RealGM

04’ Spurs
05’ Heat
06’ Mavs
07’ Suns
08’ Lakers
09’ Magic
10’ Celtics

Heat
Suns
Mavs
Spurs
Magic
Lakers
Celtics

A few stats to help set tiers/levels:
RS SRS
04 Spurs 7.51
08 Lakers +7.34
07 Suns +7.27
09 Magic +6.49
06 Mavs +5.96
05 Heat +5.76
10 Celtics +3.37

PS rNET Rating
07 Suns +10.4
06 Mavs +10.2
08 Lakers +9.7
08 Celtics +8.9
09 Magic +8.6
04 Spurs 7.2
05 Heat 6.2

Overall SRS (RS + 7xPS):
06 Mavs +8.87
07 Suns +8.83
09 Magic +8.13
08 Lakers +8.08
10 Celtics +7.93
04 Spurs +7
05 Heat +6

So the 05 Heat seem fairly consistently near the bottom, and I think that fits with my impression of them.
The Celtics are the clear worst regular season team (hurt by coasting and a drop during injured Garnett games), but rise up in the playoffs.
The Suns look pretty great -- and played the 2007 championship Spurs the best of any team those Spurs faced, it what was a winnable loss in 6.
The Lakers are fairly consistently in the upper half; they traded for Pau Gasol mid-season and went on a fairly historic run the next season after they had time to adjust, so I think their playoff championship odds might be a little higher than the full-season average (which includes pre-Gasol) would suggest.
The Spurs are a mixed bag, with one of the best regular seasons and one of the worst playoffs. One of the best defenses ever, but a team whose offense fell against the 04 Lakers. Perhaps those Lakers were a little better than the regular season suggested with a coasting Shaq, but perhaps they aren't the top team either.
The Mavs very nearly won the title, and had one of the best playoffs statistically, but also had some resilience issues in longer samples / in surrounding seasons, so I'm a little lower on them than the stats suggest.
The Magic were probably an underrated team historically, particularly 2010, and as they said in the recent Thinking basketball podcast on peak Dwight, were ahead of their time in coaching ad three point shooting.

I think I might go something like this, with uncertainty bars:
1. 07 Suns (I think most likely the best team)
2. 08 Lakers (a resilient team; the same team would go on to have an all-time run the next year with a bit more time to develop chemistry)
3. 06 Mavs
4. 04 Spurs (defaulting to the dynasty team in an off year and the team with the better regular season over the 09 Magic team; 3-5 are close and could be switched)
5. 09 Magic
6. 10 Celtics (might be higher with better health)
7 .05 Heat are near bottom

DraymondGold wrote:A few stats to help set tiers/levels:
RS SRS
04 Spurs 7.51
08 Lakers +7.34
07 Suns +7.27
09 Magic +6.49
06 Mavs +5.96
05 Heat +5.76
10 Celtics +3.37

PS rNET Rating
07 Suns +10.4
06 Mavs +10.2
08 Lakers +9.7
08 Celtics +8.9
09 Magic +8.6
04 Spurs 7.2
05 Heat 6.2

Overall SRS (RS + 7xPS):
06 Mavs +8.87
07 Suns +8.83
09 Magic +8.13
08 Lakers +8.08
10 Celtics +7.93
04 Spurs +7
05 Heat +6

So the 05 Heat seem fairly consistently near the bottom, and I think that fits with my impression of them.
The Celtics are the clear worst regular season team (hurt by coasting and a drop during injured Garnett games), but rise up in the playoffs.
The Suns look pretty great -- and played the 2007 championship Spurs the best of any team those Spurs faced, it what was a winnable loss in 6.
The Lakers are fairly consistently in the upper half; they traded for Pau Gasol mid-season and went on a fairly historic run the next season after they had time to adjust, so I think their playoff championship odds might be a little higher than the full-season average (which includes pre-Gasol) would suggest.
The Spurs are a mixed bag, with one of the best regular seasons and one of the worst playoffs. One of the best defenses ever, but a team whose offense fell against the 04 Lakers. Perhaps those Lakers were a little better than the regular season suggested with a coasting Shaq, but perhaps they aren't the top team either.
The Mavs very nearly won the title, and had one of the best playoffs statistically, but also had some resilience issues in longer samples / in surrounding seasons, so I'm a little lower on them than the stats suggest.
The Magic were probably an underrated team historically, particularly 2010, and as they said in the recent Thinking basketball podcast on peak Dwight, were ahead of their time in coaching ad three point shooting.

