Do you think Pippen ever deserved 1b status on a title team?

[lessthanjake]

Alright, then how about we do another one. This time from 1988 when Jordan was the leading shotblocker on his team and 16th in blocks in the whole NBA. The first full game that shows up from CHI vs DET on youtube is Game 3 where Jordan was one of two Chicago players to record a block:

Watch on YouTube

Spoiler:

Distribution went

Oakley 13
Corzine 9
Pippen 8
Grant 6
Jordan 3
Sam Vincient 2
Rory Sparrow 1
Elston Turner 1

(Doesn't add up exactly to 40 as there were a couple splits)

Some notes:
-> rim-load only tracks usage, not efficacy, I'd say Oakley was very effective, Corzine not, Pippen Grant and Vincient were also effective, Sparrow and Turner not.
-> Jordan was very effective the one time the other team drove, but the first 2 times he's credited as the paint-protector were quick possessions where the other team didn't really try to drive.
-> Oakley had the most possessions where if I gave secondary credit he'd also be the #2, Grant and Pippen would come after

Suprisingly (not really), extending our sample did not lead to Jordan magically growing a couple inches taller or Jordan not still being a below average paint-protector who has a fraction of the paint-presence as multiple non-bigs on his own team. Yet, as far as BPM would be concerned, Jordan, as both a guard and lead(tied) block-getter would be the best paint-protector on his team.

So yeah, DBPM doesn't mean anything here. Jordan is the greatest block-padder of history, and the box-score loves him for it.

2. You are making this point in the context of trying to downplay Jordan’s BPM. Your theory isn’t really Jordan-specific, and is more just an objection to the idea of giving higher credit to a guard for a block than to a player in a bigger position

No, it is indeed Jordan-specific. Jordan was the focus of the commentary and the tracking.

But your argument would have to be geared towards an idea that Jordan’s blocks were idiosyncratically less valuable than the blocks from guards in general—an assertion that you have not made and not provided any evidence for.

Also no. The only condition necessary for my argument is that Jordan's blocks are less valuable than the blocks of the players he's compared to, or, more precisely, less reflective of his value. The burden is yours to demonstrate that the supposedly better accuracy regarding hundreds of players, is also more accurate regarding Jordan and Pippen. And before you do that, you'd need to show that metrics that give guards more credit for blocks outperform those which do the opposite.

The burden is not on me. The stat was constructed in order to correlate with RAPM. So we should generally assume that it is consistent with impact. You are arguing that in this particular instance with this particular player, the thing that normally correlates well with RAPM isn’t right. Obviously the burden is on you to demonstrate an exception to the general baseline. Your argument for that is to do analysis that would come to the same sort of result with regards to essentially every guard—because your analysis boils down to the fact that Jordan wasn’t the main one on his team contesting shots at the rim, but that’s going to be true of like every guard. If your argument is that there’s an exception to the baseline assumption because of something that would be the case essentially all the time, then we can probably assume that there’s a flaw in your analysis, because if you were right then giving guards higher credit for blocks wouldn’t correlate better with RAPM.

3. Even if we thought that Jordan’s blocks were overvalued because of his position, they’re really not a big component of his BPM. From 1987-1993, Jordan averaged 1.3 blocks per 100 possessions. Meanwhile, the difference in coefficient between a point guard’s blocks and a center’s blocks in the formula isn’t all that big (a 0.624 difference).

Again, no. Even if we assume the inflation per block is justified, that does not preclude Jordan or the players Jordan might be compared to being misrepresented by the amount of blocks recorded in the first place, hence why I highlighted the block distribution before the samples.

Huh? Are you asserting that Jordan’s number of blocks is wrong? Maybe I missed something in what you’ve previously said, but please direct me to any substantiation for the assertion that Jordan’s actual block stats are wrong.

Oh wow. The Bulls must have then really struggled when they had to go from a few minutes a game without Jordan to ...
not having Jordan at all...


(-3 relative to the league)

Oh, huh. Well, okay maybe it was just a fluky 72 games...


(-3 relative to the league)

Oh. Crazy what happens when you look at full games instead of spot minutes.

Of course, that ignores that, when Jordan retired, the Bulls recharged with superior centers and a really good, long wing defender in Kukoc as well as eventually Ron Harper (another great defender).

Kukoc was not a good defender. You also seem to be forgetting Jordan being replaced by negatives Kerr and Pete Myers, and the Bulls losing Horace Grant in 1995. And of course, even if I took your selective contextualization at face value, that would not explain Chicago becoming far worse in "off" that same season after Jordan returned.

It also ignores that the Bulls proceeded to put up rDRTGs when Jordan came back that were significantly better than they’d been putting up without him

Nope. The Bulls defense rating got worse by 0.3 points when Jordan returned in 1995. It got significantly better with the addition of Rodman and the Bulls would get significantly worse over the next 2 years when Rodman missed games. Alas, a partially sampled approximation based on a few missed minutes a game says he basically had no impact, must be true.

Kukoc was a good defender. This is substantiated by his RAPM being really good (both in JE’s set and Squared’s data) in those years (and good for his career overall as well). It’s also substantiated by the fact that Kukoc was an extremely long defender that could guard essentially any position. You’re just wrong there, though I grant that there were some misconceptions on this at the time (with, I think, some stereotypes of European players feeding into it, as well as people not realizing how valuable his sheer length was). As for Pete Myers, he was not a good player overall, but he was essentially average on defense. Same with Kerr, who did not have great physical tools, but was a smart defensive player who made good decisions. So for these purposes that amounted to Jordan being replaced with average defensive players (obviously that’s worse than Jordan, but then the rest of the team generally improved defensively, hence why the overall defense simply stayed similar—and that’s perhaps generous since they did substantially worse defensively in the playoffs than the Bulls ever did in that era with Jordan). As for losing Grant, that is a blow defensively, but it’s one that doesn’t apply to more than half the data you are talking about, and they also got Ron Harper that year and had Longley for much more of the year.

As for being worse in that same season after Jordan returned, you’re talking about 17 games, where they were barely worse overall. And, crucially, that is entirely a product of the Bulls defense being a lot worse in Jordan’s first five games back—when he was playing himself back into basketball shape and learning to play with new teammates in the middle of a season. After those first five games (culminating with the double nickel game), the Bulls DRTG was 101.43 the rest of the way—which was a good bit better than their DRTG had been prior to Jordan coming back (which was a bit over 104). And, of course, we know their rDRTG in the playoffs in 1995 was -2.7, compared to only -1.3 in the playoffs in 1994 (while having Horace Grant in those playoffs). So the Bulls were not only substantially better defensively in the full seasons after Jordan got back, but they were also better defensively in the playoffs with Jordan, and had a better DRTG in the 1995 regular season after the first few games of Jordan being back. So your argument there is really just completely based on what happened in Jordan’s first few games back from retirement. If you want to base your view of Jordan’s defense on how the Bulls did defensively in Jordan’s first few games back from retirement, then I guess you can do so, but it’s not going to be persuasive to essentially anyone.

Finally, the notion that the Bulls defense got significantly worse in games without Rodman is just false. In the 18 games Rodman missed in the 1995-1996 season, the Bulls had a 100.95 DRTG, which was actually better than the 101.8 DRTG they had for the season. Meanwhile, in the 27 games Rodman missed in the 1996-1997 season, the Bulls had a 103.08 DRTG, which was barely above the 102.4 DRTG they had for the season. And in the 2 games Rodman missed in the 1997-1998 season, the Bulls had a 94.65 DRTG, which, needless to say, was better than their season average. Neither RAPM nor WOWY analysis indicates that Rodman was a significantly impactful defender for the Bulls. The Bulls’ substantial defensive improvement was not driven by Rodman. It was driven by Jordan.

you just wave your hand at things and act like two very different teams across different years are essentially equivalent

You mean the thing you just tried to do with Rodman's return in 96? But you're right, they were not essentially equivalent; 88 MJ's cast was significantly (defensively) better. 95 Pippen's cast was significantly (defensively) worse. And Pippen led the better defense anyway. It's almost like one of them has a case as the best non-big defender ever, and the other is Micheal Jordan.

First of all, the 1996 Bulls were a lot more similar to the 1995 Bulls than the 1988 or 1989 Bulls were to the 1994 or 1995 Bulls. It’s not even a remotely valid comparison. The main differences between 1995 and 1996 were Rodman and Jordan, and we have a lot of data that tells us Rodman wasn’t driving the defensive improvement. Furthermore, the idea that the 1988 Bulls were a better team defensively than the 1994 or 1995 Bulls is just speculative nonsense that you want to believe because it helps you make a handwaving-nonsense argument that is the only way that you can get to your preferred conclusion, because essentially no actually valid data supports it (nor is it even consistent with general contemporaneous perception).

Of course, you’re also ignoring here that we don’t need to look across seasons to refute your claim, since the DRTGs with and without Pippen in a season where he missed half the season strongly contradict your claim.

