Some Historical Plus-Minus

[Squared2020]

.

[WestGOAT]

Definitely doing God's work, and I'm actually pretty sure he can get a college intern to do the gruntwork since he's working for/with the league office. If that's what he wants is another story of course

The biggest challenges with that:

- NDA's on many of the games. it was really useful to get data from games on YouTube. With one private collector, he fancied what I was doing, so about 8 years ago he had me sign an NDA that I wouldn't even mention which games he had (or who he was in relation to the games). Just trying to separate out those games would be painful; and that's a significant chunk of my repository.

Maybe make the intern sign NDA's as well, If I understand the issue correctly? But I can imagine private collectors being really finickily about their games they probably went through great effort to obtain.

[ceiling raiser]

I have between 200 and 400 games for each season, rolling back to about 1987. Then it dwindles pretty fast

Bump with a quick question - by 1987 do you mean 86-87, or 87-88? Thanks again for this project.

[Squared2020]

.

[MrVorp]

This work is incredible. I’m getting roughly +24 On/Off per 100 for Jordan in 88’, so he MIGHT have a chance to break the On/Off record. Still a decent amount of games not sampled. Also looks like 91’ Jordan sits around +18 in the games sampled.

[lessthanjake]

This work is incredible. I’m getting roughly +24 On/Off per 100 for Jordan in 88’, so he MIGHT have a chance to break the On/Off record. Still a decent amount of games not sampled. Also looks like 91’ Jordan sits around +18 in the games sampled.

I did some calculations on this myself and I get similar numbers but slightly higher (though the difference may be that I am looking at per 100 possession numbers). Specifically, I got a +26.1 on-off for Jordan in these 1987-1988 games and a +19.3 on-off for Jordan in the 1990-1991 games. I also got +24.14 for Jordan’s rookie year, though it is down to +17.76 if you take out the playoff games that are in the data set.

Just for transparency’s sake, here’s my methodology:

I counted all the games Squared provided, except where there were games that clearly only a portion of the game was tracked for. So, for instance, in the 1987-1988 data for Jordan, there’s a game where the score is listed as 20-16. I didn’t count that. I then counted the total +/- with Jordan on, and got the +/- when he was off by comparing that to the actual total margin of victory in the games. We can get the per-100-possessions “on” number for Jordan pretty easily by totaling up the number of possessions listed in Squared’s data. Specifically, since the offensive and defensive possession numbers are sometimes slightly different, I added together the offensive possessions and defensive possessions and divided by two to get the number of possessions he was on the floor for. To get the number of possessions he was off the floor for, I aimed to estimate how many total possessions there were in these games. One could potentially actually go to basketball-reference and total up the pace in each of these individual games. I think that would be most accurate, and maybe I’ll eventually do it, but for now I was admittedly too lazy to do that. What I did instead was look at the Bulls’ overall pace for the season and multiply by the number of games in Squared’s data. That *should* give a good estimate of the number of total possessions in those games. One small adjustment needs to be made to that, because pace is a per-48-minute stat, and there are overtime games that last more than 48 minutes, so I added extra total possessions based on the number of OT periods there were in the set of games tracked (so, for instance, the estimated number of possessions added by one OT period would be 5/48*PACE). Once I had the total possessions, I could subtract out the possessions with Jordan on, in order to get the number of possessions with him off. And once we have the +/- with Jordan on, the number of possessions with Jordan on, the +/- with Jordan off, and the number of possessions with Jordan off, you have all the information you need in order to derive an on-off per 100 possessions.

I’ll note that, to sense check things, I ran this same methodology on the 1995-1996 Squared data for Jordan, to see whether the output looked completely different from the known on-off data we have from that year. And it did not. The method spit out a +16.78 “on” net rating for Jordan in 1995-1996, with a +0.02 “off” net rating. The actual net ratings for the full season were +16.7 on and +1.5 off. So this was really close. Obviously it’s not going to be exactly the same because it does not include the full sample of games, but the fact that it came out very close is suggestive of the method probably not being super fundamentally flawed.

[lessthanjake]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

[DraymondGold]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

It would be cool at some point to get together a post that has all the on/off per-season data we have for Jordan (both regular season and postseason) in one spot. So something like...

