example, for career ON Net Rtg:

+1057 * 48/7476 * 100/88.7 = +7.7

[lessthanjake]

.

I am curious how you got your overall playoff rORTG averages for Jordan over different timespans. I calculated possession-weighted averages for Jordan and got some different numbers.

Just to be clear on my methodology, here’s what I did:

To estimate the number of possessions Jordan played in each series, I looked at your “Legends Playoff Plus Minus” spreadsheet, and divided the Minutes ON by 48 and then multiplied by the series Pace. I then used the on-court rORTG for each series that you reported in your OP, and did a possession-weighted average of them (i.e. I multiplied the rORTG in a series by the estimate of the number of possessions Jordan played in that series, added all those together for each series, and then divided by the total number of possessions across all the series).

Here’s what I got for various timeframes:

Jordan Playoff On-Court rORTG

- Full Career: +8.90
- Full Career minus 1995: +9.16
- 1989-1998: +9.49
- 1989-1998 minus 1995: +9.82
- Title Years Only: +10.28
- 1985-1990: +6.92

Since we’ve discussed these years in particular, I’ll also note that I have the 1991 playoffs at +14.41 overall, and the 1993 playoffs at +14.85 overall.

For reference, I used the same technique for LeBron, Steph, and Nash, except I got the number of possessions they played each series from PBPstats, and calculated their on-court rORTG in those series using PBPstats data for their on-court ORTG and the opponents’ regular-season DRTG. Here’s the numbers I got for them (I did not calculate across all possible timeframes—so this is just what I already have):

LeBron Playoff On-Court rORTG

- 2006-2020: +6.41
- 2007-2020: +8.06
- 2009-2018: +8.78
- 2006-2010: +2.50

Steph Playoff On-Court rORTG

- 2014-2023: +7.41

Nash Playoff On-Court rORTG

- 2001-2013: +10.39
- 2001-2010: +10.65
- 2005-2010: +12.58

____________

It looks to me like your numbers for Jordan are different, though. Are you using a different method to calculate Jordan’s averages? I don’t think I have any data input errors, but of course it’s definitely possible, so I could double-check that again. But figured I’d run by you to see your thoughts.

Why would you do a possession-weighted average when you can calculate directly?

ON Net = Raw ON * 48/Minutes ON * 100/Pace
ON ORtg = ON Team * 48/Minutes ON * 100/Pace
ON DRtg = ON Opp * 48/Minutes ON * 100/Pace
OFF Net = Raw OFF * 48/Minutes OFF * 100/Pace
OFF ORtg = OFF Team * 48/Minutes OFF * 100/Pace
OFF DRtg = OFF Team * 48/Minutes OFF * 100/Pace

So for example, for career ON Net Rtg:

+1057 * 48/7476 * 100/88.7 = +7.7

Yeah, I’m on the same page on all that. Calculating net rating definitely doesn’t require a possession-weighted average. I’m talking about calculating overall rORTG across multiple series/years, which I think does require doing a weighted average because each series has its own independent rORTG number due to each opponent having a different regular-season DRTG. For instance, let’s say a player has a +10 rORTG in a series he played 200 possession in and a +5 rORTG in a series he played 400 possessions in, and we want to calculate the rORTG over the course of those two series. I think the way to calculate that would be to do 10*(200/600)+5*(400/600).

[Djoker]

I am curious how you got your overall playoff rORTG averages for Jordan over different timespans. I calculated possession-weighted averages for Jordan and got some different numbers.

Just to be clear on my methodology, here’s what I did:

To estimate the number of possessions Jordan played in each series, I looked at your “Legends Playoff Plus Minus” spreadsheet, and divided the Minutes ON by 48 and then multiplied by the series Pace. I then used the on-court rORTG for each series that you reported in your OP, and did a possession-weighted average of them (i.e. I multiplied the rORTG in a series by the estimate of the number of possessions Jordan played in that series, added all those together for each series, and then divided by the total number of possessions across all the series).

Here’s what I got for various timeframes:

Jordan Playoff On-Court rORTG

- Full Career: +8.90
- Full Career minus 1995: +9.16
- 1989-1998: +9.49
- 1989-1998 minus 1995: +9.82
- Title Years Only: +10.28
- 1985-1990: +6.92

Since we’ve discussed these years in particular, I’ll also note that I have the 1991 playoffs at +14.41 overall, and the 1993 playoffs at +14.85 overall.