Another big problem with your claim is it’s really not just “spot minutes” that supports Jordan’s case here. Pippen’s RAPM being lower in the 1997-1998 season is largely driven by what happened in the large number of games Pippen missed that season. And, indeed, if we look at the 38 games Pippen missed that season, the Bulls had an incredibly good 98.53 defensive rating. This is lower than the Bulls’ season DRTG. This is a far better defensive performance from the Bulls than they ever had without Jordan. In fact, it’s actually not at all a drop off from what they were doing *with* Pippen in that era or even that season.

So true. Luckily for Chicago, Pippen decided to become the best offensive player in the league:
https://media.discordapp.net/attachments/807803459331555363/1247382436904697948/image.png?ex=665fd2ec&is=665e816c&hm=14e336d592d72a0b5f2c378a35b6c5ccc1a834ff572eb0c6534ed65dde3dd0d3&=&format=webp&quality=lossless&width=605&height=199



Also best not handwave the RAPM "context" of Pippen being the league-leader in o-rapm in 1996.

This is why we should look to multiple pieces of data. Even though the Bulls offense did a lot better in the games Pippen played in in 1997-1998, RAPM does not tell us that Pippen was the best offensive player in the league in 1997-98. Nor does RPM (another impact data metric we have for those years). So the facetious claim you’re trying to make an analogy to is one that might look supported by raw WOWY, but isn’t supported by RAPM or RPM (though Pippen does unsurprisingly look good offensively!). In contrast, the data across the board indicates Jordan was more impactful defensively: WOWY analysis, RAPM, and RPM all support that point. Obviously the evidence for an assertion is stronger when it is supported by data across the board, rather than being supported by one piece of data and refuted by others. Your facetious assertion is also not remotely in line with contemporaneous perception of people who watched the Bulls play, but also nor is your claim that Pippen was far and away a better defender than Jordan. So your analogy is plainly silly.

I’d also note that it’s important to take a step back and think about the actual specific claims we are making here. I am presenting data that indicates Jordan was a superior defender to Pippen, but I’ve actually gone out of my way to explicitly say on multiple occasions that I don’t think this necessarily means Jordan is the superior defender, because there’s potential error/flaws in data like this. I am simply saying that this data is enough for us to reject your extreme assertion that Pippen was “far and away” the Bulls’ best defender. Your response here is to make some caricatured argument that Pippen must be the best offensive player in the league because his offensive WOWY looked good. But leaving aside the above point that that’s not supported by the data in the same way, I’m explicitly not saying we must take data entirely at face value. I’m simply saying we can use it to reject an assertion that is highly contrary to the data. So, for instance, if I made a claim that Pippen was merely an average offensive player, I think you could point to this sort of data for Pippen and say that it refutes my assertion. That’s the equivalent of what I’m doing here. You made an extreme claim that can be refuted by data without having to believe that data must be *exactly* right.

[Cavsfansince84]

Note on Harper: he wasn't really a defensive specialist or a guy with a great def rep early in his career. He was pretty good with the Cavs I think but then he went to the Clippers and took on a volume scoring role and I think had b2b season ending knee injuries then went to the Bulls in 95 and was ok. I don't think it was until 96 that he really accepted the role of part time ball handler and def specialist who could also run in transition. That's when he really figured out his spot but he wasn't just a guy playing great defense from day 1 there.

[DraymondGold]

2. You are making this point in the context of trying to downplay Jordan’s BPM. Your theory isn’t really Jordan-specific, and is more just an objection to the idea of giving higher credit to a guard for a block than to a player in a bigger position, because the player in the bigger position is typically going to be contesting more shots and therefore should get more credit for blocks if they’re used as a proxy for rim protection impact/deterrence. There’s some intuitive appeal to that assertion. But, as with the rest of the BPM formula, the reason BPM came up with different weightings for different positions is that it actually fit RAPM better (this is what the stat was geared towards doing!). So your intuition that blocks suggest higher impact for centers than for guards obviously didn’t actually show up in the data. It was apparently actually the opposite! So your argument is basically built on an intuition that you may think makes intuitive sense but that clearly wasn’t borne out by long-term RAPM data at all. Of course, that doesn’t necessarily mean that this overall correlation is true for a specific player. But your argument would have to be geared towards an idea that Jordan’s blocks were idiosyncratically less valuable than the blocks from guards in general—an assertion that you have not made and not provided any evidence for. It’s theoretically possible, I suppose. I’ve made an argument in the past that BPM downplaying offensive rebounds for centers (which it does to a way bigger degree than it downplays blocks for centers, by the way) may fit RAPM better in general, but that it idiosyncratically downplays the effect of Moses Malone’s offensive rebounding, since his rebounding seems to have had an idiosyncratically large effect on his teams’ rebounding (and therefore team ORTG) compared to other bigs. But that’s the sort of argument you’d need to make.

3. Even if we thought that Jordan’s blocks were overvalued because of his position, they’re really not a big component of his BPM. From 1987-1993, Jordan averaged 1.3 blocks per 100 possessions. Meanwhile, the difference in coefficient between a point guard’s blocks and a center’s blocks in the formula isn’t all that big (a 0.624 difference). And Jordan wasn’t a PG, so the coefficient on his blocks is surely smaller than the PG coefficient. We don’t know what exactly the coefficient for Jordan was, but overall we can be pretty confident that his BPM in that 1987-1993 period would probably only be somewhere around 0.6 lower even if we gave his blocks the *lowest possible* coefficient (which, I’ll remind you, is not something that is justified by its fit with RAPM). So you’re really just making a mountain out of a statistical molehill here. And that’s especially true when we realize that other wings like Pippen (or LeBron) have their blocks getting weighted much more highly than centers too (despite not being center-level rim deterrents)—so if you decided you want to deflate Jordan for this, then you’d need to deflate those guys at least somewhat too. It’s really an objectively extremely marginal thing, which you’re just using to try to discredit a stat that doesn’t say what you want it to say.

(0) Uncertainty of WOWY
Hey jake! Generally agree with your point here. Just want to add for those who are new to multi-year (e.g. team change / retirement) WOWY:
-the uncertainty in this kind of WOWY is *larger* than the signal. Among two-year WOWY samples for the standard top 15 players, the mean variation in overlapping samples is ~ 6.53 SRS (144% of the average signal).

By going to defense only, we're cutting the signal in half... which makes the uncertainty (relative to the signal) even larger.

Even if we take the statmuse numbers and associated logic at face value... the uncertainty is wayyyy bigger than the signal, sufficient to make it pretty unequivocal that the WOWY data does not support saying Pippen is far and away better than defense.

.. but I'm a little uninterested in re-hashing this stuff. You've made pretty clear points!

Two things you brought up that seem more interesting to me.

(1) Value of blocks on offense vs defense.
You rightfully point out blocks correlate better with value for guards than bigs, and it seems the data supports this.

One of the reasons blocks may be more valuable for guards than centers is that they have a much stronger offensive component.
For a position 1.0 Point Guard: BPM weighs blocks at 1.327, offensive BPM weighs blocks at 0.725, and the difference is 0.602
For a position 5.0 Center: BPM weighs blocks at 0.703, offensive BPM weighs blocks at 0.097, and the difference is 0.606

So the difference in the value of blocks mostly comes from offense! Which might make some amount of sense. Blocks at the guard position often start up the fast break. E.g. Blocks from the perimeter often send the ball's often sent down the court ahead of most of the defense. The guy who blocked or their teammate are often in the best position to pick it up (e.g. no need to turn their hips, they're already facing that way to take off on the break).

A block from a big doesn't start the fast break with as much frequency. This connection between the offense and defense is one of the reasons why it can be harder to get accurate offensive/defensive splits than to measure the overall value. It can be hard to accurately split the overall value into components.

The rest of your points still apply though. Really no evidence to suggest the blocks from a 3 are more valuable defensively than blocks from a 2, and there's the rest of the data we have. E.g. all the plus minus data, which is actually quite substantial.

(2) Jordan's Defensive role
I actually wasn't aware that this much data supported Jordan being a better defender than Pippen to quite this extent. I think unlike how it was characterized, it's quite a large amount of plus minus data -- I wouldn't be shocked at all if the plus minus data had smaller uncertainty bars than the WOWY (though I would love to find some article or source that describes how plus minus uncertainty goes down with sample size to be sure).

This goes against a lot of people's intuition. Most people (even among those who aren't in the acrimonious anti-Jordan association ) usually think Pippen was at least slightly better defensively than Jordan, even if the advantage was only slight. But impact metrics describe impact in a given role / fit. So I wonder... how much is the data suggesting
-Jordan's better defensively relative to Pippen than most people once thought, vs
-Uncertainty/noise causes Jordan to seem better when Pippen was actually better, vs
-Jordan's role on the team defense was more conducive to him than Pippen's was.

You've talked a bit about the first 2, but I'd love to hear if you (or others) have thoughts on the final point. Was Jordan's role on defense particularly suited for him compared to Pippen?