I think I might go something like this, with uncertainty bars:
1. 07 Suns (I think most likely the best team)
2. 08 Lakers (a resilient team; the same team would go on to have an all-time run the next year with a bit more time to develop chemistry)
3. 06 Mavs
4. 04 Spurs (defaulting to the dynasty team in an off year and the team with the better regular season over the 09 Magic team; 3-5 are close and could be switched)
5. 09 Magic
6. 10 Celtics (might be higher with better health)
7 .05 Heat are near bottom

Personally I'm not sure I'd be up for a 7x multiplier on playoff numbers.

But otoh the bigger issue here is SRS is theoretically a neutral schedule and whilst teams do vary night to night, I think in the majority of cases over a large sample that will be pretty close to evening out, not deviating too much.

But with that 7x multiplier and a "relative" playoff number it becomes really important that the playoff iteration is getting an accurate read on opponents.

With only a small number of opponents - here as few as two - if something is changed and consistently off - that thing is not just impacting probably one game in 82 like SRS - it's throwing off maybe half of 7x weighted sample.
For instance an RS number might derive the '04 Lakers RS strength off a dataset where Bryant, O'Neal and Malone (as individuals, averaged out) play, I think (calculated from Reference) 52.7490976% of possible minutes (Malone pulls it down but Shaq and Kobe are both around 62%). In the Spurs series that number goes to 87.5%.

Playoffs evaluation is complicated. Fair aggregation can be complicated ...

Owly wrote:

DraymondGold wrote:A few stats to help set tiers/levels:
RS SRS
04 Spurs 7.51
08 Lakers +7.34
07 Suns +7.27
09 Magic +6.49
06 Mavs +5.96
05 Heat +5.76
10 Celtics +3.37

PS rNET Rating
07 Suns +10.4
06 Mavs +10.2
08 Lakers +9.7
08 Celtics +8.9
09 Magic +8.6
04 Spurs 7.2
05 Heat 6.2

Overall SRS (RS + 7xPS):
06 Mavs +8.87
07 Suns +8.83
09 Magic +8.13
08 Lakers +8.08
10 Celtics +7.93
04 Spurs +7
05 Heat +6

So the 05 Heat seem fairly consistently near the bottom, and I think that fits with my impression of them.
The Celtics are the clear worst regular season team (hurt by coasting and a drop during injured Garnett games), but rise up in the playoffs.
The Suns look pretty great -- and played the 2007 championship Spurs the best of any team those Spurs faced, it what was a winnable loss in 6.
The Lakers are fairly consistently in the upper half; they traded for Pau Gasol mid-season and went on a fairly historic run the next season after they had time to adjust, so I think their playoff championship odds might be a little higher than the full-season average (which includes pre-Gasol) would suggest.
The Spurs are a mixed bag, with one of the best regular seasons and one of the worst playoffs. One of the best defenses ever, but a team whose offense fell against the 04 Lakers. Perhaps those Lakers were a little better than the regular season suggested with a coasting Shaq, but perhaps they aren't the top team either.
The Mavs very nearly won the title, and had one of the best playoffs statistically, but also had some resilience issues in longer samples / in surrounding seasons, so I'm a little lower on them than the stats suggest.
The Magic were probably an underrated team historically, particularly 2010, and as they said in the recent Thinking basketball podcast on peak Dwight, were ahead of their time in coaching ad three point shooting.

