lessthanjake wrote:
This is amazing stuff! A few observations and questions:
1. I wonder if Jordan has the most +10 on-court rORTG playoff series in NBA history. Based on this thread, it looks like Jordan had 17 of them. I’ve previously ran numbers for some other players using PBPstats data, and I found that LeBron had 16 of them. Nash had the highest average rORTG, but he had fewer series, so he had just 10 series with a +10 on-court rORTG. Steph has 9 of them, unless he added one these playoffs (which is so unlikely that I’m not even bothering to check). Magic likely has a bunch of them. But I guess Kareem probably has the best chance of having more, since he’s probably got a really similar number as Magic on the 1980s Lakers, but then also probably has ones on the Bucks. You’ve tracked rORTG for more players than I have, though, so perhaps you’re aware of more numbers in this regard?
2. Similarly, I wonder if Jordan has the most +15 on-court rORTG playoff series in NBA history. This thread indicates Jordan had 9 of those. By my count, LeBron has 7 of them, Nash has 4 of them, and Steph has 5. Again, my guess is that Kareem is the one with the best chance of having more, but maybe Magic would if the rORTGs with him on the court for the Lakers are a bit better than Kareem.
3. Related to the above, it looks to me like a lot of the analysis people have done here about how good different players’ playoff offenses were in rORTG terms really were missing the context that Jordan’s playoff offenses were absolutely awful with him off the court, which obviously pulls down the overall team numbers.
4. Jordan’s rORTG from 1985-1990 (+5.9) really sticks out to me, given the lack of talent on the team. For reference, by my calculations, LeBron’s on-court rORTG in the playoffs from 2006-2010 was +2.50. Jordan’s on-court rORTG wasn’t entirely consistent in these years, since the Bulls offense with Jordan was bad once and mediocre twice against the Pistons. But against every other team in those years, they really cooked, despite not having much talent. And it’s worth noting that those Pistons teams consistently had far better playoff rDRTGs than they had in the regular season, so some of what’s going on with the Pistons is just that their defense in that era got a lot better in the playoffs. The Bulls/Jordan did really get them back with that +21.8 on-court rORTG in 1991 though.
5. Your post makes me wonder if Jordan’s 1991 regular season on-court rORTG is the highest ever. You mentioned that the Bulls had a +12.7 rORTG with Jordan on the court in the regular season sample we have. Isn’t the sample we have for that season abnormally bad for the Bulls? I believe the Bulls went like 24-1 in the games we don’t have. So if the Bulls had a +12.7 rORTG with Jordan on the court in the sample we do have where the Bulls went 37-20, it feels like it would likely go up even more if we added the remaining games where they went 24-1. Which would be pretty crazy, since +12.7 is already ridiculously high. The highest regular season on-court rORTG I’ve seen is 2005 Nash, with a +14.2. I feel like adding a set of games where the 1991 Bulls went 24-1 is pretty likely to pull Jordan’s rORTG from +12.7 to something above that +14.2 record.
6. Those 1993 playoff numbers are really impressive to me, particularly given that Pippen really wasn’t in good form offensively in those playoffs.
7. Still on 1993, I can’t help but think that a lot of the criticism of Jordan’s 1993 series against the Knicks that we saw in the RPOY project was unwarranted, or at least inconsistent. The criticism has always been that Jordan didn’t shoot well in the series. But he had a positive rTS%, great turnover economy, and a lot of assists, and it appears to have led to the Bulls having a +18.2 on-court rORTG (which also means they had an awful rORTG in Jordan’s off minutes in the series). Of course, the supporting cast shot well in the series, but I feel like the same people that would use that to justify continuing to criticize Jordan’s 1993 Knicks series also lean very hard into team-wide data when arguing that LeBron was still really impactful offensively while genuinely shooting badly in the 2015 playoffs. The on-court rORTG for Jordan in that series is incredibly high, and, unlike in the 2015 playoffs when the Cavs barely did worse with LeBron off the court, it definitely didn’t carry over to when Jordan was off the court. So the argument that Jordan must’ve been impacting the offense in other ways besides scoring efficiently is even stronger. Just some food for thought.
8. As you note, the offensive on-off for Jordan’s career is pretty crazy. We do also see the overall defense being very slightly better with Jordan off (albeit the opposite is true in the title years and in 1985-1990, so the overall fact is basically a result of 1995). Given that the offensive on-off feels almost impossibly good, while the defensive on-off doesn’t seem as good as we’d think, my guess is there may be something systematic causing offenses on both teams to look even worse with Jordan off. One explanation may be that the Bulls tended to lean into particularly defensive lineups with Jordan off. That’s probably part of it. Another explanation I have is contingent on whether you’re assuming equal pace on and off the court. If you are, and the Bulls pace was actually lower with Jordan off the court, then the true offensive off value would be better and the true defensive off value would be worse. It’s a possible explanation, but I don’t have any particular reason to believe it’s right so it’s purely speculative.
9. It looks to me like you’re determining the pace of these series by using the basketball-reference pace. Is that right? Are there ones where you’re not doing that (perhaps because they were tracked by you or Squared)? Relatedly, are you assuming that the teams played equal pace with Jordan on and off the court?
10. Where are you getting the on-court ORTGs for the regular season? The data I find for Jordan in his on-off page on basketball-reference for 1997 and 1998 do not match what you’ve listed (they say 114.2 in 1997 and 107.1 in 1998). And if I divide the “O Point” by the “OPoss” in the Squared data for 1991, 1993, and 1996, they don’t seem to match either. I figure I must be missing something here.