(I'm not sure this undermines your point regardless... even with a favorable role, it's hard to imagine a difference in the defensive fit to explain this data away by enough to argue Pippen's far and away the better defender; but perhaps it's more interesting to discuss than rehashing the uncertainty, systematic bias, and general limitations of Jordan's WOWY for the umpteenth time )

[Djoker]

Before conversing with Lebron fans, I've never heard anyone say that Pippen is by far the best defensive player on the Bulls. I don't think the Bulls dynasty had one player be by the far best on defense anyway. And for whatever is worth, the best may well have been Grant/Rodman and not Jordan or Pippen. But it's not clear cut and we simply don't have metrics or any objective way to figure it out. It's reasonable to say they were all great defensive players and leave it at that.

[AEnigma]

Always amazed at what Jordan brain can portray as confusing. Gee, why is the 6’5” guard with a 6’9” wingspan not as good a defender as the 6’8” forward with a 7’4” wingspan who could legitimately defend 1 through 5 and was consistently cited as the team’s defensive leader; sounds like a Lebron fan conspiracy to me!

Hell, why stop there. The 1998 Jazz got 3 points worse on defence when John Stockton missed time, so maybe if the Bulls had traded Pippen for Stockton they could have been 7 points better defensively!

[lessthanjake]

(0) Uncertainty of WOWY
Hey jake! Generally agree with your point here. Just want to add for those who are new to multi-year (e.g. team change / retirement) WOWY:
-the uncertainty in this kind of WOWY is *larger* than the signal. Among two-year WOWY samples for the standard top 15 players, the mean variation in overlapping samples is ~ 6.53 SRS (144% of the average signal).

By going to defense only, we're cutting the signal in half... which makes the uncertainty (relative to the signal) even larger.

Even if we take the statmuse numbers and associated logic at face value... the uncertainty is wayyyy bigger than the signal, sufficient to make it pretty unequivocal that the WOWY data does not support saying Pippen is far and away better than defense.

.. but I'm a little uninterested in re-hashing this stuff. You've made pretty clear points!

Great stuff in this post, DraymondGold!

Yeah, I agree completely regarding WOWY having a high degree of uncertainty, which is part of why I don’t look at the data I presented and say it must mean Jordan was the better defender, but rather just am using it to refute a more extreme claim in the other direction. Definitely agree that there’s more interesting things to discuss here—including the stuff you discussed in the rest of your post!

Two things you brought up that seem more interesting to me.

(1) Value of blocks on offense vs defense.
You rightfully point out blocks correlate better with value for guards than bigs, and it seems the data supports this.

One of the reasons blocks may be more valuable for guards than centers is that they have a much stronger offensive component.
For a position 1.0 Point Guard: BPM weighs blocks at 1.327, offensive BPM weighs blocks at 0.725, and the difference is 0.602
For a position 5.0 Center: BPM weighs blocks at 0.703, offensive BPM weighs blocks at 0.097, and the difference is 0.606

So the difference in the value of blocks mostly comes from offense! Which might make some amount of sense. Blocks at the guard position often start up the fast break. E.g. Blocks from the perimeter often send the ball's often sent down the court ahead of most of the defense. The guy who blocked or their teammate are often in the best position to pick it up (e.g. no need to turn their hips, they're already facing that way to take off on the break).

A block from a big doesn't start the fast break with as much frequency. This connection between the offense and defense is one of the reasons why it can be harder to get accurate offensive/defensive splits than to measure the overall value. It can be hard to accurately split the overall value into components.

The rest of your points still apply though. Really no evidence to suggest the blocks from a 3 are more valuable defensively than blocks from a 2, and there's the rest of the data we have. E.g. all the plus minus data, which is actually quite substantial.

Really interesting point! I didn’t realize we had a breakdown of which component the value of blocks was being put in by position. This additional information and the explanation you provide makes a lot of sense to me.

(2) Jordan's Defensive role
I actually wasn't aware that this much data supported Jordan being a better defender than Pippen to quite this extent. I think unlike how it was characterized, it's quite a large amount of plus minus data -- I wouldn't be shocked at all if the plus minus data had smaller uncertainty bars than the WOWY (though I would love to find some article or source that describes how plus minus uncertainty goes down with sample size to be sure).

This goes against a lot of people's intuition. Most people (even among those who aren't in the acrimonious anti-Jordan association ) usually think Pippen was at least slightly better defensively than Jordan, even if the advantage was only slight. But impact metrics describe impact in a given role / fit. So I wonder... how much is the data suggesting
-Jordan's better defensively relative to Pippen than most people once thought, vs
-Uncertainty/noise causes Jordan to seem better when Pippen was actually better, vs
-Jordan's role on the team defense was more conducive to him than Pippen's was.

You've talked a bit about the first 2, but I'd love to hear if you (or others) have thoughts on the final point. Was Jordan's role on defense particularly suited for him compared to Pippen?

The potential explanations you provide are certainly possible explanations. I think there’s two other possible explanations I’d add to the mix:

1. Perhaps the Bulls were better able to replace what Pippen did defensively than they were with Jordan. Virtually all of Pippen’s missed games and much of the RAPM data is from 1995 onwards. This is a time period where they had Kukoc. Of course, Kukoc is another really long defender that the Bulls could slot in at the SF position and who could plausibly guard almost anyone on the floor. So perhaps part of why we see the Bulls doing so well defensively in games without Pippen is that Kukoc could fulfill Pippen’s defensive role quite competently. On the flip side, the Bulls couldn’t really replace the kind of disruption Jordan provided—to some degree Harper could and that probably helped them in 1995, as did the general fact that Kukoc was a good defender. So the Bulls had good defenders in the game in lieu of Jordan and they did well defensively. But perhaps the Bulls were able to more precisely approximate Pippen than Jordan, allowing them to reach defensive heights with Pippen off that they never could with Jordan off. One advantage of this explanation is that it actually is pretty consistent with the data we have from non-Kukoc years. At this point, Squared has seemingly actually logged 79 Bulls games from the 1993 season and 57 games from the 1991 season (sidenote: I hope Squared eventually puts back up his page on Jordan plus-minus, so we can see what games these are, since my Jordan on-off thread could definitely be updated a lot with these additional games, if I knew which games they were—it’s a real shame that acrimonious discussion here led to Squared taking that page down). In those 136 games, the Bulls had a 100.9 DRTG with Jordan on and a 100.9 DRTG with Pippen on. Obviously there’s a lot of overlap in those minutes, but it obviously suggests an essentially equal DRTG with each of them on and the other off. Granted, the RAPM numbers themselves suggest an edge for Jordan once we account for who else was on the court (since Jordan is ahead in a possession-weighted average of their DRAPMs in that season, though that’s a crude measurement). But the DRTGs being very similar in those years without Kukoc is perhaps suggestive that this explanation has some legs.

2. One other potential explanation is that something about how Jordan communicated on defense was more indispensable. That also seems at least theoretically plausible, though I don’t have any independent information that would particularly corroborate that and I do think Pippen was a good communicator on defense, so the theory would probably rest more granularly on whether other guys on the team were good at communicating the same sorts of things (i.e. whether there was overlap in what the players communicated well). Probably not something we could really ever figure out.

[DraymondGold]

(0) Uncertainty of WOWY
Hey jake! Generally agree with your point here. Just want to add for those who are new to multi-year (e.g. team change / retirement) WOWY:
-the uncertainty in this kind of WOWY is *larger* than the signal. Among two-year WOWY samples for the standard top 15 players, the mean variation in overlapping samples is ~ 6.53 SRS (144% of the average signal).

By going to defense only, we're cutting the signal in half... which makes the uncertainty (relative to the signal) even larger.

Even if we take the statmuse numbers and associated logic at face value... the uncertainty is wayyyy bigger than the signal, sufficient to make it pretty unequivocal that the WOWY data does not support saying Pippen is far and away better than defense.

.. but I'm a little uninterested in re-hashing this stuff. You've made pretty clear points!

Great stuff in this post, DraymondGold!

Yeah, I agree completely regarding WOWY having a high degree of uncertainty, which is part of why I don’t look at the data I presented and say it must mean Jordan was the better defender, but rather just am using it to refute a more extreme claim in the other direction. Definitely agree that there’s more interesting things to discuss here—including the stuff you discussed in the rest of your post!

Two things you brought up that seem more interesting to me.

(1) Value of blocks on offense vs defense.
You rightfully point out blocks correlate better with value for guards than bigs, and it seems the data supports this.

One of the reasons blocks may be more valuable for guards than centers is that they have a much stronger offensive component.
For a position 1.0 Point Guard: BPM weighs blocks at 1.327, offensive BPM weighs blocks at 0.725, and the difference is 0.602
For a position 5.0 Center: BPM weighs blocks at 0.703, offensive BPM weighs blocks at 0.097, and the difference is 0.606

So the difference in the value of blocks mostly comes from offense! Which might make some amount of sense. Blocks at the guard position often start up the fast break. E.g. Blocks from the perimeter often send the ball's often sent down the court ahead of most of the defense. The guy who blocked or their teammate are often in the best position to pick it up (e.g. no need to turn their hips, they're already facing that way to take off on the break).