For reference, I used the same technique for LeBron, Steph, and Nash, except I got the number of possessions they played each series from PBPstats, and calculated their on-court rORTG in those series using PBPstats data for their on-court ORTG and the opponents’ regular-season DRTG. Here’s the numbers I got for them (I did not calculate across all possible timeframes—so this is just what I already have):

LeBron Playoff On-Court rORTG

- 2006-2020: +6.41
- 2007-2020: +8.06
- 2009-2018: +8.78
- 2006-2010: +2.50

Steph Playoff On-Court rORTG

- 2014-2023: +7.41

Nash Playoff On-Court rORTG

- 2001-2013: +10.39
- 2001-2010: +10.65
- 2005-2010: +12.58

____________

It looks to me like your numbers for Jordan are different, though. Are you using a different method to calculate Jordan’s averages? I don’t think I have any data input errors, but of course it’s definitely possible, so I could double-check that again. But figured I’d run by you to see your thoughts.

Why would you do a possession-weighted average when you can calculate directly?

ON Net = Raw ON * 48/Minutes ON * 100/Pace
ON ORtg = ON Team * 48/Minutes ON * 100/Pace
ON DRtg = ON Opp * 48/Minutes ON * 100/Pace
OFF Net = Raw OFF * 48/Minutes OFF * 100/Pace
OFF ORtg = OFF Team * 48/Minutes OFF * 100/Pace
OFF DRtg = OFF Team * 48/Minutes OFF * 100/Pace

So for example, for career ON Net Rtg:

+1057 * 48/7476 * 100/88.7 = +7.7

Yeah, I’m on the same page on all that. Calculating net rating definitely doesn’t require a possession-weighted average. I’m talking about calculating overall rORTG across multiple series/years, which I think does require doing a weighted average because each series has its own independent rORTG number due to each opponent having a different regular-season DRTG. For instance, let’s say a player has a +10 rORTG in a series he played 200 possession in and a +5 rORTG in a series he played 400 possessions in, and we want to calculate the rORTG over the course of those two series. I think the way to calculate that would be to do 10*(200/600)+5*(400/600).

Oh.. you're right but In that case why not just calculate the ORtg for those two series and then find the rORtg by just subtracting the average DRtg of those two teams.

For instance if I want to get the average rORtg for the 1993 Knicks and 1996 Sonics series.

(596 + 505) * 48/(274 + 252) * 100/((6 * 89.7 + 6 * 83.5)/12) = 116.0 ORtg (note average pace is a weighted average by games)

Average Opp DRtg = (99.7 + 102.1)/2 = 100.9 DRtg

rORtg = 116.0 - 100.9 = +15.1 rORtg

Although I think your method should still get the same result. Maybe minor rounding errors...

[lessthanjake]

Why would you do a possession-weighted average when you can calculate directly?

ON Net = Raw ON * 48/Minutes ON * 100/Pace
ON ORtg = ON Team * 48/Minutes ON * 100/Pace
ON DRtg = ON Opp * 48/Minutes ON * 100/Pace
OFF Net = Raw OFF * 48/Minutes OFF * 100/Pace
OFF ORtg = OFF Team * 48/Minutes OFF * 100/Pace
OFF DRtg = OFF Team * 48/Minutes OFF * 100/Pace

So for example, for career ON Net Rtg:

+1057 * 48/7476 * 100/88.7 = +7.7

Yeah, I’m on the same page on all that. Calculating net rating definitely doesn’t require a possession-weighted average. I’m talking about calculating overall rORTG across multiple series/years, which I think does require doing a weighted average because each series has its own independent rORTG number due to each opponent having a different regular-season DRTG. For instance, let’s say a player has a +10 rORTG in a series he played 200 possession in and a +5 rORTG in a series he played 400 possessions in, and we want to calculate the rORTG over the course of those two series. I think the way to calculate that would be to do 10*(200/600)+5*(400/600).

Oh.. you're right but In that case why not just calculate the ORtg for those two series and then find the rORtg by just subtracting the average DRtg of those two teams.

