VanWest82 wrote:I was going to make this post a while back but sadly I got baited into a suspension by one of the usual suspects. I see these same usual suspects have come to ruin this thread...hardly surprising.
squared2020 wrote:...
Yeah, it's a shame. In the thread Squared was referencing, there's definitely some stuff from certain posters that come across pretty clearly like personal attacks (or at a minimum, passive aggressions and needlessly rude remarks), both toward him and to me as well. Perhaps the posts in this thread weren't quite as bad but... If all this conflict drove Squared2020 off the board, that would be an absolute shame. One of best contributors to this board gone after an argument gone sour (and not the first argument gone sour with this poster).
Squared, although some topics on this board invite belligerent comments, there's always a group willing to thoughtfully engage and remain respectful. You can always put people on your 'ignore' list if you think that would help. Obviously do what you think would make you happy, but there's always space for you here if you want to return to talk basketball!
Anyways...
VanWest82 wrote:Lastly, this graph shows on/off trajectory year-to-year. If we use net on-off as a proxy for impact, it helps illustrate both arugments I made above. Namely that a) role player impact improved (already discussed), b) Scottie, Horace, and BJ all improved post Jordan's first retirement.
The raw stats also corroborate this finding.
In conclusion, I think it's fair to list the second three-peat as Jordan having lots of help, though the playoffs were still largely carry jobs and 98 was a carry job for the first half of the regular season. But it isn't fair to say the first three-peat happened because the supporting cast was just so excellent. They were good, but mostly that team was good because they had Michael Jordan.
Wow, great post! I was just wanting to start looking at on/off inspired the recent Pippen post, but plotting makes it much easier to analyze trends over time.
This kind of post really highlights the kind of analysis we can do with the extended plus minus data Squared2020 gives us before 1997 (play-by-play) and 1994 (Pollack’s total on/off data).
It’s shocking to me how well this on/off corroborates the mean interpretation of Jordan (and not the oddly acrimonious anti-Jordan contingency that’s been growing round these parts). In one plot, you can see....
-Jordan's GOAT level peak and prime,
-91 being an improvement in supporting cast (while Jordan is still GOAT level),
-93 being a down year for the supporting cast (Pippen and Horace Grant in particular),
-94 being a normal year for both of the + an improvement in the rest of the supporting cast (explaining why Jordan's two-year WOWY looks lower than expected),
-96 having a great supporting cast + a GOAT level player in their prime leading to a GOAT level team,
-97 being like 96 but slightly worse from everyone.
... Cool stuff
A few more detailed observations:
1) Jordan continues to look like he has a GOAT level peak and a prime in On/Off and Plus/Minus (On). This has been well documented in other threads, but to get into this a little bit, let’s look at some context:
-Shaq has a On/Off peak of +15.7 (1 year in ’00), a long peak of +12.3 (5 years), and a prime of +11.5 (8 years).
-Duncan has an On/Off peak of +17.8 (1 year in ’05; +14.7 in ’03), a long peak of +12.6 (5 years), and a prime of +11.3 (8 years).
-Garnett has a On/Off peak of +23.6 (1 year in ’03; + 20.7 in ’04), a long peak of +15.0 (5 years), and a prime of +12.9 (8 years).
-LeBron has an On/Off peak of +21.2 (1 year in ’09; +15.3 in ’13), a long peak +15.3 (5 years), and a prime of +14.5 (8 years).
[source: Basketball Reference]
Jordan is absolutely competing with the best of these. From 1988 to 1996, he looks consistently at or above +15.0 On/Off. That’s almost unheard of. Indeed, it’s even more impressive given the strong supporting cast he had during many of the 3-peat years and the GOAT level On rating he was putting up (the On rating correlates better with APM/RAPM than Off).
While he looks around the middle of the pack for 1 year peaks in On/Off (in available years), if Jordan shows any consistency in the missing years (the box stats suggest he would), he would likely be competing for the best of these 3 year peaks and would more likely than not have the best On/Off in these 5–8 year extended peak or regular prime time spans.
2) 1988–1991: We definitely see improvement in the supporting cast. Pippen is the most obvious improvement, although Paxson also jumps. Pippen's improvement from his rookie year into his prime is probably the biggest thing that stands out from the recent Scottie Pippen Plus Minus article Squared just published.
