Compilation of Michael Jordan’s on-off data with the Bulls

[homecourtloss]

Also, the 1991 Bulls were a +9.5 NRtg (i.e., per 100) team. IF Jordan wound up with something like +9.85 ON for the season, the Bulls could NOT be a -9.77 with him off and still wind up a +9.5 team.

As I noted in my original post, the Bulls games in the Squared sample for 1991 were actually substantially less good for the Bulls than the full season. The Bulls only lost one of the unsampled games and had a bunch of blowout victories in the unsampled games. So this is not surprising at all. And it relates to something that I noted from the very beginning in my OP. As I mentioned in my OP, I’m not sure how it cuts in terms of bias, but my general inclination is to think that Jordan would probably look better in a better sample of games. We don’t know though, of course, and sampling error is definitely an issue with the Squared stuff and I think is a particular concern with the 1990-1991 season data since we know it’s not really a representative sample of games.

In the 56 sampled games, what was the Bulls NRtg? If we know this, then we can know what their NRtg was in the other 26 games (more or less since we’d assume the same pace) since we know the season NRtg has to be +9.5. The numbers in the others games would have to be WILDLY different because you already have 56 of the 82 games played leaving us with only 26 games left of possessions and ON minutes and OFF minutes. We know that:

—The 1991 Bulls were a +9.5 team in the regular season
—Jordan played 3,034 total minutes and apparently in the 56 game sample was +9.85 per 100
—(If we extend the 56 games sample, then the on-off would be +9.85 ON, +8 ON-OFF, +1.85 ON-OFF to get to a +9.5 NRtg for the team)
—We know the total off minutes given a full 82 game schedule plus the overtimes they played

If I assume that over the 56 game sample that Jordan played 37 minutes a game (like he did over the course of 1991 over 82 games), then:

—If Jordan was +14 per 100 on court in these other 26 games, then the Bulls would be about +3.8 per 100 in the off minutes, and Jordan would be about +11.2 ON, +3.8 OFF, +7.4 ON-OFF assuming he was playing the same amount in the other games though a cursory glance shows that he played fewer minutes in the blowouts

—If Jordan was +20 per 100 on court in these other 26 games, then the Bulls would be about -2.3 per 100 in the off minutes, and Jordan would be about +13 ON, -2.3 OFF, +15.3 ON-OFF

—If Jordan was +25 per 100 on court in these other 26 games, then the Bulls would be about -7.1 per 100 in the off minutes, and Jordan would be about +14.6 ON, -7.1 OFF, +21.7 ON-OFF

^These last two given the minutes/possessions adjustments in blowouts would likely have to be something like +27 per 100 to get to that final ON-OFF.

Also, if he does wind up with something around +9.85 on court in his consensus peak year (or about +10.9 for the RS+PS), well…that doesn’t seem like a GOAT peak when you have players with multiple year runs of RS+PS with on court numbers better than +9.85 or +10.9 and especially so if we’re talking single season. In a previous discussion on another thread, it was mentioned that a large on-off isn’t as impressive because someone is taking a bad team and making them good and that what was more impressive was the on court rating.

Jordan’s “on” number is surely substantially higher for the entire season, because, as noted above, the unsampled games were a lot better for the Bulls than the sampled games. So this isn't really a particularly valid point. I do agree, though, that the super high on-offs in Jordan's earlier years aren't as impressive as they would be if they were with a higher "on" value. Which is something I actually flagged in another thread when I noted that Jordan's four-year peak on-off in the data we have is higher than any four-year period I'm aware of for others but that Steph is fairly close but with a higher "on." In any event, Jordan does have some fantastic “on” values in the full data we have for the 1996 and 1997 regular seasons, so he’s not lacking in those sorts of seasons.

Even if the ON is higher (and almost surely it is), and we take the most optimistic view, it’s not an outlier peak. It’s very, very good like we have for others, and amongst the GoATs but not some mythical outlier most impressive peak as those were the claims along the lines of “most impressive/impactful peak” but the data doesn’t bear that out. Also, the claim was that 1991 and also 1988 through 1991 is the peak; referring to 1997 and 1998 takes us back to previous discussions about mid Jordan or later Jordan having perhaps more impact. Additionally we have RAPM data for 1996-1998 and that also doesn’t bear out any outlier peak.

Peak Jordan On-Off per 100 Possessions in Regular Season + Playoffs (1987-1988 to 1990-1991) (sample: 158 out of 387 games)

- On: +7.63
- Off: -13.35
- On-Off: +20.98
.

So Peak Jordan was +7.63 RS+PS over a 158 game sample. That’s…nice I guess, but not the type of astonishing high on court number that one would expect of a consensus peak stretch for a goat candidate. I recall in a different thread, KG’s on-off wasn’t all *that* impressive as it was mostly a high on-off raising a bad team, etc. Btw,

KG 2002 through 2004 (272 games): +6.21 on, +17.75 on-off
KG 2003 and 2004 (188 games): +6.73 on, +21.45 on-off

You’re not quite using four-year spans there (though I understand your point is similar number of games), but yeah, that’s a very high on-off span for Garnett—basically the highest in the play-by-play era.

And the fact is that if Jordan was rocking a +21 on-off in a four-year span, it would be unmatched by anyone in the play-by-play era, including Garnett (with the caveat again that Steph’s higher “on” value makes his best four-year span arguably more impressive in this regard). We don’t know for sure if Jordan really did have that high an on-off over four years, since we only have 41% of the data from those years, meaning there’s potentially significant sampling error. But what we do have looks undeniably incredible.
since the player voted #1 by this board has no such stretch. You seem to often create an artificial bar for Jordan that you don’t apply to anyone else (and that no one else would meet), and then say Jordan doesn’t meet it.

Again, I am not creating any artificially high bar for Jordan—you made the claim about the most impressive peak, etc. Again, now you’re also going back to on/off when before which you implied wasn’t as impressive for KG as the ON number in another thread.

I don’t actually think that Robinson’s +18.7 and Garnett’s +18.2 are super alike, and that’s kind of my point. Robinson’s +18.7 comes with around like a +10.3 net rating when he was on, while Garnett’s +18.2 comes with a +7.2 net rating when he was on. To me, the Robinson numbers are substantially better, since it is substantially easier to raise a team from really low numbers and substantially harder to raise it to the +10 sort of territory.

And even leaving aside raw achievement, I’ll repeat again that there’s diminishing marginal returns on things like on-off. It is definitely easier to raise an awful team a lot in terms of on-off than it is to raise a good team. And that also means it’s easier to accrue high RAPM on a bad team—because ultimately RAPM is aiming to isolate the effect you had on your team’s net rating and you can have a bigger effect on your team's net rating if the baseline for the team is bad. So, to me, I don't really find Garnett's on-off numbers or RAPM with the Timberwolves all *that* impressive.

I am impressed by both.

Let’s say that the more important ON (your argument) +7.63 ON from 1988 to 1991 is really +8.5 or +9.5 or +10 over this stretch. That is really impressive but not outlier peak at all when we have several,players likely around the same or higher.

As far as the “player voted #1,” from 2012 to 2017 (550+ RS+PS games) James was ~+9.4 to +9.8 ON court.

2015 to 2017 (278 games RS+PS), was +10.26 ON, -6.74 OFF, +17 ON-OFF. +17 and +21 might different enough but the +10.26 ON (what you imply has more value) is probably as high as Jordan’s optimistic ON from 1988-1991 (your calculations).

2008 to 2010 (260 RS+PS games), playing with that supporting cast is +8.99 ON. 2009-2010 (182 RS+PS games) is +12.28 ON. 2008 to 2010 + 2015 to 2017 (520 RS +PS) is +9.63 and so on.

Also, remember the optimistic “Jordan was +25 per 100 (more like +27 per 100) on court in these other 26 games, then the Bulls would be about -7.1 per 100 in the off minutes, and Jordan would be about +14.6 ON, -7.1 OFF, +21.7 ON-OFF

Well, these numbers look VERY familiar from a 2009 season, no?

So, being consistent: big on-offs are impressive especially when backed with lineup data (not just impressive for certain players), big on court numbers are impressive (especially when backed by lineup data), but these runs by these players are witching the same vicinity though some did it for longer.

I don’t see any outlier anything, “highest peak,” etc.

[AEnigma]

@Enigma

That is all fine, but when my assertion here is that:

a) when you do the same process for both points of comparison, they are very much within the margin for error of each other despite frequent suggestions to the contrary; and
b) this calculation demonstrably does not match up with what you see on NBA.com/BRef/pbpstats even when only looking at on-court measures

… then it feels like you are still speaking past the point. The error margins are separate from these measurements being fundamentally different.