For instance if I want to get the average rORtg for the 1993 Knicks and 1996 Sonics series.

(596 + 505) * 48/(274 + 252) * 100/((6 * 89.7 + 6 * 83.5)/12) = 116.0 ORtg (note average pace is a weighted average by games)

Average Opp DRtg = (99.7 + 102.1)/2 = 100.9 DRtg

rORtg = 116.0 - 100.9 = +15.1 rORtg

Although I think your method should still get the same result. Maybe minor rounding errors...

I think what I’m doing is similar but subtly different. There’s a couple differences I have with that method:

1. Pace is not actually number of possessions per game, but rather is number of possessions per 48 minutes. So if a game goes into OT, there will be more possessions than the pace for that game. Therefore, when you multiply the pace by the number of games, you won’t get the right possession number if there were overtimes.

2. Averaging the opponents’ DRTGs doesn’t account for the fact that the player isn’t playing the same number of possessions against each defense. Which can skew the overall number. For instance, let’s say a player plays 400 possessions against a -5 rDRTG team, and 200 possessions against a +5 rDRTG team. Averaging the opponents’ DRTGs would come out to a totally neutral opponent DRTG. But actually the player played two-thirds of his possessions against the good defense. So, merely averaging the opponents’ DRTGs would undersell the average quality of defense faced in those possessions. This is why I think one should do a possession-weighted average. This way, you’re not accidentally skewing the numbers if the player played different numbers of possessions against teams with different-quality defenses.

To take the example of the 1993 Knicks series and the 1996 Sonics series, here’s what I’d do:

We calculate the number of possessions Jordan played in each series separately, by taking the minutes Jordan played, dividing by 48, and multiplying by the pace of the series. For the 1993 Knicks series, that’d be (249/48)*86.3 = 447.68. For the 1996 Sonics series, that’d be (252/48)*83.5 = 438.38. Now we figure out the rORTG for each series separately, by dividing the number of points scored with Jordan ON by the number of possessions and multiplying that by 100, and then subtracting the opposing team’s regular-season DRTG. For the 1993 Knicks series, that’d be 100*(528/447.68)-99.7=18.24. For the 1996 Sonics series, that’d be 100*(505/438.38)-102.1=13.10. Now that we have the rORTG for both series, we take a weighted average of those rORTGs, weighted by the percent of total possessions Jordan faced each team. Since we estimated 447.68 possessions in the Knicks series and 438.38 possessions in the Sonics series, the total possession estimate is 886.06. Therefore, the rORTG across those two series would be 18.24*(447.68/886.06)+13.10*(438.38/886.06)=15.70.

You could also get the same result by calculating the possessions and then taking the total on-court ORTG and subtracting a possession-weighted average of the opponents’ DRTGs. In that case the math would be 100*(528+505)/(886.06)-(99.7*(447.68/886.06)+102.1*(438.38/886.06)=15.70. Those two versions are functionally identical and get to the same result.

Anyways, that 15.70 number I got is similar to your number, since there was no overtime in those series, and the number of possessions in each series is also extremely similar. In some instances, there would be a larger difference between the two methods, because of the existence of OTs and because some series have very different possession numbers from each other along with different quality opposing defenses.

EDIT: I note that your calculations above used the Jordan ON minutes and pace numbers from the 1993 Finals against the Suns but used the 1993 Knicks DRTG. So I believe that particular +15.1 number you got is a result of a data input error. I think your method would still get a *very* slightly lower number than mine though, even with the correct data put in for the 1993 Knicks (it’d be like +15.68 instead of +15.70). This is because the possession number for the Knicks series is very slightly higher and their defense was a little bit better. As noted, though, the output would diverge a lot more if we had OTs and/or series that didn’t have very similar possession numbers.

_______

FWIW, as per the couple posts I’ve made on this page that listed rORTGs across series/years (one post about rORTGs against good defenses and good teams, and the other post about rORTGs across specific timeframes), using the above-described method Jordan’s on-court rORTGs come out at a tier above others except Nash (who we’ve discussed lineup issues regarding). I find that data to be very compelling. And I just wanted to note the difference in the numbers we came to, since the numbers I got are actually a bit better for Jordan than the multi-year rORTG averages listed in your OP.