I think I might go something like this, with uncertainty bars:
1. 07 Suns (I think most likely the best team)
2. 08 Lakers (a resilient team; the same team would go on to have an all-time run the next year with a bit more time to develop chemistry)
3. 06 Mavs
4. 04 Spurs (defaulting to the dynasty team in an off year and the team with the better regular season over the 09 Magic team; 3-5 are close and could be switched)
5. 09 Magic
6. 10 Celtics (might be higher with better health)
7 .05 Heat are near bottom

Personally I'm not sure I'd be up for a 7x multiplier on playoff numbers.

But otoh the bigger issue here is SRS is theoretically a neutral schedule and whilst teams do vary night to night, I think in the majority of cases over a large sample that will be pretty close to evening out, not deviating too much.

But with that 7x multiplier and a "relative" playoff number it becomes really important that the playoff iteration is getting an accurate read on opponents.

With only a small number of opponents - here as few as two - if something is changed and consistently off - that thing is not just impacting probably one game in 82 like SRS - it's throwing off maybe half of 7x weighted sample.
For instance an RS number might derive the '04 Lakers RS strength off a dataset where Bryant, O'Neal and Malone (as individuals, averaged out) play, I think (calculated from Reference) 52.7490976% of possible minutes (Malone pulls it down but Shaq and Kobe are both around 62%). In the Spurs series that number goes to 87.5%.

Playoffs evaluation is complicated. Fair aggregation can be complicated ...

Yeah, I get those concerns. The actual numbers were taken from Sansterre's overall SRS list (https://forums.realgm.com/boards/viewtopic.php?t=2012241). It's probably the most in-depth study of the GOAT teams ever, be it statistically or contextually, but as sansterre himself admitted pretty constantly, it's just a formula. There are a few wrinkles I didn't mention (e.g. the playoff SRS are calculated rolling by series rather than all at one, to better catch playoff evolution and avoid downgrading first round losses too much compared to teams who mossed the playoffs). The weights weren't optimized to produce the most accurate results (just set by-eye to get the approximate right ranking and fit people's intuitive preference for high playoff weighting) -- so there's uncertainty and things it misses (e.g. injuries).

It's far from perfect. But I do still think it's helpful to get a rough sense of a team, even if there are contexts that can throw it off, and enough uncertainty to flip the order of teams when they're close.

For the Spurs specifically, I felt like their playoff performance was better than overall SRS was rating it, in part because I thought those Lakers opponents were better in the playoffs than the stat thought, and so raised them up a bit. I didn't dive as deeply into the minutes, but it's good to see there's some evidence behind the feeling I had.

Do you have the 04 Spurs higher than I did on my rough / first-draft list?

DraymondGold wrote:

Owly wrote:

DraymondGold wrote:A few stats to help set tiers/levels:
RS SRS
04 Spurs 7.51
08 Lakers +7.34
07 Suns +7.27
09 Magic +6.49
06 Mavs +5.96
05 Heat +5.76
10 Celtics +3.37

PS rNET Rating
07 Suns +10.4
06 Mavs +10.2
08 Lakers +9.7
08 Celtics +8.9
09 Magic +8.6
04 Spurs 7.2
05 Heat 6.2

Overall SRS (RS + 7xPS):
06 Mavs +8.87
07 Suns +8.83
09 Magic +8.13
08 Lakers +8.08
10 Celtics +7.93
04 Spurs +7
05 Heat +6