These numbers do not extrapolate the way you assume they do. If I were doing the same blind Jordan process for Lebron, I would see basketball-reference list a 92.6 postseason pace and say he should actually be +4.96 per 100 on. Or if I think there is better data on NBA.com, I could see that 93.82 team pace and conclude he was +4.89 per 100 on. Or maybe I would see the 94.29 on-court pace number and say he should be +4.87 per 100.

However, that is not what happens with any of these sites. So why would you conclude that is what would happen with Jordan? Your process inflates numbers relative to what is listed on all these “official” sites; I am not criticising that in itself, but it is why I insist on you applying the process evenly. Numbers are not meaningful without equal comparison. And this is not isolated to Lebron: this applies to everyone you are comfortable declaring to have fallen short of Jordan’s values based on completely disparate means of measurement.

Here, try it out with Curry if you do not care about running numbers for Lebron. Surely you must be at least a little curious about what would happen if you went through the same process you used on Jordan? I get +10.1 per 48 on for 2015-19 Curry, but pbp only has him at +9.74 even though it also says he played at a 97.893 pace. Do you not think that is something worth considering so long as you continue to advance this methodology?

I think this provides good reason to not be particularly certain about extrapolations about exact on and off pace differences based on per-48-minute data. Which, I’ll note, is why my last post did not provide precise numbers and was not worded in any definitive way (I talked about what the data “suggests” and said “if this were true” and that it “might underestimate” Jordan’s on-off “ if the difference were as big as it looks like”).

No, again, you are assuming that there are error measurements potentially going both ways. There are not. Your process is almost always inflationary. It does not match your intuition. You could test that out, but you refuse.

Here is another: 1998 Malone with +48 in 796 minutes -> +2.89 per 48 minutes. Basketball reference says +1.7 per 100 even though we know neither the Jazz nor Malone himself were a fast team (including by BRef’s own estimates). Time and time again, we see this is a clearly distinct process.

The truth is that we don’t really know for sure what, if any, pace differences on and off the floor there were for Jordan but, as I’ve said, we have no reason whatsoever to believe that difference would be the same direction and magnitude for Jordan as it is for LeBron, and so using data for LeBron that we know would incorporate this error (when we have data without the error) does not make sense since there’s no reason to believe that would cancel out any error from pace difference for Jordan (indeed, the equal-pace assumption might well *hurt* Jordan, rather than help like it would for the LeBron).

I literally showed you an example where Lebron played faster than his team did without him. You keep speculating that this is why there is such a regular distinction, and it is not.

As for the rest of it, you’re largely just pointing out that on-off models can be a bit wonky (especially at smaller sample sizes), likely due to possession estimations sometimes being inaccurate at small sample sizes.

At a fundamental level, on-off models are absolutely just trying to take +/- when “on” and divide by an estimate of possessions “on” and taking +/- when “off” and dividing by an estimate of possessions “off.” And that is exactly what I did, except with an equal-pace assumption to determine the breakdown between “on” and “off” minutes. So I don’t know what other error/bias you’re specifically asserting there is. Is your assertion that the possession estimate formula BBREF uses is different when calculating pace than it is when calculating on-off? That seems…quite unlikely, but if you have evidence of that then please show it.

I am telling you to check for yourself, because every time I highlight an obvious inconsistency, you ignore and dismiss it. This is not difficult: per 100 should be a reliably larger value than these per 48 hand calculations so long as pace is below 100, yet that is not what we see. So you are doing this process where you are regularly creating that increase, and when the disconnect is pointed out, you shrug and attempt to argue that maybe Jordan’s values would be even higher. I will reiterate: every time the pace is below 100, your process guarantees the per-100 number is higher than the per-48 number. That is not what happens in any of these “official” data sources you insist must be used once available.

Otherwise I don’t really see what source of error/bias you’re identifying beyond the equal-pace assumption, and we have no idea the direction or magnitude of that for Jordan (and have at least some indication it might be *hurting* his numbers). The bottom line is that if you think my method is biasing the data in a specific way beyond the equal-pace assumption, you need to explain exactly how rather than just declaring that it must be different without identifying the specific mechanism through which that’d be the case.

? What? No, I do not know the coding here, how would I know the specific mechanism. This is a quantifiable observation about what how your calculation process does not match. I am not even saying your math is bad or anything; just that it may as well be functioning as a different measurement. And it is illegitimate to compare different measurements like that.

Again, there is absolutely no “hurt”. When you multiply by 100/(#<100), the number increases.

Here is another: 1998 Jordan, +158 in 871.5 minutes, so +8.7 per 48. BRef says +8.8 per 100, so even if we assume that is secretly +8.84, Jordan could only be playing at a 98 pace. Again, every indication we have is they were a lot slower than that. 1997 Jordan, +136 in 803.7 minutes, so +8.12. Here, finally, we get something plausible at a BRef estimate of +8.8 per 100… but again, that is still at best a 92 pace, for a team estimated to have played at an 85.5 pace that postseason. It does not add up, and it pretty much never does. I do not know exactly why, but I can at least see it; can you?

[Squared2020]

.

[Djoker]

.

I think Jordan could easily be +20 or more ON court in the remaining 26 games in the 1991 season. His team went 25-1 (79-win pace) with a bunch of blowouts. Jordan himself in 1996 was +16.7 ON court and Curry had seasons of +18.0, +17.1 and +16.7 ON Court as well. So +20 over a super hot team stretch isn't at all unlikely. Since the sample is +6.0 MOV and total season MOV is +9.0 the Bulls' MOV in the remaining games is like +15.5 which is insanely high.

[OhayoKD]

I don't know about you guys, but when I read these posts, I lose interest really fast. Maybe it's the black text?

What are we trying to do here... Computing On/Off ratings for players with respect to the teams? IE: Bulls +14 when Jordan ON versus -14 when Jordan OFF... something like that?

Shouldn't I care of he's being D'd up by Long-Arms McGee instead of Waterboy Williams in those situations?

I think the idea(and one that jake supports strongly) is its harder to take a 40 win to team to 60 than a 30 win team to 50. And while I don't know how the 4-year stuff grades out(pretty odd to use with two incomplete samples tbh), at least for one-year, along those lines, Lebron's 2009 and Steph's 2016 are probably unreachable(on-court ratings of +15 and +18!), especially with evenly applied pace-adjustment. RK says Magic is in a similar light on a +5 srs team too so there's that.

And of course on/off is cool and all, but 27 to 50 =/ 19 to 61 or 30 to 60 or 40 to 65.

That said, I think people exaggerate the effect. It is generally harder to lift better teams because a better team will have less room for an individual to impart maximum value. But this should be seen as more of a tiebreaker rather than some massive fundemental difference as has been potrayed.

[TheGOATRises007]

The on-off numbers for 92 further strengthening my argument that the 92 Bulls were the best Bulls team and they barely get talked about.

[homecourtloss]

Spoiler:

86 Bulls Game 2: 1986 NBA Playoffs Round 01 -Game 2 : Chicago Bulls vs Boston Celtics
[url]

Watch on YouTube

[/url]

86 Bulls Game 3: Chicago Bulls vs Boston Celtics 1986 Playoffs Game 3
[url]

Watch on YouTube

;list=PLDD8CAF80BF715316[/url]

I have 1986 game 3 vs. the Celtics

—1st quarter, Jordan -10
—2nd quarter goes out with 4:30 or 4:40 left in 2nd quarter at -11 (have to manually count seconds to be exact since they didn’t show the game clock in these days unless it was CBS televised games and not even then sometimes)
—Bulls -3 in 4:30 to end second quarter without Jordan

—1st half:
19 minutes and 30 seconds of Jordan: -11
4 minutes and 30 seconds without Jordan: -3

After three quarters:
31 minutes and 20 seconds with Jordan: -25
4 minutes and 40 seconds without Jordan: -3

4th quarter
5 minutes left in fourth quarter: -17 (fouls out)

TOTAL
38:30–39:00 with Jordan on court: -17
9:00–9:30 with Jordan off court: -1

Tagging others who were interested in this 1986 Celtics vs. Bulls series

… entire playoff career excluding 1986 (3 games), 1987 (3 games) and 1995 (10 games).