[Djoker]

I think what I’m doing is similar but subtly different. There’s a couple differences I have with that method:

1. Pace is not actually number of possessions per game, but rather is number of possessions per 48 minutes. So if a game goes into OT, there will be more possessions than the pace for that game. Therefore, when you multiply the pace by the number of games, you won’t get the right possession number if there were overtimes.

2. Averaging the opponents’ DRTGs doesn’t account for the fact that the player isn’t playing the same number of possessions against each defense. Which can skew the overall number. For instance, let’s say a player plays 400 possessions against a -5 rDRTG team, and 200 possessions against a +5 rDRTG team. Averaging the opponents’ DRTGs would come out to a totally neutral opponent DRTG. But actually the player played two-thirds of his possessions against the good defense. So, merely averaging the opponents’ DRTGs would undersell the average quality of defense faced in those possessions. This is why I think one should do a possession-weighted average. This way, you’re not accidentally skewing the numbers if the player played different numbers of possessions against teams with different-quality defenses.

To take the example of the 1993 Knicks series and the 1996 Sonics series, here’s what I’d do:

We calculate the number of possessions Jordan played in each series separately, by taking the minutes Jordan played, dividing by 48, and multiplying by the pace of the series. For the 1993 Knicks series, that’d be (249/48)*86.3 = 447.68. For the 1996 Sonics series, that’d be (252/48)*83.5 = 438.38. Now we figure out the rORTG for each series separately, by dividing the number of points scored with Jordan ON by the number of possessions and multiplying that by 100, and then subtracting the opposing team’s regular-season DRTG. For the 1993 Knicks series, that’d be 100*(528/447.68)-99.7=18.24. For the 1996 Sonics series, that’d be 100*(505/438.38)-102.1=13.10. Now that we have the rORTG for both series, we take a weighted average of those rORTGs, weighted by the percent of total possessions Jordan faced each team. Since we estimated 447.68 possessions in the Knicks series and 438.38 possessions in the Sonics series, the total possession estimate is 886.06. Therefore, the rORTG across those two series would be 18.24*(447.68/886.06)+13.10*(438.38/886.06)=15.70.

You could also get the same result by calculating the possessions and then taking the total on-court ORTG and subtracting a possession-weighted average of the opponents’ DRTGs. In that case the math would be 100*(528+505)/(886.06)-(99.7*(447.68/886.06)+102.1*(438.38/886.06)=15.70. Those two versions are functionally identical and get to the same result.

Anyways, that 15.70 number I got is similar to your number, since there was no overtime in those series, and the number of possessions in each series is also extremely similar. In some instances, there would be a larger difference between the two methods, because of the existence of OTs and because some series have very different possession numbers from each other along with different quality opposing defenses.

EDIT: I note that your calculations above used the Jordan ON minutes and pace numbers from the 1993 Finals against the Suns but used the 1993 Knicks DRTG. So I believe that particular +15.1 number you got is a result of a data input error. I think your method would still get a *very* slightly lower number than mine though, even with the correct data put in for the 1993 Knicks (it’d be like +15.68 instead of +15.70). This is because the possession number for the Knicks series is very slightly higher and their defense was a little bit better. As noted, though, the output would diverge a lot more if we had OTs and/or series that didn’t have very similar possession numbers.

_______

FWIW, as per the couple posts I’ve made on this page that listed rORTGs across series/years (one post about rORTGs against good defenses and goods teams, and the other post about rORTGs across specific timeframes), using the above-described method Jordan’s on-court rORTGs come out at a tier above others except Nash (who we’ve discussed lineup issues regarding). I find that data to be very compelling. And I just wanted to note the difference in the numbers we came to, since the numbers I got are actually a bit better for Jordan than the multi-year rORTG averages listed in your OP.

Ok I actually understand what you mean now.

1. Yes that's correct. OT periods are taken into account with the minutes played but when averaging pace by games played, it can be a small issue.

2. My method is simply average rORtg per series. Yours is average rORtg weighted by possession in those series. Small distinction but nevertheless an important one. Your method is more accurate technically speaking but I'm still unsure what is more informative. Maybe we do actually want just a simple average of all the different series. Because your possession based method gives a single 7-game series almost twice the weight of a 4-game sweep. I'm not sure that's actually a good thing. Maybe each series should be a separate but equivalent entry?