Jordan's on/off:
...
1988 Regular Season: +X [Source: Squared2020, sample size: Y games]
1988 Playoffs: +Z [Source: that Thinking Basketball video, sample size: all games]
1989 Regular Season: +B [Source: Squared2020, sample size: C games]
...
1997 Regular Season: +D [Source: NBA.com, sample size: all games]
1997 Playoffs: +E [Source: NBA.com, sample size: all games]
...

etc. Obviously not on you to put together (it would take some time), but it would be fascinating to see all in one spot! For the stuff that has some uncertainty, e.g. the Thinking Basketball video might require looking at those graphs by eye, or the on/off you've listed from the regular season has some slight uncertainty too, it might be nice to list a rough uncertainty too.

[eminence]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

May I ask the Bulls overall MOV in those games?

[lessthanjake]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

It would be cool at some point to get together a post that has all the on/off per-season data we have for Jordan (both regular season and postseason) in one spot. So something like...

Jordan's on/off:
...
1988 Regular Season: +X [Source: Squared2020, sample size: Y games]
1988 Playoffs: +Z [Source: that Thinking Basketball video, sample size: all games]
1989 Regular Season: +B [Source: Squared2020, sample size: C games]
...
1997 Regular Season: +D [Source: NBA.com, sample size: all games]
1997 Playoffs: +E [Source: NBA.com, sample size: all games]
...

etc. Obviously not on you to put together (it would take some time), but it would be fascinating to see all in one spot! For the stuff that has some uncertainty, e.g. the Thinking Basketball video might require looking at those graphs by eye, or the on/off you've listed from the regular season has some slight uncertainty too, it might be nice to list a rough uncertainty too.

I’ve actually been aiming to put together something like this! Not sure when I’ll be able to actually get through doing it, but I’m hoping to do so (and have done much of the heavy lifting for it in terms of calculations already, while posting in this thread and another recently).

[lessthanjake]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

May I ask the Bulls overall MOV in those games?

In the 1987-1988 sample, I counted a total MOV of +104 (in 42 games, so +2.48 per game). In the 1990-1991 sample, I counted a total MOV of +338 (in 56 games, so +6.04 per game). The total season average MOV in 1987-1988 was +3.4 and the total season average MOV in 1990-1991 was +9.0. So basically, both samples of games were a bit worse than the overall season sample for the Bulls, in particular the 1990-1991 one.

Not sure how that cuts in terms of what direction it might bias Jordan’s on-off. The 1987-1988 sample isn’t super different from the overall sample so that’s probably not a big deal, but the 1990-1991 average MOV in the sample is actually pretty different from the full sample, which indicates that the remaining 26 games that were not sampled were significantly better for the Bulls. And, indeed, one can glance at the games not sampled and see that it includes a lot of blowouts for the Bulls. Not sure how that’d cut in terms of Jordan’s on-off. In general, I’d think blowouts help a star player’s on-off, since it tends to really juice up their “on” value, and the net rating when they’re off the court is often not actually good since there’s garbage time where the winning team usually doesn’t do well. But who knows—it’s also possible that the Bulls did better in those unsampled games because they did really well in Jordan’s “off” minutes.

[MrVorp]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

Well, so much for Jordan’s impact peaking during the second three peat.

[LukaTheGOAT]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

Well, so much for Jordan’s impact peaking during the second three peat.

Generally, the only people who push such a notion are doing so, in order to downplay his true peak.

[Harry Palmer]

A few points:

1) this is an astonishingly fantastic discussion. At one point I was sort of regarded as a data-head here, but long before I lost interest I was already getting out of my depth with hard analysis, just never had too much interest or any special abilities with quantitative data so as it became prevalent and I stayed ~ where I’d started I was kinda running backwards standing still, and reading this I feel pretty obsolete.