1986 Bulls vs. Celtics Game 2
1st quarter: played all 12 minutes, 33-25 Bulls, Jordan +8

2nd quarter: sits with Bulls up 33-29, +4, 9:56 left in 2nd

—Bulls 41-36, 6:20 left
—Bulls 43-36, 6:00 left
—Bulls 43-38, 5:45 left
—Bulls 43-41, 5:00 left (Jordan re-enters the game)

Jordan Played the rest of the game; he re-entered while he was +4 on court, Bulls up 2; Bulls lose by 4, Jordan goes to -2 overall

TOTAL
Jordan 52:56 seconds on court: -2
5:04 seconds Jordan off: -2

Some play by play from 2nd half of game 2:

3rd quarter
77-69 Bulls in the third
81-74 Bulls
81-81
84-81 Bulls
89-85 Bulls
91-85 Bulls
91-88 [Ainge contested pull up three to end third—12 points in third; Jordan has 36 at the end of third]

4th quarter
93-92, Celtics take lead on 30 foot three pointer at the end of the shot clock by Larry Bird, 11:05 left
97-96 on Kevin McHale shot basically from his back sitting on Corzine, 9:06 left
104-100 Celtics with 7:12 left [Jordan operating in and around lane with pull ups]
106-104, Celticss with 515 left
108-107 Celtics 3:30
110-107 Celtics with 3:00 left
110-109 Celtics with 2:50 left
111-110 Bulls with 2:15 left [Jordan with 50]
111-111 with 2:00 left
113-111 with 1:30 left
114-113 Celtics 1:00 left
116-113 Celtics with :45 left
116-114 :25, parish lost rebound, Bulls with ball with 6 seconds
116-116 on foul call on mchale after shot, Jordan makes 2 FTs

First OT
119-128 Celtics 2:20 left in OT
120-119 bulls 2:00 in OT
123-119 Bulls 1:30 left in OT
123-121 Bulls 1:20 left in OT
123-123 :50 left in OT
125-123 Bulls :22 left in OT
125-125 :12 left in OT, Danny Ainge makes left-handed drive
125-125 :02 left in OT, Jordan misses 15 foot jumper leaning slightly left after McHale stumbles that was wide open
125-125 end of OT, Larry bird misses 30 foot three pointer off of screen back rim

2nd OT
131-129 Celtics, Jordan with 61, 1:45
131-131, Jordan with 63 on lane hanging shot
133-131, Jerry Schisting 17 footer, :45 seconds left
133-131, Jordan misses pull up over mchale, :27 left
135-131, Parish makes baseline 10 footer off of pick and roll action, :09 left
135-131, Woldridge airball 28 footer with :03 left
135-131 final

[homecourtloss]

Tagging others who were interested in this 1986 Celtics vs. Bulls series

… entire playoff career excluding 1986 (3 games), 1987 (3 games) and 1995 (10 games).

1986 Bulls vs. Celtics Game 1

Since we don’t have the entire game, this one is a little bit more difficult, but it is possible to get an approximation because we have the highlights, and we know the rotation/substitution patterns, and the fact that Jordan played nearly the entire game

1986 game 1 highlights

Watch on YouTube

1st quarter, Bulls up 35-27, Jordan starts off 5-5, almost surely played entire 1st as he did in game 2 and  in game 3, +8

2nd quarter:, So this is the only ambiguity here. It’s likely that just like in game two, Jordan takes all of his rest minutes somewhere in this second quarter. In game 2 and in game 3, he played a bit in the 2nd then sat and then played the rest of the game (until fouling out in game 3)

At 3:40 in the video, Jordan steals and dunks to make it 43-31:



At 4:00 they show him sitting on the bench so he’s resting but we don’t know the score. Then the video shows Jordan scoring twice (drive and acrobatic finish, drive and dunk on non-existent defense) and then we see him bringing the ball up court at 47-46 Bulls after a either a made shot/FT(s) by the Celtics or turnover (impossible to tell). He was likely out for all the 5 minutes he wasn’t on court in the game (just like game 2 and game 3.



If we take the most optimistic view that he went out at 43-31 and came back at 43-46, he would be +12 on court and then then the Bulls would go on to trail 59-61 at the half making Jordan +13 on court at the half. We do know that the Bulls went up 51-46 and then lost the lead at the half though this doesn’t affect the overall +13 on court.

3rd quarter: He played the entire 3rd quarter and made the shot at the buzzer (see at 6:30 in the video) to cut the Celtics’ lead to 10. Jordan was -8 in the quarter, putting him at the most optimistic at +5 for the game.

4th quarter: He played likely the entire quarter. From the video, we see him bringing the ball up the court at 6:45 in the video down 84-94 (the score at the beginning of the 4th)



And see him scoring down 117-103



Then you see him on court at the end of the game at the buzzer



He was -9 in the fourth quarter.

Total (most optimistic view)

43 minutes Jordan On court: -4
5 minutes Jordan off court: -15

The Bulls were able to score with Jordan off court innthr ither games and them being -15 in 5 minutes is highly unlikely.

Realistic positive estimate would be something like

43 minutes Jordan On court: -8
5 minutes Jordan off court: -11

[homecourtloss]

1986 Bulls vs. Celtics Playoffs Jordan ON/OFF

1986 game 1 vs. the Celtics

Spoiler:

1986 Bulls vs. Celtics Game 1
Since we don’t have the entire game, this one is a little bit more difficult, but it is possible to get an approximation because we have the highlights, and we know the rotation/substitution patterns, and the fact that Jordan played nearly the entire game

1986 game 1 highlights

Watch on YouTube

1st quarter, Bulls up 35-27, Jordan starts off 5-5, almost surely played entire 1st as he did in game 2 and  in game 3, +8

2nd quarter:, So this is the only ambiguity here. It’s likely that just like in game two, Jordan takes all of his rest minutes somewhere in this second quarter. In game 2 and in game 3, he played a bit in the 2nd then sat and then played the rest of the game (until fouling out in game 3)

At 3:40 in the video, Jordan steals and dunks to make it 43-31:



At 4:00 they show him sitting on the bench so he’s resting but we don’t know the score. Then the video shows Jordan scoring twice (drive and acrobatic finish, drive and dunk on non-existent defense) and then we see him bringing the ball up court at 47-46 Bulls after a either a made shot/FT(s) by the Celtics or turnover (impossible to tell). He was likely out for all the 5 minutes he wasn’t on court in the game (just like game 2 and game 3.



If we take the most optimistic view that he went out at 43-31 and came back at 43-46, he would be +12 on court and then then the Bulls would go on to trail 59-61 at the half making Jordan +13 on court at the half. We do know that the Bulls went up 51-46 and then lost the lead at the half though this doesn’t affect the overall +13 on court.

3rd quarter: He played the entire 3rd quarter and made the shot at the buzzer (see at 6:30 in the video) to cut the Celtics’ lead to 10. Jordan was -8 in the quarter, putting him at the most optimistic at +5 for the game.

4th quarter: He played likely the entire quarter. From the video, we see him bringing the ball up the court at 6:45 in the video down 84-94 (the score at the beginning of the 4th)



And see him scoring down 117-103



Then you see him on court at the end of the game at the buzzer



He was -9 in the fourth quarter.

Total (most optimistic view)

43 minutes Jordan On court: -4
5 minutes Jordan off court: -15

The Bulls were able to score with Jordan off court in their games and them being -15 in 5 minutes is highly unlikely.

Realistic positive estimate would be something like

43 minutes Jordan On court: -8
5 minutes Jordan off court: -11

Total (most optimistic view)
43 minutes Jordan On court: -4
5 minutes Jordan off court: -15

The Bulls were able to score with Jordan off court in their games and them being -15 in 5 minutes is highly unlikely.

Realistic positive estimate would be something like
43 minutes Jordan On court: -8
5 minutes Jordan off court: -11

Pessimistic estimate (unrealistic) would be something like
43 minutes Jordan On court: -14
5 minutes Jordan off court: -5

1986 game 2 vs. the Celtics

Spoiler:

1st quarter:[/b] played all 12 minutes, 33-25 Bulls, Jordan +8

2nd quarter: sits with Bulls up 33-29, +4, 9:56 left in 2nd

—Bulls 41-36, 6:20 left
—Bulls 43-36, 6:00 left
—Bulls 43-38, 5:45 left
—Bulls 43-41, 5:00 left (Jordan re-enters the game)

Jordan Played the rest of the game; he re-entered while he was +4 on court, Bulls up 2; Bulls lose by 4, Jordan goes to -2 overall

TOTAL
Jordan 52:56 seconds on court: -2
5:04 seconds Jordan off: -2

Some play by play from 2nd half of game 2:

3rd quarter
77-69 Bulls in the third
81-74 Bulls
81-81
84-81 Bulls
89-85 Bulls
91-85 Bulls
91-88 [Ainge contested pull up three to end third—12 points in third; Jordan has 36 at the end of third]

4th quarter
93-92, Celtics take lead on 30 foot three pointer at the end of the shot clock by Larry Bird, 11:05 left
97-96 on Kevin McHale shot basically from his back sitting on Corzine, 9:06 left
104-100 Celtics with 7:12 left [Jordan operating in and around lane with pull ups]
106-104, Celticss with 515 left
108-107 Celtics 3:30
110-107 Celtics with 3:00 left
110-109 Celtics with 2:50 left
111-110 Bulls with 2:15 left [Jordan with 50]
111-111 with 2:00 left
113-111 with 1:30 left
114-113 Celtics 1:00 left
116-113 Celtics with :45 left
116-114 :25, parish lost rebound, Bulls with ball with 6 seconds
116-116 on foul call on mchale after shot, Jordan makes 2 FTs