Interestingly, Horace Grant (the first 3-peat Bulls’ third star) looks about the same in his rookie year 1988 vs his prime years in 1991. His minutes did bump from 23 MPG to 34 MPG, so that might explain some of the improved per-game impact. But it would be interesting to check if Grant’s On/Off impact remains strong in 1988 as we get a larger sample. Is this a sign that Grant was ready and should have been playing more minutes earlier in 88, is the '88 success just from a smaller (perhaps more floor-raising) role, or is this just general On/Off noise?
3) 1991–1993: Squared2020’s On/Off data pretty unequivocally supports the idea that 1993 was a major down year for the non-Jordan players, and thus a major floor raising year for Jordan. Jordan’s On/Off is slightly down, in line with someone easing into their post-peak prime years, but they’re absolutely still indicative of a GOAT level prime for someone who’s not in their 5-year peak.
Going from the 1991 supporting cast to the 1993 supporting cast is pretty stark. Cartwright, who was 4th in MPG in 1991, goes from a ~ +2.5 to a -5.0 On/Off. That’s a massive drop off from a major contributor. Paxson also experiences a drop.
Among the co-stars, Pippen has a clear down year in 1993: it’s his worst available prime year in On/Off. Horace Grant has an even bigger drop, falling over -5.0 On/off from 1991 to 1993. 93 is also Grant’s worst available prime year, significantly worse than even his rookie year. This corresponds with a drop off in the box stats for both of them (see e.g. trex's great post on the topic here
https://forums.realgm.com/boards/viewtopic.php?p=107591289#p107591289). There’s a crowd that like to avoid all box stats and stick only with pure impact data. The intentions are understandable. To me, although box stats do have limitations, there are still clear benefits: here, for example, a drop off in a player’s box performance can often correspond to a drop off in their impact.
4) 1993–1994: People love to emphasize two-year WOWY to sell that Jordan is overrated or not GOAT level. Plenty of others (me included) have pushed back, saying that the other contextual changes are significant. I recently published a quick study in the ‘Multi Year WOWY Database’ that showed that two-year WOWY has a mean uncertainty of over 100% (and thus Jordan having a two-year WOWY of +3.32 in 93–94 does not preclude him from having a GOAT level prime).
This On/Off data supports the idea that the other non-Jordan changes were significant. Among the co-stars, Pippen improves in On/Off from 1993 to 1994. Horace Grant improves significantly by well over +5.0 On/Off. Does this mean Jordan was holding them back? Maybe.
But not necessarily. Grant’s On/Off with 91 Bulls with Jordan is just as good as in 94, as are his box stats (see trex’s post). Pippen’s On/Off looks better in 91, 95, 96, and 97 than in 93, and his box stats show a similar trend (see trex’s post). Instead, this just looks like 1993 was an odd down year for the pair, a down year that bounced back in ’94.
Also notice the improved supporting cast from 1993 to 1994. BJ Armstrong remains present, but improves his On/Off slightly in 1994 (as above, note that his On/Off peak is in ’96 once Jordan returned).
Paxson’s minutes at PG get replaced largely by Kerr, a bump of ~ +4.0 On/Off. Big man minutes by Cartwright and Purdue shift to Wennington and Kukoc (whose On/Off would also peak with Jordan), a bump of ~ +12.0 On/Off and ~ +5.0 On/Off respectively. These are significant changes to the team's depth that help contextualize why they did so well in 1994 compared to 1993.
5) 1994–1995–1996: Jordan comes back in late 1995 and gets back to full speed in 1996 (plus the addition of Rodman), and the team-level performance skyrockets to GOAT level. Jordan is once again clearly the best player on the team and in the league. +15.0 On/Off on a GOAT level team is, well, GOAT level.
However, you can see that the supporting cast has improved since the first 3-peat Bulls. Pippen is certainly better in 94–97 than he is in 91–93.
Alongside in-shape Jordan, Luc Longley improves by over +5.0 On/Off. Kukoc looks like the 3rd most valuable Bull in 96, and he also improves in On/Off. Harper improves too. Between the 3 of them, there’s close to a +10.0 On/Off improvement.