And the results we got are quite different... 1 to 1.3 points difference in rORtg is actually significant. Food for thought.

[lessthanjake]

I think what I’m doing is similar but subtly different. There’s a couple differences I have with that method:

1. Pace is not actually number of possessions per game, but rather is number of possessions per 48 minutes. So if a game goes into OT, there will be more possessions than the pace for that game. Therefore, when you multiply the pace by the number of games, you won’t get the right possession number if there were overtimes.

2. Averaging the opponents’ DRTGs doesn’t account for the fact that the player isn’t playing the same number of possessions against each defense. Which can skew the overall number. For instance, let’s say a player plays 400 possessions against a -5 rDRTG team, and 200 possessions against a +5 rDRTG team. Averaging the opponents’ DRTGs would come out to a totally neutral opponent DRTG. But actually the player played two-thirds of his possessions against the good defense. So, merely averaging the opponents’ DRTGs would undersell the average quality of defense faced in those possessions. This is why I think one should do a possession-weighted average. This way, you’re not accidentally skewing the numbers if the player played different numbers of possessions against teams with different-quality defenses.

To take the example of the 1993 Knicks series and the 1996 Sonics series, here’s what I’d do:

We calculate the number of possessions Jordan played in each series separately, by taking the minutes Jordan played, dividing by 48, and multiplying by the pace of the series. For the 1993 Knicks series, that’d be (249/48)*86.3 = 447.68. For the 1996 Sonics series, that’d be (252/48)*83.5 = 438.38. Now we figure out the rORTG for each series separately, by dividing the number of points scored with Jordan ON by the number of possessions and multiplying that by 100, and then subtracting the opposing team’s regular-season DRTG. For the 1993 Knicks series, that’d be 100*(528/447.68)-99.7=18.24. For the 1996 Sonics series, that’d be 100*(505/438.38)-102.1=13.10. Now that we have the rORTG for both series, we take a weighted average of those rORTGs, weighted by the percent of total possessions Jordan faced each team. Since we estimated 447.68 possessions in the Knicks series and 438.38 possessions in the Sonics series, the total possession estimate is 886.06. Therefore, the rORTG across those two series would be 18.24*(447.68/886.06)+13.10*(438.38/886.06)=15.70.

You could also get the same result by calculating the possessions and then taking the total on-court ORTG and subtracting a possession-weighted average of the opponents’ DRTGs. In that case the math would be 100*(528+505)/(886.06)-(99.7*(447.68/886.06)+102.1*(438.38/886.06)=15.70. Those two versions are functionally identical and get to the same result.

Anyways, that 15.70 number I got is similar to your number, since there was no overtime in those series, and the number of possessions in each series is also extremely similar. In some instances, there would be a larger difference between the two methods, because of the existence of OTs and because some series have very different possession numbers from each other along with different quality opposing defenses.

EDIT: I note that your calculations above used the Jordan ON minutes and pace numbers from the 1993 Finals against the Suns but used the 1993 Knicks DRTG. So I believe that particular +15.1 number you got is a result of a data input error. I think your method would still get a *very* slightly lower number than mine though, even with the correct data put in for the 1993 Knicks (it’d be like +15.68 instead of +15.70). This is because the possession number for the Knicks series is very slightly higher and their defense was a little bit better. As noted, though, the output would diverge a lot more if we had OTs and/or series that didn’t have very similar possession numbers.

_______

FWIW, as per the couple posts I’ve made on this page that listed rORTGs across series/years (one post about rORTGs against good defenses and goods teams, and the other post about rORTGs across specific timeframes), using the above-described method Jordan’s on-court rORTGs come out at a tier above others except Nash (who we’ve discussed lineup issues regarding). I find that data to be very compelling. And I just wanted to note the difference in the numbers we came to, since the numbers I got are actually a bit better for Jordan than the multi-year rORTG averages listed in your OP.

1. Yes that's correct. OT periods are taken into account with the minutes played but when averaging pace by games played, it can be a small issue.