First OT
119-128 Celtics 2:20 left in OT
120-119 bulls 2:00 in OT
123-119 Bulls 1:30 left in OT
123-121 Bulls 1:20 left in OT
123-123 :50 left in OT
125-123 Bulls :22 left in OT
125-125 :12 left in OT, Danny Ainge makes left-handed drive
125-125 :02 left in OT, Jordan misses 15 foot jumper leaning slightly left after McHale stumbles that was wide open
125-125 end of OT, Larry bird misses 30 foot three pointer off of screen back rim

2nd OT
131-129 Celtics, Jordan with 61, 1:45
131-131, Jordan with 63 on lane hanging shot
133-131, Jerry Schisting 17 footer, :45 seconds left
133-131, Jordan misses pull up over mchale, :27 left
135-131, Parish makes baseline 10 footer off of pick and roll action, :09 left
135-131, Woldridge airball 28 footer with :03 left
135-131 final

TOTAL
Jordan 52:56 on court: -2
5:04 Jordan off court: -2

1986 game 3 vs. the Celtics

Spoiler:

—1st quarter, Jordan -10
—2nd quarter goes out with 4:30 or 4:40 left in 2nd quarter at -11 (have to manually count seconds to be exact since they didn’t show the game clock in these days unless it was CBS televised games and not even then sometimes)
—Bulls -3 in 4:30 to end second quarter without Jordan

—1st half:
19 minutes and 30 seconds of Jordan: -11
4 minutes and 30 seconds without Jordan: -3

After three quarters:
31 minutes and 20 seconds with Jordan: -25
4 minutes and 40 seconds without Jordan: -3

4th quarter
5 minutes left in fourth quarter: -17 (fouls out)

TOTAL
Jordan 38:30-39:00 on court: -17
Jordan 9:00–9:30 with off court: -1

[DraymondGold]

1986 game 2 vs. the Celtics

Spoiler:

1st quarter:[/b] played all 12 minutes, 33-25 Bulls, Jordan +8

2nd quarter: sits with Bulls up 33-29, +4, 9:56 left in 2nd

—Bulls 41-36, 6:20 left
—Bulls 43-36, 6:00 left
—Bulls 43-38, 5:45 left
—Bulls 43-41, 5:00 left (Jordan re-enters the game)

Jordan Played the rest of the game; he re-entered while he was +4 on court, Bulls up 2; Bulls lose by 4, Jordan goes to -2 overall

TOTAL
Jordan 52:56 seconds on court: -2
5:04 seconds Jordan off: -2

Some play by play from 2nd half of game 2:

3rd quarter
77-69 Bulls in the third
81-74 Bulls
81-81
84-81 Bulls
89-85 Bulls
91-85 Bulls
91-88 [Ainge contested pull up three to end third—12 points in third; Jordan has 36 at the end of third]

4th quarter
93-92, Celtics take lead on 30 foot three pointer at the end of the shot clock by Larry Bird, 11:05 left
97-96 on Kevin McHale shot basically from his back sitting on Corzine, 9:06 left
104-100 Celtics with 7:12 left [Jordan operating in and around lane with pull ups]
106-104, Celticss with 515 left
108-107 Celtics 3:30
110-107 Celtics with 3:00 left
110-109 Celtics with 2:50 left
111-110 Bulls with 2:15 left [Jordan with 50]
111-111 with 2:00 left
113-111 with 1:30 left
114-113 Celtics 1:00 left
116-113 Celtics with :45 left
116-114 :25, parish lost rebound, Bulls with ball with 6 seconds
116-116 on foul call on mchale after shot, Jordan makes 2 FTs

First OT
119-128 Celtics 2:20 left in OT
120-119 bulls 2:00 in OT
123-119 Bulls 1:30 left in OT
123-121 Bulls 1:20 left in OT
123-123 :50 left in OT
125-123 Bulls :22 left in OT
125-125 :12 left in OT, Danny Ainge makes left-handed drive
125-125 :02 left in OT, Jordan misses 15 foot jumper leaning slightly left after McHale stumbles that was wide open
125-125 end of OT, Larry bird misses 30 foot three pointer off of screen back rim

2nd OT
131-129 Celtics, Jordan with 61, 1:45
131-131, Jordan with 63 on lane hanging shot
133-131, Jerry Schisting 17 footer, :45 seconds left
133-131, Jordan misses pull up over mchale, :27 left
135-131, Parish makes baseline 10 footer off of pick and roll action, :09 left
135-131, Woldridge airball 28 footer with :03 left
135-131 final

TOTAL
Jordan 52:56 on court: -2
5:04 Jordan off court: -2

Fantastic stuff homecourtloss! Really appreciate the tracking data.

I might have a slightly different number for 1986 Game 2 -- I'll show how I got it, but let me know if we disagree.

-33:50 in the video: Jordan goes out. Score: 33-29, +4 in favor of bulls. 9:56 left in 2nd (we agree).
-43:32 in the video: Jordan still out. Score 43–38 (we agree).
Then Bird scores a 2 at 42:55, so the score is 43–40. (we agree)
-43:20 in the video: 43-40, +3 in favor of Bulls 5:00 left in 2nd. (we agree)
*Then Jordan comes back in* before Bird makes the foul shot, making it 43-41 only after Jordan comes back in (we disagree).

So I guess the main disagreement is whether to count the foul shot by Bird as while Jordan was off (the foul happened before the shot) or while Jordan was on (he came on the court during the break before the foul shot).

I'm genuinely not sure what the standard way to count this is. It's only 1 point, so not much of a difference regardless, but still fun to get in the weeds of this.

As for the time, if Jordan left at 9:56 then gets back on the court at 5:00, I think Jordan would be off for 4:56, not 5:04 (and thus play for 53:04, not 52:26). Again just a small difference, not anything huge.

Converting to per 48 (if we use my count for the score/time):
Without Jordan: -1 per 4:56 minutes * (per 48 minutes / per 4:56 minutes) = -9.73 per 48 minutes
With Jordan: -3 per 53:04 minutes) * (per 48 minutes / per 53:04 minutes) = -2.71 per 48 minutes
change: +7.02 per 48 minutes.

[homecourtloss]

I might have a slightly different number for 1986 Game 2 -- I'll show how I got it, but let me know if we disagree.

-33:50 in the video: Jordan goes out. Score: 33-29, +4 in favor of bulls. 9:56 left in 2nd (we agree).
-43:32 in the video: Jordan still out. Score 43–38 (we agree).
Then Bird scores a 2 at 42:55, so the score is 43–40. (we agree)
-43:20 in the video: 43-40, +3 in favor of Bulls 5:00 left in 2nd. (we agree)
*Then Jordan comes back in* before Bird makes the foul shot, making it 43-41 only after Jordan comes back in (we disagree).

So I guess the main disagreement is whether to count the foul shot by Bird as while Jordan was off (the foul happened before the shot) or while Jordan was on (he came on the court during the break before the foul shot).

I'm genuinely not sure what the standard way to count this is. It's only 1 point, so not much of a difference regardless, but still fun to get in the weeds of this. .

To be honest, going through some of the tracking data, I’ve seen this and counted both ways, but most of the time I have seen it that the free throw would not count against Jordan since he was not on the court when the foul occurred, and that free-throw was generated when he was not on the court.

One thing I do want to mention about this game especially, but the other two games as well is that for all the crying that people have about fouls, there were dozens of touch fouls for both teams in this 1986 series. Obviously, we know the free-throw rate in the 1980s in the 1990s was much higher than it is now, but many people attribute this to so much more three-point shooting now, but when you watch a lot of the games, you just see a whole bunch of touch fouls. People want to simultaneously say that more physicality was allowed but yet somehow more physicality was allowed, but they were a whole bunch more free throws so then there must be some really hard and brutal, fouls, right? When you watch the games, it’s just really not true. Watching these games, you see maybe a few hard fouls, but nothing out of the ordinary and just one touch foul after another. Jordan had a few touch fouls go his way, bird had a few touch fouls go his way, McAle, John Paxson got a bunch of them in game three, and they were just a whole bunch of on the floor foul calls, but both teams were over the foul limit and it was an endless parade of free throws.

I started watching these games as a kid in 1985, and of course, I remember them differentl because of the emotion involved while watching them. Now, when you watch them to track data or to do a film analysis, you really see the absurd amount of touch vowels, and how much it ruined the rhythm of the game. People only tend to remember the few hard fouls that they’ve seen on the highlights but unless one spends time, tracking these games, one doesn’t really see what these fouls look like.

[Squared2020]

.