Kerr is interesting. After 1994, his On/Off drops (as does Wennington’s in 1995, and perhaps those two drops contribute to the worse team performance without Jordan in 1995). What gives? Did Kerr just peak at age 28 in '94 then fall off? Does the shortened 3 point line take away some of what made Kerr’s shooting impactful? Does Kerr’s value get split with Ron Harper once he joins the team in '95? Does Kerr's play style not fit with Jordan, or does his On/Off get deflated by playing in a rotation that's opposite to Jordan? Not sure here, but this might be interesting to dive into.
Finally: Rodman. Despite the reputation, he does not look like the clear 3rd star in On/Off. Indeed his On/Off is also worse than Kukoc, Longley, and Harper, right in line with 1995 Perdue. I’m also less sure what the best explanation is here. Is this On/Off noise? Perhaps a slow adjustment to the triangle? Are people overrating the value of rebounding? Are we underrating the limitations of a non-passing non-scoring threat? Perhaps Rodman added some playoff resilience that isn’t seen in the regular season, or some subtle scalability that allowed others’ On/Off to flourish without much of a boost to his? Whatever the issue in 1996, Rodman does appear like the clear 3rd star in 1997 On/Off. Perhaps it’s not time to entirely re-interpret those 2nd 3-peat Bulls. Although his 1996 Squared2020 RAPM is also 6th on the Bulls, and his 1997 Goldstein RAPM only goes up to 5th (moving ahead of Luc Longley). Food for thought…
6) 1996–1997. Basically everyone declines going from 1996 to 1997. Rodman is the only exception (along with Kerr, whose value remains at basically -5.0 On/Off either way, so not exactly great). Mind-you, the ’97 Bulls are still in contention for a top 10 team ever, so a drop off from the possible GOAT team still produces great results.
Some of this might be health. The ’97 Bulls struggled much more with health than the ’96 Bulls, and from a WOWY perspective, the healthy ’97 Bulls had a team level performance that was much closer in line with ’96 than the full-season mixed-health results. Rodman only played 55 games, Kukoc 57, Longly 59. Could the worse health cause a decline in players On rating (thus limiting On/Off), or perhaps could the change in lineup order cause a decline in fit (thus limiting the On/Off)? Probably a little.
Jordan has a major drop off going from 96 to 97 (a loss of ~ -7.0 On/Off). Some loss in Jordan’s offensive creation due to aging or regular season coasting could lead to harder shots for the rest of the cast, which might be another possible explanation for the others’ decline. By 1997, Jordan’s On/Off looks identical to Pippen’s. Of course, Jordan’s +8.0 On/Off at age 33 is still better than anything 2010+ Duncan put up at or after age 33 (whom many put as the ~5th best player of all time, just below GOAT level), so the drop off in '97 isn’t exactly precluding Jordan from being GOAT level. He does gain separation from Pippen after adjusting the plus minus data (Jordan’s +5.84 vs Pippen’s +5.44 in Goldstein RAPM).
And we see a major improvement for Jordan in the playoffs. In 1997, Jordan goes from +8.0 RS On/Off -> +23.6 PS On/Off. In 1998, he goes from +6.7 RS On/Off -> 13.1 PS On/off. Is this evidence that Jordan was coasting more as he got older in the regular season, that he revved his motor up more in the playoffs, or evidence of his playoff improvement? Maybe. I wouldn’t want to make this argument just using two years of playoff On/Off (notoriously noisy, even more so small playoff-sized samples). But it could be used as part of a larger argument, if people had other evidence (e.g. box stats, film, more plus minus data).
Anyways, fun stuff to dive into. Thanks as always to Squared2020 for making this data available (and somehow free!), and thanks to VanWest82 for the helpful plot!
Indeed. It's times like this where I wish the moderators were a little more proactive in defending posters and preventing poor behavior, especially when some of the posters involved have a history of making personal attacks (or continuous passive aggressions that just slip under the threshold required to get moderator intervention). Moderating is volunteer basis though, and they don't always see the history, and I suppose there's a philosophical preference to hope people can manage themselves on their own. But when there is a more passive moderation, that allows some amount of toxic history to build up (not necessarily just in this thread, also in other previous ones over time), you get outcomes like this. You get people leaving projects (e.g. ~4 posters, me included, left the latest Peaks project after personal attacks from posters in this thread) and people leaving the forum. 
. Bizarre behavior and personal attacks are going insane in this thread now. All this cuz people who actually like and respect his work told him maybe he'd be have better data focusing on complete seasons vs jumping around damn. Was not on my bingo card this morning...