2) that said I feel there are a couple of generalized points re: Hakeem and data. The first has already been mentioned in terms of how less acute/invariably contextualized defensive metrics are and how much of Hakeem’s real value was contingent on 48 minutes* of being dominant at both ends. Then there is the degree to which his actual defensive play will be obvious watching but whose numerical credit will be dissipated in other’s yield. For example the regularity with which he flashed to prevent a post entry, closed out on the secondary valve n the perimeter to force an error or invite the seemingly open bypass, then was back to protect the rim all in a matter of a few seconds almost as a kind of one man trap. The degree to which everyone who played with him talks about how unicornish he was in this specific regard is pretty eloquent. Personally the only guy I’ve ever seen who came close to even showing the capacity to do this was KG, but he had neither the power nor the post-anchoring role that Hakeem did to fully realize this impact, nor did he do it with a fraction of the regularity. It’s just that he had the capacity that stands out. Possibly Russell playing in a more advanced era would also be here, I cannot tell from what we see on film. So that’s just one example of how do you measure the value of doing something few others would even try?

3) Next, the degree to which any standardized method automatically homogenizes, and least accurately represents uniqueness. It’s not remotely an issue with the rubric or the people applying it, it’s just the nature of any broadly applied measurement requiring/generating its own gravitational force. Think of how much more accurately scholastic rubrics are tailored to differentiate in the middle/mass rather than at either extreme, Of course this significantly reduces objectivity and leaves analysis wide open to biased ideas of who was/was not the exception, so there’s a futility in highlighting this blind spot, and I don’t feel qualified or unbiased enough to die on this hill. But it definitely helps me understand aspects like almost every player who played with Hakeem and, say, Shaq or Robinson or Duncan consistently saying that not only was Hakeem the best, but that it wasn’t particularly close. And they almost immediately launch into examples like the defensive play I mentioned earlier where he is the primary disruptive force at three different levels on the same play, and how much easier that made their job. The steals/blocks thing is kind of the caveman drawings that indicate but do not especially prove the kind of thing I’m talking about here.

4) Normally years back I would spend a lot of time at this point highlighting how extreme his lack of supporting cast was relative to any comparable peer but to my delighted surprise in my absence this seems to have become fairly understood. I wish I knew how, but I’ll take it. I’m less sure we have figured out a way to quantify that, but my limitations with data analysis don’t provide any insights into how to move that discussion forward. So call it a big asterisk but not a complete argument.

5) And I guess the last point I’d make is the degree to which his making everyone around him’s job easier at both ends, less energy intensive and therefore logically yielding them greater reserves that don’t necessarily pay off in a way that’s reflected in any stat you can assign him, or even necessarily require his being on the floor when it does, is again more likely to serve as an impressionist painting rather than a photograph. But again it is a point his peers repeatedly raise, it’s the multidimensional aspect thar for example both Kenny Smith and Michael Jordan raise when calling him the best ever, from opposite points of view. And though I think Hakeem is definitely a highlight friendly player in terms of unique skill sets, the dominant feature that stands out to me that separates what watching him felt like compared with anyone else is the degree to which he was the overwhelming and sideline to sideline focus of play at both ends, relentlessly, up and down and back up again ad infinatum.

I suppose this might mean that my best future objective advocates for this impact is if biometric observation ever approaches baseball’s level, but the nature of their controls makes that seem far off. And so to appreciate that you have to watch entire games and compare that with watching others. For fairness sake I would offer that doing so might also reveal less complimentary angles, for example the degree to which, imo, his passing is overrated, consisting as it did with fairly basic kicks due to the auto-double and triple teaming, and the Rockets being kind of the forerunner of the inside out 3 point game that would become much more sophisticated in other hands. Not necessarily a point against him, if the simple pass is there, take it, but I think the impression of his is exaggerated because of how much that was the bread and butter of their offence when it got to the championship runs.

I’ll close with something that obviously impresses people because it’s mentioned a lot but usually ends up in an ad-hom cul-de-sac about character or w/e and which I have no intention of contradicting but am somewhat more interested in what it tells us about his physical capabilities, and that’s the whole Ramadan thing. In case you’ve somehow missed it, Hakeem’s religion and piety meant that for an entire month in the middle of the season he would neither eat nor drink anything between dawn and dusk. That…should not be able to happen. I mean not without a huge impact. And who knows, maybe it did, maybe that’s part of the idea of him being a playoff raiser, maybe it’s just that Ramadan is in the rear view come the playoffs. And there’s an argument here that if the last point is true, that still represents a loss in value to his team who receives no secondary value in his doing that. I’d accept that, if proven to be true, that knocks him down a peg or two regardless of how unrelated to ability it is.