[DraymondGold]

-33:50 in the video: Jordan goes out. Score: 33-29, +4 in favor of bulls. 9:56 left in 2nd (we agree).
-43:32 in the video: Jordan still out. Score 43–38 (we agree).
Then Bird scores a 2 at 42:55, so the score is 43–40. (we agree)
-43:20 in the video: 43-40, +3 in favor of Bulls 5:00 left in 2nd. (we agree)
*Then Jordan comes back in* before Bird makes the foul shot, making it 43-41 only after Jordan comes back in (we disagree).

The Corzine foul occurred while Jordan was on the bench. The free throw does not count against Jordan's plus-minus even though he substituted in before the free throw.

Edit: The free throw counts against George Gervin, who Jordan subbed in for.

Thanks! Makes sense.

So the updated Game 2 plus minus per 48 would be:

Converting to per 48 (if we use my count for the score/time):
Without Jordan: -2 per 4:56 minutes * (per 48 minutes / per 4:56 minutes) = -19.46 per 48 minutes
With Jordan: -2 per 53:04 minutes * (per 48 minutes / per 53:04 minutes) = -1.81 per 48 minutes
plus minus: +17.65 per 48 minutes.

If we want to calculate the per 100 directly, there were 8 possessions (for each team), so
without Jordan would be: -2 per 8 possessions * (per 100 / per 8 possessions) = -25 per 100 possessions.

Haven't counted the possessions with Jordan yet.

[lessthanjake]

1986 Bulls vs. Celtics Playoffs Jordan ON/OFF

1986 game 1 vs. the Celtics

Spoiler:

1986 Bulls vs. Celtics Game 1
Since we don’t have the entire game, this one is a little bit more difficult, but it is possible to get an approximation because we have the highlights, and we know the rotation/substitution patterns, and the fact that Jordan played nearly the entire game

1986 game 1 highlights

Watch on YouTube

1st quarter, Bulls up 35-27, Jordan starts off 5-5, almost surely played entire 1st as he did in game 2 and  in game 3, +8

2nd quarter:, So this is the only ambiguity here. It’s likely that just like in game two, Jordan takes all of his rest minutes somewhere in this second quarter. In game 2 and in game 3, he played a bit in the 2nd then sat and then played the rest of the game (until fouling out in game 3)

At 3:40 in the video, Jordan steals and dunks to make it 43-31:



At 4:00 they show him sitting on the bench so he’s resting but we don’t know the score. Then the video shows Jordan scoring twice (drive and acrobatic finish, drive and dunk on non-existent defense) and then we see him bringing the ball up court at 47-46 Bulls after a either a made shot/FT(s) by the Celtics or turnover (impossible to tell). He was likely out for all the 5 minutes he wasn’t on court in the game (just like game 2 and game 3.



If we take the most optimistic view that he went out at 43-31 and came back at 43-46, he would be +12 on court and then then the Bulls would go on to trail 59-61 at the half making Jordan +13 on court at the half. We do know that the Bulls went up 51-46 and then lost the lead at the half though this doesn’t affect the overall +13 on court.

3rd quarter: He played the entire 3rd quarter and made the shot at the buzzer (see at 6:30 in the video) to cut the Celtics’ lead to 10. Jordan was -8 in the quarter, putting him at the most optimistic at +5 for the game.

4th quarter: He played likely the entire quarter. From the video, we see him bringing the ball up the court at 6:45 in the video down 84-94 (the score at the beginning of the 4th)



And see him scoring down 117-103



Then you see him on court at the end of the game at the buzzer



He was -9 in the fourth quarter.

Total (most optimistic view)

43 minutes Jordan On court: -4
5 minutes Jordan off court: -15

The Bulls were able to score with Jordan off court in their games and them being -15 in 5 minutes is highly unlikely.

Realistic positive estimate would be something like

43 minutes Jordan On court: -8
5 minutes Jordan off court: -11

Total (most optimistic view)
43 minutes Jordan On court: -4
5 minutes Jordan off court: -15

The Bulls were able to score with Jordan off court in their games and them being -15 in 5 minutes is highly unlikely.

Realistic positive estimate would be something like
43 minutes Jordan On court: -8
5 minutes Jordan off court: -11

Pessimistic estimate (unrealistic) would be something like
43 minutes Jordan On court: -14
5 minutes Jordan off court: -5

1986 game 2 vs. the Celtics

Spoiler:

1st quarter:[/b] played all 12 minutes, 33-25 Bulls, Jordan +8

2nd quarter: sits with Bulls up 33-29, +4, 9:56 left in 2nd

—Bulls 41-36, 6:20 left
—Bulls 43-36, 6:00 left
—Bulls 43-38, 5:45 left
—Bulls 43-41, 5:00 left (Jordan re-enters the game)

Jordan Played the rest of the game; he re-entered while he was +4 on court, Bulls up 2; Bulls lose by 4, Jordan goes to -2 overall

TOTAL
Jordan 52:56 seconds on court: -2
5:04 seconds Jordan off: -2

Some play by play from 2nd half of game 2:

3rd quarter
77-69 Bulls in the third
81-74 Bulls
81-81
84-81 Bulls
89-85 Bulls
91-85 Bulls
91-88 [Ainge contested pull up three to end third—12 points in third; Jordan has 36 at the end of third]

4th quarter
93-92, Celtics take lead on 30 foot three pointer at the end of the shot clock by Larry Bird, 11:05 left
97-96 on Kevin McHale shot basically from his back sitting on Corzine, 9:06 left
104-100 Celtics with 7:12 left [Jordan operating in and around lane with pull ups]
106-104, Celticss with 515 left
108-107 Celtics 3:30
110-107 Celtics with 3:00 left
110-109 Celtics with 2:50 left
111-110 Bulls with 2:15 left [Jordan with 50]
111-111 with 2:00 left
113-111 with 1:30 left
114-113 Celtics 1:00 left
116-113 Celtics with :45 left
116-114 :25, parish lost rebound, Bulls with ball with 6 seconds
116-116 on foul call on mchale after shot, Jordan makes 2 FTs

First OT
119-128 Celtics 2:20 left in OT
120-119 bulls 2:00 in OT
123-119 Bulls 1:30 left in OT
123-121 Bulls 1:20 left in OT
123-123 :50 left in OT
125-123 Bulls :22 left in OT
125-125 :12 left in OT, Danny Ainge makes left-handed drive
125-125 :02 left in OT, Jordan misses 15 foot jumper leaning slightly left after McHale stumbles that was wide open
125-125 end of OT, Larry bird misses 30 foot three pointer off of screen back rim

2nd OT
131-129 Celtics, Jordan with 61, 1:45
131-131, Jordan with 63 on lane hanging shot
133-131, Jerry Schisting 17 footer, :45 seconds left
133-131, Jordan misses pull up over mchale, :27 left
135-131, Parish makes baseline 10 footer off of pick and roll action, :09 left
135-131, Woldridge airball 28 footer with :03 left
135-131 final

TOTAL
Jordan 52:56 on court: -2
5:04 Jordan off court: -2

1986 game 3 vs. the Celtics

Spoiler:

—1st quarter, Jordan -10
—2nd quarter goes out with 4:30 or 4:40 left in 2nd quarter at -11 (have to manually count seconds to be exact since they didn’t show the game clock in these days unless it was CBS televised games and not even then sometimes)
—Bulls -3 in 4:30 to end second quarter without Jordan

—1st half:
19 minutes and 30 seconds of Jordan: -11
4 minutes and 30 seconds without Jordan: -3

After three quarters:
31 minutes and 20 seconds with Jordan: -25
4 minutes and 40 seconds without Jordan: -3

4th quarter
5 minutes left in fourth quarter: -17 (fouls out)

TOTAL
Jordan 38:30-39:00 on court: -17
Jordan 9:00–9:30 with off court: -1

This is good stuff. I’m not going to add it in to the OP just yet because of the uncertainty over Game 1. If we can get something more definite on that, then I will (in theory I could include just the two games we have more definite data on, but I’d rather have the full sample for a playoffs).

Anyways, on a per-48-minute basis and using the middle-ground game 1 estimate, that’d put it at about -9.6 per 48 minutes with Jordan on and about -35.4 per 48 minutes with Jordan off. If I try to get a per-100-possession estimate using an equal-pace estimate based on BBREF pace, that’d be about -9.8 on the court and -36.2 off the court.

And if I did add that to the full playoff estimate, the total estimate with 1985-1986 + 1988-1993 + 1996-1998 would now be +7.99 on, -9.09 off, and a +17.08 on-off. Which is up from +16.94 without the 1986 series.

So it doesn’t make a huge difference (as probably to be expected from such a short series), but probably moves things forward in Jordan’s favor a bit in terms of on-off. Of course, the actual numbers are probably slightly different than this because we have a fairly wide uncertainty band for Game 1.