A block from a big doesn't start the fast break with as much frequency. This connection between the offense and defense is one of the reasons why it can be harder to get accurate offensive/defensive splits than to measure the overall value. It can be hard to accurately split the overall value into components.

The rest of your points still apply though. Really no evidence to suggest the blocks from a 3 are more valuable defensively than blocks from a 2, and there's the rest of the data we have. E.g. all the plus minus data, which is actually quite substantial.

Really interesting point! I didn’t realize we had a breakdown of which component the value of blocks was being put in by position. This additional information and the explanation you provide makes a lot of sense to me.

Yep! For further reading, it's on the page explaining the BPM (2.0) formula here: https://www.basketball-reference.com/about/bpm2.html

The BPM was fit to 5-year RAPM (prior based on team quality and minutes played). It's actually an interesting looking version of RAPM -- I wish they ran the full data, and not just the 4 non-overlapping timespans.

(2) Jordan's Defensive role
I actually wasn't aware that this much data supported Jordan being a better defender than Pippen to quite this extent. I think unlike how it was characterized, it's quite a large amount of plus minus data -- I wouldn't be shocked at all if the plus minus data had smaller uncertainty bars than the WOWY (though I would love to find some article or source that describes how plus minus uncertainty goes down with sample size to be sure).

This goes against a lot of people's intuition. Most people (even among those who aren't in the acrimonious anti-Jordan association ) usually think Pippen was at least slightly better defensively than Jordan, even if the advantage was only slight. But impact metrics describe impact in a given role / fit. So I wonder... how much is the data suggesting
-Jordan's better defensively relative to Pippen than most people once thought, vs
-Uncertainty/noise causes Jordan to seem better when Pippen was actually better, vs
-Jordan's role on the team defense was more conducive to him than Pippen's was.

You've talked a bit about the first 2, but I'd love to hear if you (or others) have thoughts on the final point. Was Jordan's role on defense particularly suited for him compared to Pippen?

The potential explanations you provide are certainly possible explanations. I think there’s two other possible explanations I’d add to the mix:

1. Perhaps the Bulls were better able to replace what Pippen did defensively than they were with Jordan. Virtually all of Pippen’s missed games and much of the RAPM data is from 1995 onwards. This is a time period where they had Kukoc. Of course, Kukoc is another really long defender that the Bulls could slot in at the SF position and who could plausibly guard almost anyone on the floor. So perhaps part of why we see the Bulls doing so well defensively in games without Pippen is that Kukoc could fulfill Pippen’s defensive role quite competently. On the flip side, the Bulls couldn’t really replace the kind of disruption Jordan provided—to some degree Harper could and that probably helped them in 1995, as did the general fact that Kukoc was a good defender. So the Bulls had good defenders in the game in lieu of Jordan and they did well defensively. But perhaps the Bulls were able to more precisely approximate Pippen than Jordan, allowing them to reach defensive heights with Pippen off that they never could with Jordan off. One advantage of this explanation is that it actually is pretty consistent with the data we have from non-Kukoc years. At this point, Squared has seemingly actually logged 79 Bulls games from the 1993 season and 57 games from the 1991 season (sidenote: I hope Squared eventually puts back up his page on Jordan plus-minus, so we can see what games these are, since my Jordan on-off thread could definitely be updated a lot with these additional games, if I knew which games they were—it’s a real shame that acrimonious discussion here led to Squared taking that page down). In those 136 games, the Bulls had a 100.9 DRTG with Jordan on and a 100.9 DRTG with Pippen on. Obviously there’s a lot of overlap in those minutes, but it obviously suggests an essentially equal DRTG with each of them on and the other off. Granted, the RAPM numbers themselves suggest an edge for Jordan once we account for who else was on the court (since Jordan is ahead in a possession-weighted average of their DRAPMs in that season, though that’s a crude measurement). But the DRTGs being very similar in those years without Kukoc is perhaps suggestive that this explanation has some legs.

2. One other potential explanation is that something about how Jordan communicated on defense was more indispensable. That also seems at least theoretically plausible, though I don’t have any independent information that would particularly corroborate that and I do think Pippen was a good communicator on defense, so the theory would probably rest more granularly on whether other guys on the team were good at communicating the same sorts of things (i.e. whether there was overlap in what the players communicated well). Probably not something we could really ever figure out.

Good points! #1 (Replacement) seems quite probable, and points to a trait that I have a hard time evaluating -- positional scarcity.

Even if Pippen is a better defender than Jordan 'in a vacuum', he might be less of an outlier relative to small forwards than Jordan is relative to shooting guards, which might boost Jordan's defense from a team-building perspective when you add him to a lineup vs Pippen. To be clear, and like you say, the raw plus minus would only show impact relative to his various backups on his team (not relative to the overall shooting guard position at the time). But it still brings to mind the more general idea of positional scarcity.

I haven't figured out quite how to incorporate positional scarcity into my evaluation of players. If 2 players are both e.g. +2 defense over the average defender in the NBA, but one player's a point guard and one's a Center (so the point guard is more likely to be replacing a worse defender and thus boosting their team defense more)... do we (option 1) call them equal defenders (relative to an average player in the nba) or do we (option 2) give the point guard credit for boosting the team more (relative to an average player *at their position* in the nba)? I would imagine most impact metrics are measures of option 2, but are they ever measures of option 1, or is this not a clearly defined question for some metrics?

Re: #2 (communication), I would definitely be surprised if Jordan were the superior communicator to Pippen (my impression was the reverse), but like you say, it's hard to tell objectively. Two ways forward would be either a) look back at reports from the time to see if anyone, e.g. the coach, ever talked about who were the most active communicators for the team, or b) look at film to see if we can catch any time they're gesturing or moving their mouths defensively (or if we're lucky, listen for if the mic picks anything up). Both have clear limitations, but at least it's a way forward. I forget which, but in one of Thinking Basketball's videos, he actually pointed out the communication habits of certain players in his film analysis -- one of the more impressive spots in his frequently impressive film analysis.

If I were NBA commissioner for a day, I would mic up every one on the court and publish the game transcript with each game. I'm sure there are tons of gems we miss every game. You could even start to do some basic stats (e.g. frequency of calls). Maybe even some sort of machine learning speech to text analysis could allow us to ballpark quantify how often players trash talk, give hype calls, communicate defensively vs offensively, etc. Never going to happen obviously, but it's a fun what if!

[lessthanjake]

I haven't figured out quite how to incorporate positional scarcity into my evaluation of players. If 2 players are both e.g. +2 defense over the average defender in the NBA, but one player's a point guard and one's a Center (so the point guard is more likely to be replacing a worse defender and thus boosting their team defense more)... do we (option 1) call them equal defenders (relative to an average player in the nba) or do we (option 2) give the point guard credit for boosting the team more (relative to an average player *at their position* in the nba)? I would imagine most impact metrics are measures of option 2, but are they ever measures of option 1, or is this not a clearly defined question for some metrics?

I get the impression that most impact metrics are actually more like Option 1, since it certainly seems like on average centers typically have good defensive impact ratings, while the opposite is true for PGs. It might depend on the metric, but that’s what I’ve generally observed. And I think it would make sense, since something like RAPM doesn’t know what position someone is, so it’d have no way of adjusting compared to position.

In terms of our own general analysis, I’m pretty agnostic between whether we should analyze using Option 1 or Option 2. Both should theoretically lead to the same result for a player *overall*, with the difference just being how much of their impact we are attributing to offense and defense respectively. I tend to go more for Option 1, since I think most metrics are oriented towards that, so I think using that approach requires layering on less subjective adjustments to my analysis. But either approach seems fine to me in general. What I think is definitely important, though, is to not mix and match approaches in order to artificially increase or decrease our assessment of the player. For instance, we currently have a player in Jokic that is a great example of where that could come into play. He’s an incredible offensive player at the position that is generally the least good offensively, while not being particularly strong at the defensive side of the ball that players of his position normally excel in. Under Option 1, we’d say he’s an incredible offensive player and an above-average defender. Under Option 2, we’d say he’s an even more incredible offensive player and probably a slightly below-average defender (note to others: not interested in discussing this particular assessment—it is an example for illustrative purposes only). Both options would arrive at the same overall place for Jokic’s impact, just with different allocations of how much of his impact comes from each side of the ball. There’s also what I’ll label Option 3 and Option 4—which are invalid because they just mix and match these two, to get to an invalidly positive or negative conclusion. Option 3 would be to assess Jokic’s offense relative to the all positions and his defense relative to centers. This would downplay him as much as possible. Option 4 would be to assess Jokic’s offense relative to centers and his defense relative to all positions. This would play him up as much as possible. I think the most important thing here is to try to avoid those sorts of approaches, though I think unfortunately they’re quite common (and I’m sure I’ve fallen into this trap myself before!).