2. My method is simply average rORtg per series. Yours is average rORtg weighted by possession in those series. Small distinction but nevertheless an important one. Your method is more accurate technically speaking but I'm still unsure what is more informative. Maybe we do actually want just a simple average of all the different series. Because your possession based method gives a single 7-game series almost twice the weight of a 4-game sweep. I'm not sure that's actually a good thing. Maybe each series should be a separate but equivalent entry?

And the results we got are quite different... 1 to 1.3 points difference in rORtg is actually significant. Food for thought.

Yeah, I agree that there’s validity to a weighted-average approach and a simple average approach. I tend to think weighted-average is better if trying to assess how well a team played in the aggregate. But arguably each series matters equally, so I can see an argument for weighting each series equally.

That said, I also get a very different result with a simple average. For instance, if I take a simple average of Jordan’s rORTG by series, I get +9.71 for Jordan’s career, +10.39 for 1989-1998, and +11.16 for just Jordan’s title years.

I’ll also note for others that, intuitively, I think the weighted-average approach is going to be worse for players from prior eras than a simple-average approach would be, because the NBA used to have best-of-5 first rounds (and, if we go far enough back, we had best-of-3 first rounds too). Since these guys were usually on very good teams (though not always), those shorter rounds tended to be against the worst teams. Worse teams tend to have worse DRTGs. A simple average approach would effectively weigh the DRTGs of those teams the same as the other opponents, while a weighted-average approach would tend to weigh it less because best-of-5 series typically have fewer possessions. As noted above, I see validity in either approach. But I just note this because it actually indicates that the approach I took was less charitable to Jordan as compared to the other guys I was comparing him to (LeBron, Steph, and Nash). Indeed, if we compare Jordan from 1985-1998 with LeBron from 2007-2010, the weighted-average approach had Jordan at +8.90 and LeBron at +8.06. Meanwhile, a simple-average approach for those same years has Jordan at +9.71 and LeBron at +8.09. So if we think the simple-average approach is better, then I think it actually works to Jordan’s advantage.

[Djoker]

I think what I’m doing is similar but subtly different. There’s a couple differences I have with that method:

1. Pace is not actually number of possessions per game, but rather is number of possessions per 48 minutes. So if a game goes into OT, there will be more possessions than the pace for that game. Therefore, when you multiply the pace by the number of games, you won’t get the right possession number if there were overtimes.

2. Averaging the opponents’ DRTGs doesn’t account for the fact that the player isn’t playing the same number of possessions against each defense. Which can skew the overall number. For instance, let’s say a player plays 400 possessions against a -5 rDRTG team, and 200 possessions against a +5 rDRTG team. Averaging the opponents’ DRTGs would come out to a totally neutral opponent DRTG. But actually the player played two-thirds of his possessions against the good defense. So, merely averaging the opponents’ DRTGs would undersell the average quality of defense faced in those possessions. This is why I think one should do a possession-weighted average. This way, you’re not accidentally skewing the numbers if the player played different numbers of possessions against teams with different-quality defenses.

To take the example of the 1993 Knicks series and the 1996 Sonics series, here’s what I’d do:

We calculate the number of possessions Jordan played in each series separately, by taking the minutes Jordan played, dividing by 48, and multiplying by the pace of the series. For the 1993 Knicks series, that’d be (249/48)*86.3 = 447.68. For the 1996 Sonics series, that’d be (252/48)*83.5 = 438.38. Now we figure out the rORTG for each series separately, by dividing the number of points scored with Jordan ON by the number of possessions and multiplying that by 100, and then subtracting the opposing team’s regular-season DRTG. For the 1993 Knicks series, that’d be 100*(528/447.68)-99.7=18.24. For the 1996 Sonics series, that’d be 100*(505/438.38)-102.1=13.10. Now that we have the rORTG for both series, we take a weighted average of those rORTGs, weighted by the percent of total possessions Jordan faced each team. Since we estimated 447.68 possessions in the Knicks series and 438.38 possessions in the Sonics series, the total possession estimate is 886.06. Therefore, the rORTG across those two series would be 18.24*(447.68/886.06)+13.10*(438.38/886.06)=15.70.