It might be an interesting project to evaluate his output during that month compared with the rest, with full context like game times, etc. but that’s way beyond my ambition. But maybe someone with a lot more skill than I possess can do that more easily. Anyways, the point I’m making about that is not how it’s distinct from ability but rather what extra information it gives us on ability, specifically endurance. Been around pro sports my whole life and nothing I’ve ever learned suggests it’s doable, the margins at elite levels are way too narrow to allow for what should be a fairly crippling practice. But if it made him worse no one has ever had the temerity to suggest it, and some if his career games happened during that period. But the point I am making is how this dovetails with that 3-d end to end impact he had game after game…to be unique in that way would probably require either ~ superhuman endurance or some kind of unicornish energy efficiency. And I guess maybe his coping mechs for surviving that while being a constant focal point of both teams may even have informed his play when back on healthy practices. But imo it lines up enough to suggest that in addition to his basketball-built physique (in particular relatively short legs allowing for a small man’s footwork and long, long arms allowing for a big man’s space occupation, his at the time pretty unique sports background in sports sceptically dependant on footwork, eye hand/eye ball coordination and multiple jumping (though to be fair here volleyball was a lot more common for basketball players than soccer or handball, for example I think Wilt was considered at or near world level) we have a freakish engine and we are left with what I feel was a perfect storm undermined by a lot of terrible externals beyond his control.

Anyways, my thanks to the contributors to this thread, you have my respect and admiration, even if you’re making me feel like a bit of a dinosaur. Cheers.

*obviously not really 48, for effect.

[DraymondGold]

I’ll close with something that obviously impresses people because it’s mentioned a lot but usually ends up in an ad-hom cul-de-sac about character or w/e and which I have no intention of contradicting but am somewhat more interested in what it tells us about his physical capabilities, and that’s the whole Ramadan thing. In case you’ve somehow missed it, Hakeem’s religion and piety meant that for an entire month in the middle of the season he would neither eat nor drink anything between dawn and dusk. That…should not be able to happen. I mean not without a huge impact. And who knows, maybe it did, maybe that’s part of the idea of him being a playoff raiser, maybe it’s just that Ramadan is in the rear view come the playoffs. And there’s an argument here that if the last point is true, that still represents a loss in value to his team who receives no secondary value in his doing that. I’d accept that, if proven to be true, that knocks him down a peg or two regardless of how unrelated to ability it is.

It might be an interesting project to evaluate his output during that month compared with the rest, with full context like game times, etc. but that’s way beyond my ambition. But maybe someone with a lot more skill than I possess can do that more easily. Anyways, the point I’m making about that is not how it’s distinct from ability but rather what extra information it gives us on ability, specifically endurance. Been around pro sports my whole life and nothing I’ve ever learned suggests it’s doable, the margins at elite levels are way too narrow to allow for what should be a fairly crippling practice. But if it made him worse no one has ever had the temerity to suggest it, and some if his career games happened during that period. But the point I am making is how this dovetails with that 3-d end to end impact he had game after game…to be unique in that way would probably require either ~ superhuman endurance or some kind of unicornish energy efficiency. And I guess maybe his coping mechs for surviving that while being a constant focal point of both teams may even have informed his play when back on healthy practices. But imo it lines up enough to suggest that in addition to his basketball-built physique (in particular relatively short legs allowing for a small man’s footwork and long, long arms allowing for a big man’s space occupation, his at the time pretty unique sports background in sports sceptically dependant on footwork, eye hand/eye ball coordination and multiple jumping (though to be fair here volleyball was a lot more common for basketball players than soccer or handball, for example I think Wilt was considered at or near world level) we have a freakish engine and we are left with what I feel was a perfect storm undermined by a lot of terrible externals beyond his control.

Anyways, my thanks to the contributors to this thread, you have my respect and admiration, even if you’re making me feel like a bit of a dinosaur. Cheers.

*obviously not really 48, for effect.