[lessthanjake]

@Enigma

That is all fine, but when my assertion here is that:

a) when you do the same process for both points of comparison, they are very much within the margin for error of each other despite frequent suggestions to the contrary; and
b) this calculation demonstrably does not match up with what you see on NBA.com/BRef/pbpstats even when only looking at on-court measures

… then it feels like you are still speaking past the point. The error margins are separate from these measurements being fundamentally different.

These numbers do not extrapolate the way you assume they do. If I were doing the same blind Jordan process for Lebron, I would see basketball-reference list a 92.6 postseason pace and say he should actually be +4.96 per 100 on. Or if I think there is better data on NBA.com, I could see that 93.82 team pace and conclude he was +4.89 per 100 on. Or maybe I would see the 94.29 on-court pace number and say he should be +4.87 per 100.

However, that is not what happens with any of these sites. So why would you conclude that is what would happen with Jordan? Your process inflates numbers relative to what is listed on all these “official” sites; I am not criticising that in itself, but it is why I insist on you applying the process evenly. Numbers are not meaningful without equal comparison. And this is not isolated to Lebron: this applies to everyone you are comfortable declaring to have fallen short of Jordan’s values based on completely disparate means of measurement.

Here, try it out with Curry if you do not care about running numbers for Lebron. Surely you must be at least a little curious about what would happen if you went through the same process you used on Jordan? I get +10.1 per 48 on for 2015-19 Curry, but pbp only has him at +9.74 even though it also says he played at a 97.893 pace. Do you not think that is something worth considering so long as you continue to advance this methodology?

I think this provides good reason to not be particularly certain about extrapolations about exact on and off pace differences based on per-48-minute data. Which, I’ll note, is why my last post did not provide precise numbers and was not worded in any definitive way (I talked about what the data “suggests” and said “if this were true” and that it “might underestimate” Jordan’s on-off “ if the difference were as big as it looks like”).

No, again, you are assuming that there are error measurements potentially going both ways. There are not. Your process is almost always inflationary. It does not match your intuition. You could test that out, but you refuse.

Here is another: 1998 Malone with +48 in 796 minutes -> +2.89 per 48 minutes. Basketball reference says +1.7 per 100 even though we know neither the Jazz nor Malone himself were a fast team (including by BRef’s own estimates). Time and time again, we see this is a clearly distinct process.

The truth is that we don’t really know for sure what, if any, pace differences on and off the floor there were for Jordan but, as I’ve said, we have no reason whatsoever to believe that difference would be the same direction and magnitude for Jordan as it is for LeBron, and so using data for LeBron that we know would incorporate this error (when we have data without the error) does not make sense since there’s no reason to believe that would cancel out any error from pace difference for Jordan (indeed, the equal-pace assumption might well *hurt* Jordan, rather than help like it would for the LeBron).

I literally showed you an example where Lebron played faster than his team did without him. You keep speculating that this is why there is such a regular distinction, and it is not.

As for the rest of it, you’re largely just pointing out that on-off models can be a bit wonky (especially at smaller sample sizes), likely due to possession estimations sometimes being inaccurate at small sample sizes.

At a fundamental level, on-off models are absolutely just trying to take +/- when “on” and divide by an estimate of possessions “on” and taking +/- when “off” and dividing by an estimate of possessions “off.” And that is exactly what I did, except with an equal-pace assumption to determine the breakdown between “on” and “off” minutes. So I don’t know what other error/bias you’re specifically asserting there is. Is your assertion that the possession estimate formula BBREF uses is different when calculating pace than it is when calculating on-off? That seems…quite unlikely, but if you have evidence of that then please show it.

I am telling you to check for yourself, because every time I highlight an obvious inconsistency, you ignore and dismiss it. This is not difficult: per 100 should be a reliably larger value than these per 48 hand calculations so long as pace is below 100, yet that is not what we see. So you are doing this process where you are regularly creating that increase, and when the disconnect is pointed out, you shrug and attempt to argue that maybe Jordan’s values would be even higher. I will reiterate: every time the pace is below 100, your process guarantees the per-100 number is higher than the per-48 number. That is not what happens in any of these “official” data sources you insist must be used once available.

Otherwise I don’t really see what source of error/bias you’re identifying beyond the equal-pace assumption, and we have no idea the direction or magnitude of that for Jordan (and have at least some indication it might be *hurting* his numbers). The bottom line is that if you think my method is biasing the data in a specific way beyond the equal-pace assumption, you need to explain exactly how rather than just declaring that it must be different without identifying the specific mechanism through which that’d be the case.

? What? No, I do not know the coding here, how would I know the specific mechanism. This is a quantifiable observation about what how your calculation process does not match. I am not even saying your math is bad or anything; just that it may as well be functioning as a different measurement. And it is illegitimate to compare different measurements like that.

Again, there is absolutely no “hurt”. When you multiply by 100/(#<100), the number increases.

Here is another: 1998 Jordan, +158 in 871.5 minutes, so +8.7 per 48. BRef says +8.8 per 100, so even if we assume that is secretly +8.84, Jordan could only be playing at a 98 pace. Again, every indication we have is they were a lot slower than that. 1997 Jordan, +136 in 803.7 minutes, so +8.12. Here, finally, we get something plausible at a BRef estimate of +8.8 per 100… but again, that is still at best a 92 pace, for a team estimated to have played at an 85.5 pace that postseason. It does not add up, and it pretty much never does. I do not know exactly why, but I can at least see it; can you?

I think there’s a bit of work here being done by the term “almost always.” For instance, I randomly looked at Kawhi in 2019 and he looked better in BBREF’s numbers than he would’ve using my method, and then I looked at him in 2017 too (full disclosure: I didn’t check other playoffs for Kawhi, so it’s possible this is unique to those years for him, I don’t know) and BBREF’s numbers were higher than my method would’ve gotten that year too. Given your language, I imagine you’ve seen other examples too, and I didn’t look hard at all. And, notably, I also looked at a fair number of random bench players on various teams, and more often than not they tended to look better in BBREF’s numbers too. That is actually suggestive of pace differences with stars on and off the court being a key factor here. Perhaps it’s the case that stars’ teams generally tend to play at higher pace with them off the court. That seems plausible from the numbers I’ve looked up, and would go a long way in explaining what you’re describing.

I take your point, though, and actually if it is true that stars’ teams generally tend to play at higher pace with them off the court (and that that’s at least part of why my method usually spits out a higher number for stars than BBREF does), then perhaps that gives us at least some reason to think that’s more likely to be the case for Jordan himself in the years where I’ve made that equal-pace assumption. The data we have from Jordan in 1997 and 1998 is suggestive of that too (though not necessarily to a massive extent), but that’s not super probative about other years since the Bulls teams in virtually all the other years were very different. It's also definitely worth noting that we can of course zero out this issue entirely and see that Jordan looks incredible in per-48-minute playoff on-off too. In fact, that's something I've already posted about—noting that in the data we have, Jordan’s on-off in per-48-minutes terms was +15.7 (and that goes up to more like +15.8 if we incorporate the 1986 playoff data homecourtloss provided with the mid-range estimate of game 1, though there’s some missing data there, so we don’t totally know). I think per-100-possession data is preferable, but it’s not like any argument about Jordan’s immense playoff on-off is dependent on it.

I do also take your point that BBREF’s numbers don’t always make sense. Like, for instance, I’m not really sure how John Stockton has a negative playoff “on” value in the 1998 playoffs when BBREF has a positive overall +/- for him. There’s some weird output like that that I find a bit difficult to reconcile with the fact that on-off is fundamentally just about dividing +/- on by possessions on and dividing +/- off by possessions off. I can’t imagine they’re actually intending to do something different than that, but there’s some weird stuff, and obviously my method would never come to a weird result like a negative “on” for someone with a positive +/-. This seems to raise a broader discussion about why BBREF’s on-off data has these weird instances. It actually makes me wonder if the +/- they list is not the +/- they use in their on-off data (like maybe they treat FTs differently between the two or something). I’d generally say that that sounds unlikely, but I’m struggling to figure out any explanation for 1998 Stockton.

[lessthanjake]

Also, the 1991 Bulls were a +9.5 NRtg (i.e., per 100) team. IF Jordan wound up with something like +9.85 ON for the season, the Bulls could NOT be a -9.77 with him off and still wind up a +9.5 team.

As I noted in my original post, the Bulls games in the Squared sample for 1991 were actually substantially less good for the Bulls than the full season. The Bulls only lost one of the unsampled games and had a bunch of blowout victories in the unsampled games. So this is not surprising at all. And it relates to something that I noted from the very beginning in my OP. As I mentioned in my OP, I’m not sure how it cuts in terms of bias, but my general inclination is to think that Jordan would probably look better in a better sample of games. We don’t know though, of course, and sampling error is definitely an issue with the Squared stuff and I think is a particular concern with the 1990-1991 season data since we know it’s not really a representative sample of games.