I’ll probably have more thoughts/questions, but those are the ones that come to mind for now.
1. 2. Very doubtful about prime Kareem as before 1980, his teams only had one series above +10 rORtg and that was +16.6 against the Bulls in the 1974 WCF. The only other one close to +10 rORtg is his very first series with the Bucks which is +9.2 against the Sixers in the 1970 ECSF. Especially since Kareem played so many minutes in his prime, his on court ratings are probably very close to the overall team ratings.
Magic is probably the most likely candidate to beat Jordan but given that the Lakers didn't explode offensively till like 1984, he may not have played enough series in total. But I'm currently doing Magic and his ON-OFF based on the bits and pieces we have so far seems to be quite dominant.
3. Yep good point. For guys like Magic and Lebron, their team offenses look a lot better than Jordan's when they sit on the bench.
4. Yea Jordan is often underappreciated for his floor-raising efforts. +5.9 rORtg from 1985-1990 is the most surprising number to me. I didn't expect it to be that high.
5. It's very possible that with full data, Jordan has the regular season record.
If we look at the 1991 Bulls ORtg and DRtg in the sampled vs. all games, we can estimate the team's ratings in the unsampled games. Note that I'm doing a very crude estimate here and assuming that pace is the same in the unsampled games. A better method would use game-by-game pace estimates but I doubt it makes a big difference.
56 Sampled Games: 114.3 ORtg (+6.4), 108.0 DRtg (+0.1), +6.3 Net Rtg
All 82 Games: 114.6 ORtg (+6.7), 105.2 DRtg (-2.7), +9.4 Net Rtg
26 Unsampled Games: 115.2 (+7.3), 99.2 DRtg (-8.7), +16.0 Net Rtg
It's interesting but it seems that in the unsampled games (25-1 MOV, +15.5 MOV), it's the Bulls' defense that is elite and driving these blowouts.
Thus I would not be confident that MJ would pass Nash. I predict it's most likely he ends up in the +13 to +14 range (assuming the same difference between Jordan ON and Jordan OFF) and falls just short. But that would still be the 2nd best ever so that's not too bad.
6. Yea. That +33.5 delta is absolutely ridiculous on a title team.
7. Well yea. I watched and tracked this series recently. MJ's playmaking and ballhandling was really good. A big reason the Bulls won that series is the possession differential and MJ is probably the lowest turnover centerpiece ever. And MJ's presence on the court also helps with offensive rebounding not just with him getting them but his tendency to post at the elbow and attract doubles makes it very easy for his free teammate to slide over to the offensive glass.
8. The Bulls definitely played more defensively without Jordan in the game. And that's not unusual. It happens with most stars. The defensive data is also more noisy than the offensive one. Not sure why that happens.
9. Pace estimates are from Basketball-Reference. This makes for easy comparisons.
10. The Basketball-Reference data is super wonky for 1997 and 1998 and it pissed me off a bit. For instance in 1997, it shows the Bulls have a 114.2 ORtg with Jordan ON, 106.8 ORtg with Jordan OFF and yet the Bulls team total is 114.4 ORtg. That's literally impossible.
I simply used the team data in terms of points scored and allowed with Jordan ON and OFF and calculated the ORtg and DRtg (*48/minutes * 100/pace). It gives results that are actually sensible and fit with the overall team data on Basketball-Reference. For instance in 1997, the Bulls with Jordan have a 116.3 ORtg and without him a 106.9 ORtg. Check and do a weighted average by minutes played with and without MJ and you get a team result of 114.4 ORtg.
As far as the Squared2020 data, I noticed that both offensive ratings and defensive are much lower and pace much higher than on Basketball-Reference. I inquired with him as to why and he explained to me that this has to do with the method of tracking possessions. Every time there is a substitution mid-possession and he estimates that this happens 4-5 times a game, it is counted as two possessions, one with the old lineup and one with the new lineup. For example Harper/Jordan/Pippen/Rodman/Longley are on the court on offense. They use 15 seconds on the shot clock and the ball goes out of bounds off of the opposing team. Phil takes out Rodman and puts in Kukoc. The last 9 seconds of the offense are a new possession so the Bulls attack is two offensive possessions instead of one! I asked him why he did it this way instead of counting say two half-possessions (would avoid the over-counting of possessions) and he said because doing half-possessions would produce very wonky results for lineups with very few possessions. You could have stints with like 600 ORtg if the team scored a 3pt shot with a half-possession. That could totally mess up RAPM and elevate some random scrubs to the top of the list. Long story short, possessions are inflated in Squared2020's tracking which makes offenses seem weak, defenses look insane and slightly depresses Net Rtg as well.
For example, from the 1996 Bulls data which has all 82 games done:
Squared2020: 109.3 ORtg (+1.7), 96.7 DRtg (-10.9), 12.6 Net Rtg, 96.1 Pace
Basketball-Reference: 115.2 ORtg (+7.6), 101.8 DRtg (-5.8), +13.4 Net Rtg, 91.1 Pace
So what I did is take the team points scored and allowed with and without Jordan and used the Basketball-Reference pace estimates to compute ORtg and DRtg. And this way the ORtg and DRtg with Jordan ON and Jordan OFF are in agreement with the overall team data.