Interesting stuff Harry

I wanted to do a first-pass check of the Ramadan thing, so I looked at team Margin of Victory during Ramadan vs throughout the rest of the regular season. Obviously this is team performance, and it's not on a massive number of games per season, so we'd expect this to be a super noisy measurement. There's absolutely a more exact ways to do this (e.g. Check SRS rather than MoV since that adjusts for opponents, or actually start getting into the box stats and on/off data discussed in this thread)... I just wanted to do a first pass using Statmuse.

The results...

~Hakeem's Team's performance during Ramadan vs throughout the rest of the regular season~
(pre 1988 is exclusively in the playoffs)
1988: -2.2 (5 games) vs 1.39 full season average (-3.59 change)
1989: -1.38 (8 games) vs 0.95 full season average (-2.33 change)
1990: 3.92 (13 games) vs 1.46 full season average (2.46 change)
1991: 3.57 (14 games) vs 3.5 full season average (0.07 change)
1992: 0.07 (15 games) vs -1.72 full season average (1.79 change)
1993 9.93 (15 games) vs 4.23 full season average (5.7 change)
1994 2.58 (12 games) vs 4.32 full season average (-1.74 change)
1995 5.73 (15 games) vs 2.12 full season average (3.61 change) (Drexler joined for 8/15 games)
1996 -1.15 (13 games) vs 1.74 full season average (-2.89 change)
1997 -0.57 (14 games) vs 4.48 full season average (-5.05 change) (Barkley missed 8/14 games)

This doesn't jump out as a clear signal. On average, it's actually a slight positive bump in Ramadan. There are years where Hakeem's team performs worse, such as 1996, and 1988–89 on slightly smaller samples. But there's also years where they get better, such as 1990, 92, and 93.

Seems like not much of a change, at least at the team Margin of Victory level. Though this absolutely doesn't preclude a signal appearing if we looked closer. Regardless, it's super impressive stuff that Hakeem could perform at such a high level while fasting over the month of Ramadan!

[lessthanjake]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

It would be cool at some point to get together a post that has all the on/off per-season data we have for Jordan (both regular season and postseason) in one spot. So something like...

Jordan's on/off:
...
1988 Regular Season: +X [Source: Squared2020, sample size: Y games]
1988 Playoffs: +Z [Source: that Thinking Basketball video, sample size: all games]
1989 Regular Season: +B [Source: Squared2020, sample size: C games]
...
1997 Regular Season: +D [Source: NBA.com, sample size: all games]
1997 Playoffs: +E [Source: NBA.com, sample size: all games]
...

etc. Obviously not on you to put together (it would take some time), but it would be fascinating to see all in one spot! For the stuff that has some uncertainty, e.g. the Thinking Basketball video might require looking at those graphs by eye, or the on/off you've listed from the regular season has some slight uncertainty too, it might be nice to list a rough uncertainty too.

FYI, I took a stab at doing this. See the thread here: viewtopic.php?f=64&t=2314587

[Harry Palmer]

I’ll close with something that obviously impresses people because it’s mentioned a lot but usually ends up in an ad-hom cul-de-sac about character or w/e and which I have no intention of contradicting but am somewhat more interested in what it tells us about his physical capabilities, and that’s the whole Ramadan thing. In case you’ve somehow missed it, Hakeem’s religion and piety meant that for an entire month in the middle of the season he would neither eat nor drink anything between dawn and dusk. That…should not be able to happen. I mean not without a huge impact. And who knows, maybe it did, maybe that’s part of the idea of him being a playoff raiser, maybe it’s just that Ramadan is in the rear view come the playoffs. And there’s an argument here that if the last point is true, that still represents a loss in value to his team who receives no secondary value in his doing that. I’d accept that, if proven to be true, that knocks him down a peg or two regardless of how unrelated to ability it is.