So the 05 Heat seem fairly consistently near the bottom, and I think that fits with my impression of them.
The Celtics are the clear worst regular season team (hurt by coasting and a drop during injured Garnett games), but rise up in the playoffs.
The Suns look pretty great -- and played the 2007 championship Spurs the best of any team those Spurs faced, it what was a winnable loss in 6.
The Lakers are fairly consistently in the upper half; they traded for Pau Gasol mid-season and went on a fairly historic run the next season after they had time to adjust, so I think their playoff championship odds might be a little higher than the full-season average (which includes pre-Gasol) would suggest.
The Spurs are a mixed bag, with one of the best regular seasons and one of the worst playoffs. One of the best defenses ever, but a team whose offense fell against the 04 Lakers. Perhaps those Lakers were a little better than the regular season suggested with a coasting Shaq, but perhaps they aren't the top team either.
The Mavs very nearly won the title, and had one of the best playoffs statistically, but also had some resilience issues in longer samples / in surrounding seasons, so I'm a little lower on them than the stats suggest.
The Magic were probably an underrated team historically, particularly 2010, and as they said in the recent Thinking basketball podcast on peak Dwight, were ahead of their time in coaching ad three point shooting.

I think I might go something like this, with uncertainty bars:
1. 07 Suns (I think most likely the best team)
2. 08 Lakers (a resilient team; the same team would go on to have an all-time run the next year with a bit more time to develop chemistry)
3. 06 Mavs
4. 04 Spurs (defaulting to the dynasty team in an off year and the team with the better regular season over the 09 Magic team; 3-5 are close and could be switched)
5. 09 Magic
6. 10 Celtics (might be higher with better health)
7 .05 Heat are near bottom

Personally I'm not sure I'd be up for a 7x multiplier on playoff numbers.

But otoh the bigger issue here is SRS is theoretically a neutral schedule and whilst teams do vary night to night, I think in the majority of cases over a large sample that will be pretty close to evening out, not deviating too much.

But with that 7x multiplier and a "relative" playoff number it becomes really important that the playoff iteration is getting an accurate read on opponents.

With only a small number of opponents - here as few as two - if something is changed and consistently off - that thing is not just impacting probably one game in 82 like SRS - it's throwing off maybe half of 7x weighted sample.
For instance an RS number might derive the '04 Lakers RS strength off a dataset where Bryant, O'Neal and Malone (as individuals, averaged out) play, I think (calculated from Reference) 52.7490976% of possible minutes (Malone pulls it down but Shaq and Kobe are both around 62%). In the Spurs series that number goes to 87.5%.

Playoffs evaluation is complicated. Fair aggregation can be complicated ...

Yeah, I get those concerns. The actual numbers were taken from Sansterre's overall SRS list (https://forums.realgm.com/boards/viewtopic.php?t=2012241). It's probably the most in-depth study of the GOAT teams ever, be it statistically or contextually, but as sansterre himself admitted pretty constantly, it's just a formula. There are a few wrinkles I didn't mention (e.g. the playoff SRS are calculated rolling by series rather than all at one, to better catch playoff evolution and avoid downgrading first round losses too much compared to teams who mossed the playoffs). The weights weren't optimized to produce the most accurate results (just set by-eye to get the approximate right ranking and fit people's intuitive preference for high playoff weighting) -- so there's uncertainty and things it misses (e.g. injuries).

It's far from perfect. But I do still think it's helpful to get a rough sense of a team, even if there are contexts that can throw it off, and enough uncertainty to flip the order of teams when they're close.

For the Spurs specifically, I felt like their playoff performance was better than overall SRS was rating it, in part because I thought those Lakers opponents were better in the playoffs than the stat thought, and so raised them up a bit. I didn't dive as deeply into the minutes, but it's good to see there's some evidence behind the feeling I had.

Do you have the 04 Spurs higher than I did on my rough / first-draft list?

Can't go too deep ... life stuff etc.

Basic thoughts and answer to the question asked. Caveat ... I'm not too deep into this ... my "thing" is more just caution with the playoffs than any actual solution.

1) "Rolling" not at first glance convinced of the merits. Unless I'm missing something (and I haven't dived deep here) this just seems even worse to me. Treating a team like a whole over an 82 game sample then significantly changing them ... (based on the last series, I guess?) ... for what they're like maybe 20 days later. Is "rolling" solving one bug by introducing another ... IDK (and off the description here seems like high weighting on playoffs would be the cause of dinging first round exits too much versus not making it).