[DraymondGold]

I haven't figured out quite how to incorporate positional scarcity into my evaluation of players. If 2 players are both e.g. +2 defense over the average defender in the NBA, but one player's a point guard and one's a Center (so the point guard is more likely to be replacing a worse defender and thus boosting their team defense more)... do we (option 1) call them equal defenders (relative to an average player in the nba) or do we (option 2) give the point guard credit for boosting the team more (relative to an average player *at their position* in the nba)? I would imagine most impact metrics are measures of option 2, but are they ever measures of option 1, or is this not a clearly defined question for some metrics?

I get the impression that most impact metrics are actually more like Option 1, since it certainly seems like on average centers typically have good defensive impact ratings, while the opposite is true for PGs. It might depend on the metric, but that’s what I’ve generally observed. And I think it would make sense, since something like RAPM doesn’t know what position someone is, so it’d have no way of adjusting compared to position.

In terms of our own general analysis, I’m pretty agnostic between whether we should analyze using Option 1 or Option 2. Both should theoretically lead to the same result for a player *overall*, with the difference just being how much of their impact we are attributing to offense and defense respectively. I tend to go more for Option 1, since I think most metrics are oriented towards that, so I think using that approach requires layering on less subjective adjustments to my analysis. But either approach seems fine to me in general. What I think is definitely important, though, is to not mix and match approaches in order to artificially increase or decrease our assessment of the player. For instance, we currently have a player in Jokic that is a great example of where that could come into play. He’s an incredible offensive player at the position that is generally the least good offensively, while not being particularly strong at the defensive side of the ball that players of his position normally excel in. Under Option 1, we’d say he’s an incredible offensive player and an above-average defender. Under Option 2, we’d say he’s an even more incredible offensive player and probably a slightly below-average defender (note to others: not interested in discussing this particular assessment—it is an example for illustrative purposes only). Both options would arrive at the same overall place for Jokic’s impact, just with different allocations of how much of his impact comes from each side of the ball. There’s also what I’ll label Option 3 and Option 4—which are invalid because they just mix and match these two, to get to an invalidly positive or negative conclusion. Option 3 would be to assess Jokic’s offense relative to the all positions and his defense relative to centers. This would downplay him as much as possible. Option 4 would be to assess Jokic’s offense relative to centers and his defense relative to all positions. This would play him up as much as possible. I think the most important thing here is to try to avoid those sorts of approaches, though I think unfortunately they’re quite common (and I’m sure I’ve fallen into this trap myself before!).

Fair push back re: what stats do! I think that's closer to right.

Like you say, something that adjusts for teammates and opponents across the league (at least something 'pure' like RAPM) doesn't know what position someone plays, and so is position-agnostic / doesn't consider positional scarcity (what I called option 1).

One slight qualm with your current position (at least as stated): I'm not sure un-adjusted raw stats do quite the same. Something like raw on/off or raw WOWY doesn't consider positional scarcity relative to average person at that position (so it's not option 2), but I'm also not quiet comfortable saying it's option 1 either. It looks value relative to your specific replacement (+ no corrections for other teammate/opponent changes). So that might be... option 5?... which Is a bit of a confusing numbering scheme, but options 3 and 4 are already claimed

To generalize, the options might be:
-Value relative to your specific replacement (option 5; stats include raw WOWY, raw on/off)
-Value relative to average player in nba (option 1; stats include RAPM)
-Value relative to average player at your position (option 2, includes positional scarcity).

Or various hybrids of these options, choosing different ones for offense/defense, etc. And agreed It's best to try to avoid Option 3/4. They're a bit inconsistent philosophically, and lead to evaluation that's biased for/against certain players.

I wonder... where does BPM, which does have some measure for position, fall in this spectrum? Option 2, including positional scarcity? It does change the value of box stats for different positions after all. Or is it still using option 1(doesn't include positional scarcity in evaluation), while including positional data just to try to improve the fit for option 1?

[lessthanjake]

I haven't figured out quite how to incorporate positional scarcity into my evaluation of players. If 2 players are both e.g. +2 defense over the average defender in the NBA, but one player's a point guard and one's a Center (so the point guard is more likely to be replacing a worse defender and thus boosting their team defense more)... do we (option 1) call them equal defenders (relative to an average player in the nba) or do we (option 2) give the point guard credit for boosting the team more (relative to an average player *at their position* in the nba)? I would imagine most impact metrics are measures of option 2, but are they ever measures of option 1, or is this not a clearly defined question for some metrics?

I get the impression that most impact metrics are actually more like Option 1, since it certainly seems like on average centers typically have good defensive impact ratings, while the opposite is true for PGs. It might depend on the metric, but that’s what I’ve generally observed. And I think it would make sense, since something like RAPM doesn’t know what position someone is, so it’d have no way of adjusting compared to position.

In terms of our own general analysis, I’m pretty agnostic between whether we should analyze using Option 1 or Option 2. Both should theoretically lead to the same result for a player *overall*, with the difference just being how much of their impact we are attributing to offense and defense respectively. I tend to go more for Option 1, since I think most metrics are oriented towards that, so I think using that approach requires layering on less subjective adjustments to my analysis. But either approach seems fine to me in general. What I think is definitely important, though, is to not mix and match approaches in order to artificially increase or decrease our assessment of the player. For instance, we currently have a player in Jokic that is a great example of where that could come into play. He’s an incredible offensive player at the position that is generally the least good offensively, while not being particularly strong at the defensive side of the ball that players of his position normally excel in. Under Option 1, we’d say he’s an incredible offensive player and an above-average defender. Under Option 2, we’d say he’s an even more incredible offensive player and probably a slightly below-average defender (note to others: not interested in discussing this particular assessment—it is an example for illustrative purposes only). Both options would arrive at the same overall place for Jokic’s impact, just with different allocations of how much of his impact comes from each side of the ball. There’s also what I’ll label Option 3 and Option 4—which are invalid because they just mix and match these two, to get to an invalidly positive or negative conclusion. Option 3 would be to assess Jokic’s offense relative to the all positions and his defense relative to centers. This would downplay him as much as possible. Option 4 would be to assess Jokic’s offense relative to centers and his defense relative to all positions. This would play him up as much as possible. I think the most important thing here is to try to avoid those sorts of approaches, though I think unfortunately they’re quite common (and I’m sure I’ve fallen into this trap myself before!).

Fair push back re: what stats do! I think that's closer to right.

Like you say, something that adjusts for teammates and opponents across the league (at least something 'pure' like RAPM) doesn't know what position someone plays, and so is position-agnostic / doesn't consider positional scarcity (what I called option 1).

One slight qualm with your current position (at least as stated): I'm not sure un-adjusted raw stats do quite the same. Something like raw on/off or raw WOWY doesn't consider positional scarcity relative to average person at that position (so it's not option 2), but I'm also not quiet comfortable saying it's option 1 either. It looks value relative to your specific replacement (+ no corrections for other teammate/opponent changes). So that might be... option 5?... which Is a bit of a confusing numbering scheme, but options 3 and 4 are already claimed

To generalize, the options might be:
-Value relative to your specific replacement (option 5; stats include raw WOWY, raw on/off)
-Value relative to average player in nba (option 1; stats include RAPM)
-Value relative to average player at your position (option 2, includes positional scarcity).

Or various hybrids of these options, choosing different ones for offense/defense, etc. And agreed It's best to try to avoid Option 3/4. They're a bit inconsistent philosophically, and lead to evaluation that's biased for/against certain players.

I wonder... where does BPM, which does have some measure for position, fall in this spectrum? Option 2, including positional scarcity? It does change the value of box stats for different positions after all. Or is it still using option 1(doesn't include positional scarcity in evaluation), while including positional data just to try to improve the fit for option 1?

Yeah, agreed that raw on-off and WOWY are definitely different since they’re looking at value relative to replacement. I was more talking about RAPM and metrics that incorporate RAPM when I said that I think metrics are typically analyzing relative to the average player in the NBA. For those, the teammate adjustment at least theoretically makes it be valuing relative to average, because it is adjusting for who the replacement is, while raw on-off and WOWY don’t do that.

As for what BPM is doing in these terms, I guess I’d say since it was created to correlate with RAPM then we can probably say it is trying to approximate the same sort of approach RAPM is taking.

All interesting things to think about!

[OhayoKD]

2. You are making this point in the context of trying to downplay Jordan’s BPM. Your theory isn’t really Jordan-specific, and is more just an objection to the idea of giving higher credit to a guard for a block than to a player in a bigger position

No, it is indeed Jordan-specific. Jordan was the focus of the commentary and the tracking.

But your argument would have to be geared towards an idea that Jordan’s blocks were idiosyncratically less valuable than the blocks from guards in general—an assertion that you have not made and not provided any evidence for.

Also no. The only condition necessary for my argument is that Jordan's blocks are less valuable than the blocks of the players he's compared to, or, more precisely, less reflective of his value. The burden is yours to demonstrate that the supposedly better accuracy regarding hundreds of players, is also more accurate regarding Jordan and Pippen. And before you do that, you'd need to show that metrics that give guards more credit for blocks outperform those which do the opposite.