In the 56 sampled games, what was the Bulls NRtg? If we know this, then we can know what their NRtg was in the other 26 games (more or less since we’d assume the same pace) since we know the season NRtg has to be +9.5. The numbers in the others games would have to be WILDLY different because you already have 56 of the 82 games played leaving us with only 26 games left of possessions and ON minutes and OFF minutes. We know that:

—The 1991 Bulls were a +9.5 team in the regular season
—Jordan played 3,034 total minutes and apparently in the 56 game sample was +9.85 per 100
—(If we extend the 56 games sample, then the on-off would be +9.85 ON, +8 ON-OFF, +1.85 ON-OFF to get to a +9.5 NRtg for the team)
—We know the total off minutes given a full 82 game schedule plus the overtimes they played

If I assume that over the 56 game sample that Jordan played 37 minutes a game (like he did over the course of 1991 over 82 games), then:

—If Jordan was +14 per 100 on court in these other 26 games, then the Bulls would be about +3.8 per 100 in the off minutes, and Jordan would be about +11.2 ON, +3.8 OFF, +7.4 ON-OFF assuming he was playing the same amount in the other games though a cursory glance shows that he played fewer minutes in the blowouts

—If Jordan was +20 per 100 on court in these other 26 games, then the Bulls would be about -2.3 per 100 in the off minutes, and Jordan would be about +13 ON, -2.3 OFF, +15.3 ON-OFF

—If Jordan was +25 per 100 on court in these other 26 games, then the Bulls would be about -7.1 per 100 in the off minutes, and Jordan would be about +14.6 ON, -7.1 OFF, +21.7 ON-OFF

^These last two given the minutes/possessions adjustments in blowouts would likely have to be something like +27 per 100 to get to that final ON-OFF.

I’ve not calculated their net rating in those games, but I can tell you that I previously calculated that their MOV in the 56 games sampled was like +6, and we know the average MOV for the season was +9, so the Bulls outscored opponents by about +15.5 per game in the unsampled games (which suggests a bit over a +16 net rating). I don’t think it’s at all implausible that Jordan’s “on” rating in those other games was up towards the +20 or +25 range. In fact, that seems more likely than not the case. And if it was then that’d be a very high “on” value overall. And given other seasons (as well as the fact that there’s a +10 SRS season we don’t have data on where the “on” value is almost certainly very high), I really don’t think we can reasonably say that Jordan is lacking in seasons with really high “on” values compared to other players, except maybe Steph Curry.

Jordan’s “on” number is surely substantially higher for the entire season, because, as noted above, the unsampled games were a lot better for the Bulls than the sampled games. So this isn't really a particularly valid point. I do agree, though, that the super high on-offs in Jordan's earlier years aren't as impressive as they would be if they were with a higher "on" value. Which is something I actually flagged in another thread when I noted that Jordan's four-year peak on-off in the data we have is higher than any four-year period I'm aware of for others but that Steph is fairly close but with a higher "on." In any event, Jordan does have some fantastic “on” values in the full data we have for the 1996 and 1997 regular seasons, so he’s not lacking in those sorts of seasons.

Even if the ON is higher (and almost surely it is), and we take the most optimistic view, it’s not an outlier peak. It’s very, very good like we have for others, and amongst the GoATs but not some mythical outlier most impressive peak as those were the claims along the lines of “most impressive/impactful peak” but the data doesn’t bear that out.

I think the best guess we could have is that the 1990-1991 season has a combination of “on” value and total on-off that has almost never been matched. Again, you seem to be suggesting that Jordan needs to meet some “mythical outlier” standard that you don’t apply to anyone else. Something doesn’t need to be a “mythical outlier” to be the most impactful peak. Jordan doesn’t need to be a “mythical outlier” to have been better than everyone else. If Jordan had about a +21 on-off in a four-year period, then that’s *definitely* got a good case for most impactful peak (with Steph being the other candidate for it, due to his higher “on” value), even if you don’t define it as a “mythical outlier.”

Also, the claim was that 1991 and also 1988 through 1991 is the peak; referring to 1997 and 1998 takes us back to previous discussions about mid Jordan or later Jordan having perhaps more impact. Additionally we have RAPM data for 1996-1998 and that also doesn’t bear out any outlier peak.

The overall “on” value for his peak years of 1988-1991 would never be super high because his team for most of that time period was nowhere near good enough to have super high “on” values (and even in 1991, we have good reason to believe they were very negative without him). But of course the data we have also suggests that his on-off in those years was up around +21. Is that +21 with a +8 “on” as good as a +21 where the “on” number is like +18? No, it’s not. But a +21 on-off over four years is still extraordinary and unprecedented. I’m not aware of any four-year peak that competes with that except for Steph having a +17 on-off with like a +14 “on” value. Steph’s might arguably even be preferable in sheer on-off terms due to the higher “on,” but at worst that’d put Jordan at the #2 peak in impact terms, and of course in box terms Jordan’s peak obviously clears Steph. If you want to argue Steph Curry has the #1 peak and Jordan has the #2 peak, then I’m okay with that though. It’s a defensible position, given Steph’s impact. But Steph’s not really competing with Jordan in all-time terms.

I don’t want to go down the RAPM rabbit hole on Jordan since I think the points have been clearly made already in other threads, but I just want to note that Jordan absolutely looks incredible in RAPM data that we have. He’s basically always 1st (or occasionally 2nd behind a random non-star) all the time in every regular season and playoffs we have data on, except for his rookie season and his final regular season with the Bulls (and even in those he’s still one of the top several). And that’s not a surprise, since that RAPM data is ultimately based on this same incredible on-off stuff (obviously with teammate adjustments, as RAPM does).

Again, I am not creating any artificially high bar for Jordan—you made the claim about the most impressive peak, etc. Again, now you’re also going back to on/off when before which you implied wasn’t as impressive for KG as the ON number in another thread.

You absolutely are creating an artificially high bar for Jordan. Your demand for Jordan is pretty obviously that in order to be GOAT-like, his numbers must be a “mythical outlier” that is well above anyone else ever. But obviously one doesn’t need numbers to be WAY above everyone else to be the GOAT. If you did, then there’d just be no GOAT at all, because no one has that.

I think you’re misrepresenting what I said about Garnett. In this and other posts, you keep suggesting that I’ve argued that we should only care about the “on” number and not the on-off. That’s never been what I’ve said at all. What I have said (and indeed what I said in the stuff you quoted) is there’s diminishing marginal returns on “on” values, so it is harder to get a high on-off with a good team than with a bad one. We need to keep that in mind when evaluating on-off. That means that if two players have a very similar on-off but one has a better “on” number, then I’d say the player with the higher “on” value has more impressive numbers. So yeah, where Robinson and Garnett had essentially the same on-off, but Robinson’s came with a higher “on” value, I’d say Robinson’s numbers were better. But at a certain point a higher on-off with a lower “on” is more impressive if the on-off is enough higher. I haven’t pinpointed exactly where that line is in my mind, but I’ll say that if Jordan rocked a +21 on-off over a four-year span, then it’s mighty hard to think of any four-year stretch that was more impressive except for maybe a Steph stretch that had lower on-off but notably higher “on.” LeBron certainly doesn’t have a similar stretch that’s as impressive IMO.

Also, to be clear, when I said Garnett’s on-off with the Timberwolves isn’t *that* impressive, I was talking about the whole time period, not specifically the 2003 and 2004 seasons where he had his best on-off numbers. Those two seasons were definitely very impressive in on-off terms. Not as impressive as they’d be with higher “on” values, but nevertheless in the top tier of impressive in all the on-off data we know about. But Garnett’s overall on-off with the Timberwolves wasn’t as good as that. The overall picture (or even just the prime-years picture) is something like a +12 or +13 on-off with a +3 or +4 on. That’s still really good, but I think it was perfectly justified to say that that’s not *that* impressive, particularly in the context of a discussion comparing his and Steph Curry’s impact.

I don’t actually think that Robinson’s +18.7 and Garnett’s +18.2 are super alike, and that’s kind of my point. Robinson’s +18.7 comes with around like a +10.3 net rating when he was on, while Garnett’s +18.2 comes with a +7.2 net rating when he was on. To me, the Robinson numbers are substantially better, since it is substantially easier to raise a team from really low numbers and substantially harder to raise it to the +10 sort of territory.

And even leaving aside raw achievement, I’ll repeat again that there’s diminishing marginal returns on things like on-off. It is definitely easier to raise an awful team a lot in terms of on-off than it is to raise a good team. And that also means it’s easier to accrue high RAPM on a bad team—because ultimately RAPM is aiming to isolate the effect you had on your team’s net rating and you can have a bigger effect on your team's net rating if the baseline for the team is bad. So, to me, I don't really find Garnett's on-off numbers or RAPM with the Timberwolves all *that* impressive.

I am impressed by both.