It might be an interesting project to evaluate his output during that month compared with the rest, with full context like game times, etc. but that’s way beyond my ambition. But maybe someone with a lot more skill than I possess can do that more easily. Anyways, the point I’m making about that is not how it’s distinct from ability but rather what extra information it gives us on ability, specifically endurance. Been around pro sports my whole life and nothing I’ve ever learned suggests it’s doable, the margins at elite levels are way too narrow to allow for what should be a fairly crippling practice. But if it made him worse no one has ever had the temerity to suggest it, and some if his career games happened during that period. But the point I am making is how this dovetails with that 3-d end to end impact he had game after game…to be unique in that way would probably require either ~ superhuman endurance or some kind of unicornish energy efficiency. And I guess maybe his coping mechs for surviving that while being a constant focal point of both teams may even have informed his play when back on healthy practices. But imo it lines up enough to suggest that in addition to his basketball-built physique (in particular relatively short legs allowing for a small man’s footwork and long, long arms allowing for a big man’s space occupation, his at the time pretty unique sports background in sports sceptically dependant on footwork, eye hand/eye ball coordination and multiple jumping (though to be fair here volleyball was a lot more common for basketball players than soccer or handball, for example I think Wilt was considered at or near world level) we have a freakish engine and we are left with what I feel was a perfect storm undermined by a lot of terrible externals beyond his control.

Anyways, my thanks to the contributors to this thread, you have my respect and admiration, even if you’re making me feel like a bit of a dinosaur. Cheers.

*obviously not really 48, for effect.

Interesting stuff Harry

I wanted to do a first-pass check of the Ramadan thing, so I looked at team Margin of Victory during Ramadan vs throughout the rest of the regular season. Obviously this is team performance, and it's not on a massive number of games per season, so we'd expect this to be a super noisy measurement. There's absolutely a more exact ways to do this (e.g. Check SRS rather than MoV since that adjusts for opponents, or actually start getting into the box stats and on/off data discussed in this thread)... I just wanted to do a first pass using Statmuse.

The results...

~Hakeem's Team's performance during Ramadan vs throughout the rest of the regular season~
(pre 1988 is exclusively in the playoffs)
1988: -2.2 (5 games) vs 1.39 full season average (-3.59 change)
1989: -1.38 (8 games) vs 0.95 full season average (-2.33 change)
1990: 3.92 (13 games) vs 1.46 full season average (2.46 change)
1991: 3.57 (14 games) vs 3.5 full season average (0.07 change)
1992: 0.07 (15 games) vs -1.72 full season average (1.79 change)
1993 9.93 (15 games) vs 4.23 full season average (5.7 change)
1994 2.58 (12 games) vs 4.32 full season average (-1.74 change)
1995 5.73 (15 games) vs 2.12 full season average (3.61 change) (Drexler joined for 8/15 games)
1996 -1.15 (13 games) vs 1.74 full season average (-2.89 change)
1997 -0.57 (14 games) vs 4.48 full season average (-5.05 change) (Barkley missed 8/14 games)

This doesn't jump out as a clear signal. On average, it's actually a slight positive bump in Ramadan. There are years where Hakeem's team performs worse, such as 1996, and 1988–89 on slightly smaller samples. But there's also years where they get better, such as 1990, 92, and 93.

Seems like not much of a change, at least at the team Margin of Victory level. Though this absolutely doesn't preclude a signal appearing if we looked closer. Regardless, it's super impressive stuff that Hakeem could perform at such a high level while fasting over the month of Ramadan!

Fantastic, much appreciated. I agree it’s pretty opaque if it’s there at all, but also agree it might require more distillation. I know a lot of people think that his actual peak was earlier than perceived, that by the time of the rings he was already less explosive an athlete than earlier on, and the numbers were more about externals and system mesh than play. I think there might be some truth there, I know in his college and early career highlights he was much more above the rim, knock you down and send your shot into the cheap seats kind of guy, but I also think he just got smarter about how much he actually had to do it that way to have the effect he wanted, ie less wasted motion. And I suppose that might have come from learning to play through Ramadan, but that smells a bit like narrative.