The burden is not on me. The stat was constructed in order to correlate with RAPM. So we should generally assume that it is consistent with impact.

"The stat was intended to be accurate, and is therefore accurate!"

You know, if you're going to use an appeal to authority to avoid actually supporting your claims, you might want to check what said authority has actually said on the matter

Going to clear up a few questions about BPM/VORP here:
I would say BPM is fairly balanced guards vs. big men. There are different weights for many of the box score statistics based on the estimated position of the player and the estimated offensive role of the player.

BPM IS biased towards players that produce good box scores. The argument should be primarily about whether the player's box score statistics accurately reflect the player's actual impact.

As a rule of thumb, I would say the box score captures 80% of a player's offensive impact and 50% of a player's defensive impact. That is based on numerous regressions I have run onto RAPM-style metrics.

I personally would argue that John Stockton had a profile that would be overrated by the box score. His steals and his assists both tended to overstate his impact on the game in those areas, I feel. And his BPM is reflective of that. In other words--some of the credit that his assists and steals are collecting for his BPM/VORP should actually have gone to his teammates.

Just think of BPM as a reasonable approximation of what the box score + the overall team quality says about a player. Remember that is blind to around 20% of offense and 50% of defense.

As it so happens, BPM's creator, after seeing this same point made with this same tracking(as well as seeing what a much larger sample of play-by-play tracking had to say about Embid)...
https://forums.realgm.com/boards/viewtopic.php?p=108568441#p108568441

Proceeded to shift the the way the input of "size" was calculated to better address the highlighted discrepancy in a 3rd version of the metric

("Size" is one of two factors(the other being "creation") conceived as a replacement for "position" estimations in the old BPM)

IDK how exactly it would math out, but there might be a way to better approximate paint-protectoin(and to a large extent defensive value) by tying size and blocks to rebound totals. Both can be gamed by smaller players, but i'm guessing it's harder to game both.

Not a big sample, but even something crude like just adding rebounds and blocks make Draymond and Embid standout from their teammates closer to how they stand out if you were to look at synergy deterrence/rim protecting numbers

Having looked at it further, this last sentence makes a lot of sense. I realized there was a lack of robustness in my original approach--averaging % of team's TRB and % of team's blocks doesn't work in some scenarios. Consider a team where they are mostly guards and get almost no blocks--if one player got a couple of blocks and ends up as 100% of the team's blocks--that is not at all the same "size role" as a twin-tower team where the biggest big gets 40% of the blocks. So I'm going to be pivoting toward giving TRB and BLK each a weight and then summing them and calculating the % of that summed value each player has.

In light of the discussion here and at APBR, I have revised my methodology significantly.

For the Size dimension, the new methodology is to take % of team's (TRB + 3*BLK). In other words, blocks are worth 3x rebounds. This was found by regressing on actual player size (height and wingspan), based on a measurement dataset. Using an additive approach makes this setup far more robust. Hassan Whiteside's 2015 season is still an outlier, but that can't be helped...

If exalting Jordan means you will ignore what the metric’s creator says about the metric, while using their intentions as proof of concept, we may as well be done here.

[trex_8063]

Kukoc was not a good defender. You also seem to be forgetting Jordan being replaced by negatives Kerr and Pete Myers, and the Bulls losing Horace Grant in 1995. And of course, even if I took your selective contextualization at face value, that would not explain Chicago becoming far worse in "off" that same season after Jordan returned.

Kukoc was a good defender. This is substantiated by his RAPM being really good (both in JE’s set and Squared’s data) in those years (and good for his career overall as well). It’s also substantiated by the fact that Kukoc was an extremely long defender that could guard essentially any position. You’re just wrong there, though I grant that there were some misconceptions on this at the time (with, I think, some stereotypes of European players feeding into it, as well as people not realizing how valuable his sheer length was). As for Pete Myers, he was not a good player overall, but he was essentially average on defense. Same with Kerr, who did not have great physical tools, but was a smart defensive player who made good decisions. So for these purposes that amounted to Jordan being replaced with average defensive players (obviously that’s worse than Jordan, but then the rest of the team generally improved defensively, hence why the overall defense simply stayed similar—and that’s perhaps generous since they did substantially worse defensively in the playoffs than the Bulls ever did in that era with Jordan). As for losing Grant, that is a blow defensively, but it’s one that doesn’t apply to more than half the data you are talking about, and they also got Ron Harper that year and had Longley for much more of the year.

Going to chip in my 2c.....

Disagree with the implication that Steve Kerr was average on defense. While I generally agree that he was "smart"(ish) defensively and he gave decent effort, his physical limitations were simply too much. He was not tall or long, nor an explosive leaper (hindrance in shot contesting); he was not big/strong (dies easily on screens and can be abused in the post or on mismatches); his lateral quickness was merely average(ish) [at best]; and he didn't have especially quick hands or sniping instincts (to create deflections or generate turnovers).
These limitations added up.
He wasn't a gushing wound on defense, but he wasn't average either. I think generously we could call him a "marginal negative" defensively, but that's about as nice as we can get.***

***However, Steve Kerr was a very nice offensive supporting piece: spread the floor, super-efficient on small scoring volume, with GOAT-tier turnover economy (was very smart about playing within his limits: VERY low rate of error on that half of the court). And he (along with Kukoc, among other changes) is a big part of why I sort of role my eyes when people make statements that imply the '94 Bulls were basically the '93 Bulls minus Jordan.


Agree regarding Pete Myers. He was a poor offensive player (not quite what I'd call an empty uniform, but not terribly much above it). But he was totally passable defensively; seriously, it's the only reason he was getting 24.8 mpg on that team (instead of giving ~30 mpg to Kerr). Don't know that I'd go as far as calling him "good"; but calling him "bad" or "weak" or "poor" is not an accurate distinction, imo.


And I generally agree (at least sort of) regarding Kukoc. I think he was better defensively than he's typically given credit for, though his value was scattered. Against strong, bruising/bullish PF's who wanted to bang in the post.....Kukoc was a swinging door. He just did NOT defend well against them (at least not in his Chicago years; not sure about later career).
But he was surprisingly quick (for his size) and moved his feet well on the perimeter. He had length (which made simply getting a hand up more effective than it would for shorter players), and wasn't too bad about generating turnovers (WITHOUT excessive gambling). So to some degree, he could defend quite a few positions (except the quickest PG's, or the strongest PF's).
And he was a respectable defensive rebounding SF (a little less than guys like Pippen or Jerome Kersey, but notably better than SF's like Dennis Scott, Reggie Williams, Walt Williams, Sean Elliott, or Stacey Augmon).
And indeed, his DRAPM's were semi-consistently small positives.

[lessthanjake]

No, it is indeed Jordan-specific. Jordan was the focus of the commentary and the tracking.


Also no. The only condition necessary for my argument is that Jordan's blocks are less valuable than the blocks of the players he's compared to, or, more precisely, less reflective of his value. The burden is yours to demonstrate that the supposedly better accuracy regarding hundreds of players, is also more accurate regarding Jordan and Pippen. And before you do that, you'd need to show that metrics that give guards more credit for blocks outperform those which do the opposite.

The burden is not on me. The stat was constructed in order to correlate with RAPM. So we should generally assume that it is consistent with impact.

"The stat was intended to be accurate, and is therefore accurate!"

You know, if you're going to use an appeal to authority to avoid actually supporting your claims, you might want to check what said authority has actually said on the matter

Going to clear up a few questions about BPM/VORP here:
I would say BPM is fairly balanced guards vs. big men. There are different weights for many of the box score statistics based on the estimated position of the player and the estimated offensive role of the player.

BPM IS biased towards players that produce good box scores. The argument should be primarily about whether the player's box score statistics accurately reflect the player's actual impact.

As a rule of thumb, I would say the box score captures 80% of a player's offensive impact and 50% of a player's defensive impact. That is based on numerous regressions I have run onto RAPM-style metrics.

I personally would argue that John Stockton had a profile that would be overrated by the box score. His steals and his assists both tended to overstate his impact on the game in those areas, I feel. And his BPM is reflective of that. In other words--some of the credit that his assists and steals are collecting for his BPM/VORP should actually have gone to his teammates.

Just think of BPM as a reasonable approximation of what the box score + the overall team quality says about a player. Remember that is blind to around 20% of offense and 50% of defense.

As it so happens, BPM's creator, after seeing this same point made with this same tracking(as well as seeing what a much larger sample of play-by-play tracking had to say about Embid)...
https://forums.realgm.com/boards/viewtopic.php?p=108568441#p108568441

Proceeded to shift the the way the input of "size" was calculated to better address the highlighted discrepancy in a 3rd version of the metric

("Size" is one of two factors(the other being "creation") conceived as a replacement for "position" estimations in the old BPM)

IDK how exactly it would math out, but there might be a way to better approximate paint-protectoin(and to a large extent defensive value) by tying size and blocks to rebound totals. Both can be gamed by smaller players, but i'm guessing it's harder to game both.