You could also get the same result by calculating the possessions and then taking the total on-court ORTG and subtracting a possession-weighted average of the opponents’ DRTGs. In that case the math would be 100*(528+505)/(886.06)-(99.7*(447.68/886.06)+102.1*(438.38/886.06)=15.70. Those two versions are functionally identical and get to the same result.

Anyways, that 15.70 number I got is similar to your number, since there was no overtime in those series, and the number of possessions in each series is also extremely similar. In some instances, there would be a larger difference between the two methods, because of the existence of OTs and because some series have very different possession numbers from each other along with different quality opposing defenses.

EDIT: I note that your calculations above used the Jordan ON minutes and pace numbers from the 1993 Finals against the Suns but used the 1993 Knicks DRTG. So I believe that particular +15.1 number you got is a result of a data input error. I think your method would still get a *very* slightly lower number than mine though, even with the correct data put in for the 1993 Knicks (it’d be like +15.68 instead of +15.70). This is because the possession number for the Knicks series is very slightly higher and their defense was a little bit better. As noted, though, the output would diverge a lot more if we had OTs and/or series that didn’t have very similar possession numbers.

_______

FWIW, as per the couple posts I’ve made on this page that listed rORTGs across series/years (one post about rORTGs against good defenses and goods teams, and the other post about rORTGs across specific timeframes), using the above-described method Jordan’s on-court rORTGs come out at a tier above others except Nash (who we’ve discussed lineup issues regarding). I find that data to be very compelling. And I just wanted to note the difference in the numbers we came to, since the numbers I got are actually a bit better for Jordan than the multi-year rORTG averages listed in your OP.

1. Yes that's correct. OT periods are taken into account with the minutes played but when averaging pace by games played, it can be a small issue.

2. My method is simply average rORtg per series. Yours is average rORtg weighted by possession in those series. Small distinction but nevertheless an important one. Your method is more accurate technically speaking but I'm still unsure what is more informative. Maybe we do actually want just a simple average of all the different series. Because your possession based method gives a single 7-game series almost twice the weight of a 4-game sweep. I'm not sure that's actually a good thing. Maybe each series should be a separate but equivalent entry?

And the results we got are quite different... 1 to 1.3 points difference in rORtg is actually significant. Food for thought.

Yeah, I agree that there’s validity to a weighted-average approach and a simple average approach. I tend to think weighted-average is better if trying to assess how well a team played in the aggregate. But arguably each series matters equally, so I can see an argument for weighting each series equally.

That said, I also get a different result with a simple average. For instance, if I take a simple average of Jordan’s rORTG by series, I get +9.71 for Jordan’s career, +10.39 for Jordan for 1989-1998, and +11.16 for just Jordan’s title years.

Yes taking a simple series average gives me those same averages too. 1st rounds are usually shorter than later rounds (way more sweeps and gentleman sweeps) so this tends to boost all players' numbers.

However I realize one thing generally hurting Jordan when it comes to all ON-OFF based metrics including rORtg is the 1st round being a best of 5 series. Usually the 1st round is the one round where the best teams annihilate opponents and playing 1-2 games less can really skew the result. For instance, MJ's Bulls had a +13.0 rORtg in 1993 but if the 1st round against the Hawks was 1 extra game (4-0 win instead of 3-0) and we assume the same ORtg as the other three games in the same series, the 1993 playoff average jumps up to +13.5 rORtg. Almost every playoffs except 1991, lengthening the 1st round by a game will improve the overall numbers.

From 2003 onwards, all 1st rounds are best of 7 so the likes of Nash/Lebron/Curry get an average of about one extra game against a minnow in every playoff run. It's not a huge deal but one can mentally correct Jordan's numbers up by +0.3 easily maybe even +0.5. Maybe this should be mathematically corrected for in all playoffs runs up to and including 2002. And obviously for other players too like Magic, Shaq etc.

Anyways it's interesting how the results are different depending on which method one uses to compute and average. I think I'll stick with the method in the OP because a) I've always done that and changing it would force me to recalculate a whole bunch of data for everything I've ever done! and b) Ben Taylor uses that method. I specifically asked him how his rORtg and rDRtg numbers are computed and he told me it's like that. He doesn't do a weighed average by possessions. But I'll definitely still think about it. And honestly any method is justifiable as long as it's consistently applied.