Still may well be better than things people call "playoff SRS" where, as I understand it, there's no symmetry to it. If you think '94 Denver are great because they beat this powerful (RS) Supersonics team and the Supersonics are awful because they lost to this middling (RS) Denver team that seems out of whack.

2) A couple of otoh maybe satisficing solutions ...
a) Just SRS (or points dif) and playoff points dif.
pros
-A good team doesn't feel the need to mash -2 SRS team by 2 points more than another team facing a +2 side ... often they've kind of earned that opponent ... I don't love then punishing them for the RS success (worse if heavy weighting per original post)
- A simple numbers, easily, presently available as a baseline you could then adjust from.

cons
- I really don't love it because playoff opponent difficulty varies wildly.
- Whilst routes correlate pretty well with what is "deserved", weak conference, division lead seedings, changes in team standard in-season, luck etc mean that won't always be the case

b) Proper SRS (as calculated for RS) on the full seasons data
pros
-Does mean consistency in team (and opponent) perception rather than messing around with the playoffs.

Cons
- Doesn't solve roster/coaching/injury changes.
- Doesn't seem to be (to my knowledge) out there already.

c) Player-based models? (Or lineup based? but data limitations on full lineups so noise etc)
Don't know how do-able it is. But you could improve regular SRS by knowing who opponents put out there (and also maybe eliminate garbage time). And then if the who's out there changes in the playoffs you accurately convey opponent strength.

a and b would be best complemented by knowledge of opponent health, roster composition etc. How much anyone has that off the top of their heads and for all NBA history. I don't know if there's a simple numerical adjustment possible for that though (kinda hence option C). I may be missing something here though.


Tangent
'04 Spurs may have been a regular season of two halves. Per Hollinger they were +260 (translating to +6.341463415 per game) in the first 41 - +332 (8.097560976 per game). Don't know how unusual that shift is but therein lies an important point - most of us don't know how much a team typically varies from the norm over 4 to 7 game samples - I imagine ideally a model would build that in. Anyhow Hollinger suggests a Duncan injury made them from a pretty static post-Duncan and stand out at the arc offense to a "motion-oriented offense" - though at first glance Manu (who with Parker is noted as being made better use of) doesn't seem to seem to have benefited statistically.

Other odd notes
Spurs are only defending champs so could conceivably ease off RS, though not insignificant turnover might somewhat mitigate against that. Turnover might explain the above too. Boston were recent champs and older so they too have a claim to being undersold by their regular season.

I will say looking at the numbers '09 Magic look better than I'd have remembered. And that SRS is marginally improved upon the next year. I think I'd thought of Boston and LeBron's Cavaliers as the top tier and Orlando sneaking through but they were good too.


I'm generally more interested in process than outcome but as you ask ... you don't seem rigidly tied to the Sansterre "Overall SRS" but I imagine in might still have an anchor there? I might guess I would be higher on the Spurs is what I'm saying. They lost 4 percentage points from the line at the wrong time (the playoffs ... but then the Lakers do worse ... don't know if it's a change in who's playing and getting fouled or luck in either case). Lakers are a tougher opponent than on paper. Their seeding is pretty rough for a 7.51. They did have HCA over LA ... they weren't unlucky with regards to points dif over the series ... at the same time it's pretty easy to flip G5 and see them in G7 on their home court. Hedo shouldn't be as bad as he looks in the playoffs over a larger sample (there'll typically be some variance both directions over a roster, and I'm not looking systematically but it pops as a pretty bad run for him). So probably inclined to go higher but that's without a proper process or actually having to list them out.

Anyway I really like that you grant uncertainty the list doesn't look bad eyeballing it. Like you I could, and for me probably might go higher on the Celtics. I do think there was probably some conservation going on. Scalabrine lineups probably don't reflect what you can reasonably expect to see in the playoffs. And that core does have a title (and is old) already as I discussed.

Sorry, that was rambly. Anyway got to go.