Not a big sample, but even something crude like just adding rebounds and blocks make Draymond and Embid standout from their teammates closer to how they stand out if you were to look at synergy deterrence/rim protecting numbers

Having looked at it further, this last sentence makes a lot of sense. I realized there was a lack of robustness in my original approach--averaging % of team's TRB and % of team's blocks doesn't work in some scenarios. Consider a team where they are mostly guards and get almost no blocks--if one player got a couple of blocks and ends up as 100% of the team's blocks--that is not at all the same "size role" as a twin-tower team where the biggest big gets 40% of the blocks. So I'm going to be pivoting toward giving TRB and BLK each a weight and then summing them and calculating the % of that summed value each player has.

In light of the discussion here and at APBR, I have revised my methodology significantly.

For the Size dimension, the new methodology is to take % of team's (TRB + 3*BLK). In other words, blocks are worth 3x rebounds. This was found by regressing on actual player size (height and wingspan), based on a measurement dataset. Using an additive approach makes this setup far more robust. Hassan Whiteside's 2015 season is still an outlier, but that can't be helped...

If exalting Jordan means you will ignore what the metric’s creator says about the metric, while using their intentions as proof of concept, we may as well be done here.

I think you’re really getting a bit mixed up about what we’re even talking about. I’m certainly not saying that BPM is a good measure of overall defense. I think it’s pretty obvious that box-score stats are not very good measures of overall defense, since there’s so little of what happens on defense that ends up in the box score. We aren’t arguing whether BPM is a great measure of defense in general. Indeed, no one made that assertion at all. Oddly enough, it’s not even clear why you started talking about BPM in the first place. As far as I can tell, it seems to be that someone mentioned the term “metrics” in a discussion about whether Pippen deserved his MVP voting placement in 1994 and you used that as a launching pad to talk about BPM. Amusingly, the “metrics” that that person was referring to actually clearly was *not* basketball-reference BPM, since they said “In '94, Shaq and Stockton both finished with better metrics” than Pippen and yet Shaq actually had lower BPM than Pippen that season. So you really just were going on a tangent that didn’t relate to anything anyone had argued. No one actually based an argument about Pippen’s and Jordan’s defensive prowess on BPM. In fact, the discussion you used as a launching pad actually wasn’t even about Jordan at all—it was comparing Pippen to other players in 1994—and, as mentioned above, it wasn’t actually about BPM either. You were truly beating the most straw-filled of straw men.

So yeah, you were straw manning about this subject. Alas, I responded to that (I’ll note that I didn’t realize at the time how much of a tangent you were on). But your own straw manning wasn’t actually to assert that BPM isn’t a great measure of defense in general. If it was, then I wouldn’t have disagreed! It was specifically to take issue with how BPM “distributes credit for blocks” and to assert that the stat specifically attributes too much value to Jordan for his blocks. That’s a much more specific claim, since the composition of DBPM (and OBPM) is affected by everything in the box score, not just blocks, and also a measure not generally being very accurate in measuring defense doesn’t mean it inflates (or deflates) any particular player. We can think that BPM distributes credit for blocks in a way that correlates reasonably well with impact and likely doesn’t meaningfully overcredit Jordan for his blocks, while not actually thinking that BPM is particularly accurate at measuring defense more generally. Those aren’t mutually exclusive at all. Indeed, the main problem with box-score-based defense metrics is that there’s not a lot of defensive stuff directly in the box score (which is why the components of DBPM end up having to include lots of offensive stats, which, at best are quite indirectly related to defense). But blocks actually are a defensive action that is in the box score, so “distributing credit for blocks” specifically is something that isn’t subject to the metric’s biggest issue regarding defense overall. In the case of blocks, the regression is actually attributing defensive value to a direct defensive action, rather than attributing defensive value to offensive actions under the theory that those offensive actions correlate with defense (i.e. for instance, weighting assists positively in DBPM because assists correlate with court awareness more generally). Seems obvious we should have more confidence in the former than the latter, which also means we’d have more confidence in the value attributed to blocks than in the defensive component of BPM more generally. In other words, you made a specific assertion that really cannot be justified by simply pointing to the fact that BPM isn’t particularly accurate in measuring defense overall. The defensive component of BPM should be understood to be a very rough guide overall, but that doesn’t mean that the credit for blocks specifically is wildly off, and it certainly doesn’t mean that the credit for Jordan’s blocks specifically are inflated. So you’re just pointing to things that are at best tangentially relevant to what we’ve been discussing.

EDIT: By the way, to the extent you’re making an assertion that is less about how the value of blocks is accounted for by BPM and is instead about how accruing blocks affects what position the formula assigns to someone and therefore what their “position adjustment” ends up being (which is what the last half your post seems to be aimed at—though most of that relates to discussions about an as-yet-unreleased version of BPM under development and therefore isn’t really relevant), I should note that I just quickly ran what Jordan’s BPM position would be in the year he had his most blocks per game, and it came out to 2.1, so this really isn’t a situation where Jordan’s blocks are making the formula unduly think he’s a big man that should avoid the negative position adjustment (which is basically designed to penalize guards for generally having weaker defense). The stat correctly identifies his position, even with his blocks being pretty high.

[lessthanjake]

Going to chip in my 2c.....

Disagree with the implication that Steve Kerr was average on defense. While I generally agree that he was "smart"(ish) defensively and he gave decent effort, his physical limitations were simply too much. He was not tall or long, nor an explosive leaper (hindrance in shot contesting); he was not big/strong (dies easily on screens and can be abused in the post or on mismatches); his lateral quickness was merely average(ish) [at best]; and he didn't have especially quick hands or sniping instincts (to create deflections or generate turnovers).
These limitations added up.
He wasn't a gushing wound on defense, but he wasn't average either. I think generously we could call him a "marginal negative" defensively, but that's about as nice as we can get.***

***However, Steve Kerr was a very nice offensive supporting piece: spread the floor, super-efficient on small scoring volume, with GOAT-tier turnover economy (was very smart about playing within his limits: VERY low rate of error on that half of the court). And he (along with Kukoc, among other changes) is a big part of why I sort of role my eyes when people make statements that imply the '94 Bulls were basically the '93 Bulls minus Jordan.


Agree regarding Pete Myers. He was a poor offensive player (not quite what I'd call an empty uniform, but terribly not much above it). But he was totally passable defensively; seriously, it's the only reason he was getting 24.8 mpg on that team (instead of giving ~30 mpg to Kerr). Don't know that I'd go as far as calling him "good"; but calling him "bad" or "weak" or "poor" is not an accurate distinction, imo.


And I generally agree (at least sort of) regarding Kukoc. I think he was better defensively than he's typically given credit for, though his value was scattered. Against strong, bruising/bullish PF's who wanted to bang in the post.....Kukoc was a swinging door. He just did NOT defend well against them (at least not in his Chicago years; not sure about later career).
But he was surprisingly quick (for his size) and moved his feet well on the perimeter. He had length (which made simply getting a hand up more effective than it would for shorter players), and wasn't too bad about generating turnovers (WITHOUT excessive gambling). So to some degree, he could defend quite a few positions (except the quickest PG's, or the strongest PF's).
And he was a respectable defensive rebounding SF (a little less than guys like Pippen or Jerome Kersey, but notably better than SF's like Dennis Scott, Reggie Williams, Walt Williams, Sean Elliott, or Stacey Augmon).
And indeed, his DRAPM's were semi-consistently small positives.

This is all generally a fair perspective IMO. A couple things I’d note:

- On Kerr, I would note that Kerr’s career DRAPM is actually essentially exactly average (it is +0.1). And that’s with RAPM only accounting for his age 31 to age 37 seasons, so we’d expect it would be underselling what he was defensively when he was younger. Of course, Kerr having an average RAPM doesn’t definitively mean that’s what he was. You’re certainly free to think he was worse than that. But, for me, that RAPM is in line with how I perceived Kerr at the time—I never felt like the defense would collapse with him in.

- I think we’re essentially on the same page regarding Kukoc. I’d just note that I think it undersells him to say his DRAPM’s were merely “semi-consistently small positives.” I suppose that’s accurate if we’re talking about his late-career years in his mid-late 30s. But his DRAPM in his earlier years we have were actually great! For instance, in the JE season-by-season PI RAPM, Kukoc was ranked 67th, 18th, and 27th in the league in DRAPM in 1996-1997, 1997-1998, and 1998-1999 respectively. And he was like 15th in the league in DRAPM in Squared’s 1995-1996 sampling. In terms of actual numbers, Kukoc’s DRAPMs listed above in chronological order for 1996-1999 were: 2.41, 1.29, 3.11, and 2.49. I’d say DRAPM from that era actually portrays Kukoc as a pretty elite defender! And the fact that that held through the 1998-1999 season suggests this isn’t just a collinearity thing.