Let’s say that the more important ON (your argument) +7.63 ON from 1988 to 1991 is really +8.5 or +9.5 or +10 over this stretch. That is really impressive but not outlier peak at all when we have several,players likely around the same or higher.

As far as the “player voted #1,” from 2012 to 2017 (550+ RS+PS games) James was ~+9.4 to +9.8 ON court.

2015 to 2017 (278 games RS+PS), was +10.26 ON, -6.74 OFF, +17 ON-OFF. +17 and +21 might different enough but the +10.26 ON (what you imply has more value) is probably as high as Jordan’s optimistic ON from 1988-1991 (your calculations).

2008 to 2010 (260 RS+PS games), playing with that supporting cast is +8.99 ON. 2009-2010 (182 RS+PS games) is +12.28 ON. 2008 to 2010 + 2015 to 2017 (520 RS +PS) is +9.63 and so on.

Also, remember the optimistic “Jordan was +25 per 100 (more like +27 per 100) on court in these other 26 games, then the Bulls would be about -7.1 per 100 in the off minutes, and Jordan would be about +14.6 ON, -7.1 OFF, +21.7 ON-OFF

Well, these numbers look VERY familiar from a 2009 season, no?

So, being consistent: big on-offs are impressive especially when backed with lineup data (not just impressive for certain players), big on court numbers are impressive (especially when backed by lineup data), but these runs by these players are witching the same vicinity though some did it for longer.

I don’t see any outlier anything, “highest peak,” etc.

Again, I’ve never said that the “on” is “more important.” That’s a misrepresentation of what I’ve said. I’ve said that our evaluation of on-off has to keep in mind that a given on-off is more impressive with a higher “on.” Evaluating these things requires applying a balancing test looking at both the on-off and the on. Again, I can’t tell you an exact formula for that balancing test, but I can definitely say that peak LeBron does not have any four-year stretch that looks as impressive in this regard as the data we have on Jordan’s four-year peak looks. Peak LeBron (2009-2012) was at about a +14 on-off with a +11 “on.” I think we can all agree that that’s not as impressive as what we have on peak Jordan (of course with the caveat that it’s not full data for Jordan so there’s potential sampling error).

As for 2009, yeah the 1991 numbers would probably end up fairly similar to that year. But Jordan’s on-off in the 1991 playoffs is substantially better than LeBron’s in the 2009 playoffs, so there’s a pretty good chance 1991 would look like superior impact overall (though we don’t really know for sure). And this is also something of a non-sequitur, since I was talking about Jordan’s peak four years.

And, again, your demands for “outlier” on-off numbers from Jordan are just artificial demands that you’d never apply to others (including LeBron James).

[lessthanjake]

I don't know about you guys, but when I read these posts, I lose interest really fast. Maybe it's the black text?

What are we trying to do here... Computing On/Off ratings for players with respect to the teams? IE: Bulls +14 when Jordan ON versus -14 when Jordan OFF... something like that?

Shouldn't I care of he's being D'd up by Long-Arms McGee instead of Waterboy Williams in those situations?

Yes, we should. But we don’t have that data (well, actually, I suppose you kind of do for some of these games!). I think this comment goes to the fact that on-off data definitely needs to be taken with a grain of salt. I wouldn’t think anyone discussing things on this thread would disagree with that. But on-off numbers are relevant/interesting data points that are at least probative regarding a player’s impact, even if we know that they’re flawed and don’t account for a boatload of context. People still talk about players’ on-off a lot, and so I do think it’s interesting to piece together data on that for older players. We shouldn't over-index on it though.

[Djoker]

So I ended up adding the estimates for 1986 (3 games) and 1995 (10 games).

Total Playoff On-Off per 100 Possessions with 1985, 1986 + 1988-1993 + 1995-1998
- On: +7.55
- Off: -7.16
- On-Off: +14.70

This number includes Jordan's entire playoff career except 1987 (3 games).

EDIT: I checked the numbers in the OP and I think 1985 Playoffs has a mistake. The OFF number is -75.6 per 100.

[lessthanjake]

So I ended up adding the estimates for 1986 (3 games) and 1995 (10 games).

Total Playoff On-Off per 100 Possessions with 1985, 1986 + 1988-1993 + 1995-1998
- On: +7.55
- Off: -7.16
- On-Off: +14.70

This number includes Jordan's entire playoff career except 1987 (3 games).

EDIT: I checked the numbers in the OP and I think 1985 Playoffs has a mistake. The OFF number is -75.6 per 100.

The numbers I get when including the rough 1995 estimate (using the *worse* of the two 1995 estimates—which makes a big difference since the different estimates using the two charts that include 1995 spit out *super* different results) are similar to that but slightly different. I get as follows:

- On: +7.43
- Off: -7.51
- On-Off: +14.94

Given the various forms of error in the data, these are functionally the same estimates though. I assume yours are slightly different because of amalgamating the numbers slightly differently (or perhaps because of the 1985 thing—discussed below). Either way, though, we’re basically at +15 on-off and only without 1987. And, again, that’s using the by far worse estimate for 1995 (which is so different that, for instance, the “off” value in 1995 is almost +20 in what I used above but more like +8 if we used the other 1995 derivation we could use).

________

As for the 1985 playoffs, I don’t think my OP is wrong. I think what’s happening is you’re getting to a different result because you’re not using the “on” possessions data that actually have for that playoffs (which leads to a different number of estimated “on” and “off” possessions than if we just used the minutes and made an equal-pace assumption). Basically, because Squared provides “on” possessions data, the way I got that number was different than for the other playoffs. So, basically, here’s where the number comes from:

We know from the Squared data that the Bulls were +10 with Jordan on the court. And since they were -22 overall, that means they were -32 with Jordan off the court. If we add up the offensive and defensive possessions in the Squared data, we get an estimate of 360 possessions with Jordan on the court. If you go game-by-game and add up the pace BBREF reports for each game, you get an estimate of 387.2 total possessions in the series. Which means we can estimate 27.2 possessions with Jordan off the floor (which, if anything, might actually be high because BBREF generally gives slightly higher numbers of possessions than Squared, and we’re getting the “on” possessions from Squared). Given that they were -32 in those possessions, that gets us to an “off” estimate of -117.65 (because (-32/27.2)*100 = -117.65).

[Djoker]

So I ended up adding the estimates for 1986 (3 games) and 1995 (10 games).

Total Playoff On-Off per 100 Possessions with 1985, 1986 + 1988-1993 + 1995-1998
- On: +7.55
- Off: -7.16
- On-Off: +14.70

This number includes Jordan's entire playoff career except 1987 (3 games).

EDIT: I checked the numbers in the OP and I think 1985 Playoffs has a mistake. The OFF number is -75.6 per 100.

The numbers I get when including the rough 1995 estimate (using the *worse* of the two 1995 estimates—which makes a big difference since the different estimates using the two charts that include 1995 spit out *super* different results) are similar to that but slightly different. I get as follows:

- On: +7.43
- Off: -7.51
- On-Off: +14.94

Given the various forms of error in the data, these are functionally the same estimates though. I assume yours are slightly different because of amalgamating the numbers slightly differently (or perhaps because of the 1985 thing—discussed below). Either way, though, we’re basically at +15 on-off and only without 1987. And, again, that’s using the by far worse estimate for 1995 (which is so different that, for instance, the “off” value in 1995 is almost +20 in what I used above but more like +8 if we used the other 1995 derivation we could use).

________

As for the 1985 playoffs, I don’t think my OP is wrong. I think what’s happening is you’re getting to a different result because you’re not using the “on” possessions data that we actually have for that playoffs (which leads to a different number of estimated “on” and “off” possessions than if we just used the minutes and made an equal-pace assumption). So, basically, here’s where the number comes from:

We know from the Squared data that the Bulls were +10 with Jordan on the court. And since they were -22 overall, that means they were -32 with Jordan off the court. If we add up the offensive and defensive possessions in the Squared data, we get an estimate of 360 possessions with Jordan on the court. If you go game-by-game and add up the pace BBREF reports for each game, you get an estimate of 387.2 total possessions in the series. Which means we can estimate 27.2 possessions with Jordan off the floor (which, if anything, might actually be high because BBREF generally gives slightly higher numbers of possessions than Squared, and we’re getting the “on” possessions from Squared). Given that they were -32 in those possessions, that gets us to an “off” estimate of -117.65 (because (-32/27.2)*100 = -117.65).

Yea I used the worse 1995 estimate too, the -1.9 ON and +19.4 OFF.

For 1985 it totally escaped me that squared2020 tracked the exact possessions. I used the BRef pace. Yes your number makes sense. In fact the 1985 ON-OFF is the most reliable after the official data for 1997 and 1998 postseasons because the possessions were tracked.

I will recalculate everything again. I don't understand why my ON value is a bit off from yours. It shouldn't happen.