One thing I should say is I don’t know for sure when he began observing that absolutely. I know he was well into it by the prime of his career, but one part of his story is that earlier on he was much more of a physical beast, dominating with power as much as skill, and had more of a temper than his fully evolved stoic gentleman spartan deal that Shaq found so unshakable/admirable. To a degree his growing impatience with management’s incompetence, coupled with some fairly dog-whistly comments about his religion/Africanness from above contributed to a ~ forgotten phase where it got pretty ugly, I think they accused him of nursing an injury out of grievance and there was even talk of trading him and were encouraging coaches to emphasize other players more, and none of those guys have been heard from since, it was pretty openly petty. And they **** talked him a lot ‘off the record’, saying he had a temper issue, wishing his religion was basketball not Islam, I mean pretty ugly stuff. It plays even worse when you know what kind of person everyone who played with him says he was, eg Kenny and Barkley both saying that he was ~ the only truly honourable player they ever met, how if your car broke down in the middle of nowhere and you had one call left on your phone he’s the guy, and his saying he’d be there…or at the rim getting your back if you’re worried about heavily pressing a guy…that was cast iron, he’d be there. Anyways, it mostly looks bad on them but they’re not around to give their side, so who knows, but it was certainly not a happy marriage.

Then saner heads prevailed and they found a coach who imo is nowhere near great but at least had the good sense to build an offensive system around his unique skill set and that got you B2B in fairly short order beginning with Otis Thorpe? as his Robin. Somewhere in there, according to legend, he decided to stop fighting against what he could not control and adopted the pious, stoic ‘honourable’ persona that by all accounts was 3-D and is still to this day, and what I cannot say is if strict observation of Ramadan came in here or was always part of him. I can say that a lot of people think he was pretty much always the guy he came to be known as and the blame rested pretty squarely on incompetent management, but I don’t know if that’s true, or how much. Anyways, this is my long winded way of saying that he might or might not have always observed, it was at least true for the bulk of his career in the limelight, but whether or not he did that as a Twin Tower I don’t know. Maybe someone from Houston as opposed to Toronto and who was actually watching him play in those years rather than reviewing him later would have more insight. And then I don’t even know in exactly what year this supposed road to Damascus moment happened, though my vague impression is not too long before the rings.

[OhayoKD]

Update on the above: I used the above methodology, but instead of estimating the total possessions in those games using the Bulls’ full-season pace, I added together the pace for each of the specific matches as reported on basketball-reference (and appropriately adjusting up the number of possessions in games with that went to OT, since, again, pace is a per-48-minute stat). Unsurprisingly, if I do that, the number of total possessions in those games is very similar to what I’d calculated before, but it is slightly different (in this case, slightly lower).

The result is that, now, in those 42 games in the 1987-1988 season, we can calculate even more precisely that Michael Jordan’s on-off was +26.93.

EDIT: I also did the same for the 1990-1991 season. Again, the numbers were unsurprisingly very similar. The more specific Jordan on-off number (using actual game-by-game pace as opposed to estimating using season pace) for the 56 tracked games in 1990-1991 is +19.62.

Well, so much for Jordan’s impact peaking during the second three peat.

Generally, the only people who push such a notion are doing so, in order to downplay his true peak.

So you two were downplaying his true peak when you quoted me with graphs where he "peaked" during the second-three peat?

Regardless, 88 being his impact peak also would make sense(he was doing the most there) and is the one i would expect to have the most regular-season impact across time. Will be interesting to see how Magic and Jordan matchup in the rest of squared's data. Magic has had a consistent advantage in wowy and attempts at replicating RAPM but perhaps actual apm will offer a counterpoint

[rk2023]

I ran through 1988 and 1991 Magic at a higher level (didn't adjust for pace or minutes yet, and filtered out the partial games).

1988 (48 Games):
8.38 +/- Per-Game with Magic, -2.36 +/- Per-Game without

1991 (47 Games):
5.38 +/- Per-Game with Magic, -1.83 +/- Per-Game without

The former is more in-line with the full 82 MOV/game Lakers results, FWIW.

[homecourtloss]

I ran through 1988 and 1991 Magic at a higher level (didn't adjust for pace or minutes yet, and filtered out the partial games).

1988 (48 Games):
8.38 +/- Per-Game with Magic, -2.36 +/- Per-Game without

1991 (47 Games):
5.38 +/- Per-Game with Magic, -1.83 +/- Per-Game without

The former is more in-line with the full 82 MOV/game Lakers results, FWIW.

Interesting. Can you convert to per 48?