[Redmoon]

Note on Harper: he wasn't really a defensive specialist or a guy with a great def rep early in his career. He was pretty good with the Cavs I think but then he went to the Clippers and took on a volume scoring role and I think had b2b season ending knee injuries then went to the Bulls in 95 and was ok. I don't think it was until 96 that he really accepted the role of part time ball handler and def specialist who could also run in transition. That's when he really figured out his spot but he wasn't just a guy playing great defense from day 1 there.

Rewatched some of the bulls 96/97 PS runs recently and Harpers d stood out to me. Long arms (longer than MJ?), really great at rear contests and picking pockets. Am curious about his earlier cav days, who do you think was a better defender, Ehlo or Harp?
Of course Pippen was a game changer and had a ton of high leverage plays on that end as well. His athleticism before injuries took over was eye popping, making those bron esque chase down blocks. I think your comment on Pip devoting a larger part of his energy to the defensive end makes alot of sense now thinking back.

[Owly]

Going to chip in my 2c.....

Disagree with the implication that Steve Kerr was average on defense. While I generally agree that he was "smart"(ish) defensively and he gave decent effort, his physical limitations were simply too much. He was not tall or long, nor an explosive leaper (hindrance in shot contesting); he was not big/strong (dies easily on screens and can be abused in the post or on mismatches); his lateral quickness was merely average(ish) [at best]; and he didn't have especially quick hands or sniping instincts (to create deflections or generate turnovers).
These limitations added up.
He wasn't a gushing wound on defense, but he wasn't average either. I think generously we could call him a "marginal negative" defensively, but that's about as nice as we can get.***

***However, Steve Kerr was a very nice offensive supporting piece: spread the floor, super-efficient on small scoring volume, with GOAT-tier turnover economy (was very smart about playing within his limits: VERY low rate of error on that half of the court). And he (along with Kukoc, among other changes) is a big part of why I sort of role my eyes when people make statements that imply the '94 Bulls were basically the '93 Bulls minus Jordan.


Agree regarding Pete Myers. He was a poor offensive player (not quite what I'd call an empty uniform, but terribly not much above it). But he was totally passable defensively; seriously, it's the only reason he was getting 24.8 mpg on that team (instead of giving ~30 mpg to Kerr). Don't know that I'd go as far as calling him "good"; but calling him "bad" or "weak" or "poor" is not an accurate distinction, imo.


And I generally agree (at least sort of) regarding Kukoc. I think he was better defensively than he's typically given credit for, though his value was scattered. Against strong, bruising/bullish PF's who wanted to bang in the post.....Kukoc was a swinging door. He just did NOT defend well against them (at least not in his Chicago years; not sure about later career).
But he was surprisingly quick (for his size) and moved his feet well on the perimeter. He had length (which made simply getting a hand up more effective than it would for shorter players), and wasn't too bad about generating turnovers (WITHOUT excessive gambling). So to some degree, he could defend quite a few positions (except the quickest PG's, or the strongest PF's).
And he was a respectable defensive rebounding SF (a little less than guys like Pippen or Jerome Kersey, but notably better than SF's like Dennis Scott, Reggie Williams, Walt Williams, Sean Elliott, or Stacey Augmon).
And indeed, his DRAPM's were semi-consistently small positives.

This is all generally a fair perspective IMO. A couple things I’d note:

- On Kerr, I would note that Kerr’s career DRAPM is actually essentially exactly average (it is +0.1). And that’s with RAPM only accounting for his age 31 to age 37 seasons, so we’d expect it would be underselling what he was defensively when he was younger. Of course, Kerr having an average RAPM doesn’t definitively mean that’s what he was. You’re certainly free to think he was worse than that. But, for me, that RAPM is in line with how I perceived Kerr at the time—I never felt like the defense would collapse with him in.

- I think we’re essentially on the same page regarding Kukoc. I’d just note that I think it undersells him to say his DRAPM’s were merely “semi-consistently small positives.” I suppose that’s accurate if we’re talking about his late-career years in his mid-late 30s. But his DRAPM in his earlier years we have were actually great! For instance, in the JE season-by-season PI RAPM, Kukoc was ranked 67th, 18th, and 27th in the league in DRAPM in 1996-1997, 1997-1998, and 1998-1999 respectively. And he was like 15th in the league in DRAPM in Squared’s 1995-1996 sampling. In terms of actual numbers, Kukoc’s DRAPMs listed above in chronological order for 1996-1999 were: 2.41, 1.29, 3.11, and 2.49. I’d say DRAPM from that era actually portrays Kukoc as a pretty elite defender! And the fact that that held through the 1998-1999 season suggests this isn’t just a collinearity thing.

Some thoughts in response to what#s been offered here ...

Regarding Kerr ... if you trust that number - after accounting for position depending if you're thinking of average as a point or a band you could argue that that's a little above average.

He was -0.38 in 97-14 RAPM (negative bad)
0.2 in 97-22 (positive bad)
0.1 in 97-24 (positive bad)

So a little below break even on all without accounting for position but ... smalls numbers tend to be weaker on that end. That this is later career is a fair contextual note (though I will elaborate on what could in some context be seen as a counterpoint).

I might argue Kerr's sterotype marginally underrated him as a defender. In the 1991 offseason as well as working on his handle he hired an aerobics instructor to help him improve his mobility.

There's probably an argument that the Bulls system with mulitple capable exterior defenders (Brown, Harper, Jordan, Pippen) may have helped cover him and control matchups. On the other it makes him visibly the less rangy, less athletic, less aggressive (and race could be a factor too, I suppose, in terms of visibility) and thereby provide an ungenerous context with a very high bar.

I suppose it depends on the context and expectation too. 94-98 he plays between 22.4 and 24.8 mpg. Given how late in the day MJ ultimately decided on leaving maybe it was initially intended to be less but that's where it fell. Those that would grade on a curve versus norms at that level of rotation could get to a more generous view.

Qualitatively, 2 of the three Barry reviews I glance at wax positive. Now those merely end in a B grade (on a AAA-D scale), but despite some limitations they suggest he's pretty solid at staying in front of his man and knew the schemes well (one is in outright disagreement with an opinion here regarding picks, suggesting he "gets through screens well").

We can only relate to the contexts we actually saw so as above (and below) opinion may vary interpreting him more globally.

I think generously we could call him a "marginal negative" defensively, but that's about as nice as we can get

The "marginal negative" doesn't instinctively seem at all wild ... that that would be "generously ... as nice as we can get" does seem to require taking quite a dim view of the numbers given (a) where I think ... it depends what position we're calling him here but ... "guards" tend to rank on the defensive end on RAPM, (b) that as lessthanjake notes this sample is on older Kerr. Now it's possible to poke holes, he played with a lot of good defensive teammates, maybe RAPM is giving him credit that should go to them. Still, granting the uncertainty, context limitations etc, I think I might be inclined to argue against marginal negative being framed as a best case "the nicest we can get". I suppose at the margins I would note that ending plays without a turnover and with a make are the sorts of things that helped Paxson and Kerr contribute to defenses indirectly.


The Kukoc side stuff is interesting. I think I'd tended to swallow the general idea (not exactly fought against by some Bulls sources) that he was a "soft Euro" who didn't really contribute on that end. My eyeballing those career RAPMs, depending somewhat on data of choice, and with a very crude guess at a fair positional adjustment would be, would seem to suggest at very least the negative reputation was too strong (maybe circa neutral for known career? a touch better? again impression vary depending on version). I don't think I'd particularly looked at JE's individual year numbers from that source (more from Googlesites, which didn't go back that far for yearly numbers) but looking at it I think I have got them and fwiw, that '98 peak defensive number is actually more than matched by AScreaming...'s DRAPM ranks, where he comes out 2nd in NPI and 18th in "RPI". Obviously take individual numbers of this type with a pinch of salt.

[Cavsfansince84]

Rewatched some of the bulls 96/97 PS runs recently and Harpers d stood out to me. Long arms (longer than MJ?), really great at rear contests and picking pockets. Am curious about his earlier cav days, who do you think was a better defender, Ehlo or Harp?
Of course Pippen was a game changer and had a ton of high leverage plays on that end as well. His athleticism before injuries took over was eye popping, making those bron esque chase down blocks. I think your comment on Pip devoting a larger part of his energy to the defensive end makes alot of sense now thinking back.

My impression would be that Harper had a lot more natural ability and athleticism than Ehlo but that Harper was still very young in his Cavs days while Ehlo had a decent rep as a defender back then. It's been so long I couldn't really make any definitive statements about who was better. I only really saw one full year of Harper and even that feels hazy now.

[SportsGuru08]

Note on Harper: he wasn't really a defensive specialist or a guy with a great def rep early in his career. He was pretty good with the Cavs I think but then he went to the Clippers and took on a volume scoring role and I think had b2b season ending knee injuries then went to the Bulls in 95 and was ok. I don't think it was until 96 that he really accepted the role of part time ball handler and def specialist who could also run in transition. That's when he really figured out his spot but he wasn't just a guy playing great defense from day 1 there.

Well that and he didn't really fit the triangle as a scorer, which is what I'm presuming was the reason he was brought over in '94. Even before MJ returned, he was only averaging around 8 or 9 PPG.