RealGM

99’ - Robinson
03’ - ??
05’ - Ginobili
07’ - Parker
14’ - ??

99 - Robinson
03 - Robinson
05 - Manu
07 - Manu
14 - Kawhi

Well, if we go by raw +/- during the playoff runs that led to titles, here are the #2's from each of the 5 chips, with the #1's in parentheses):

1999 - Avery Johnson (David Robinson)
2003 - Tim Duncan (Manu Ginobili)
2005 - Robert Horry (Manu Ginobili)
2007 - Fabricio Oberto (Manu Ginobili)
2014 - Kawhi Leonard (Manu Ginobili)

Really impressive collection of betas helping Robinson & Ginobili along the way!

99 - D.Rob
03 - Manu
05 - Manu
07 - Manu
14 - Kawhi

Tough to pick a #2 n 03, the support cast was so bad. I certainly wasn't D.Rob though. I considered Parker in 07. 05 Manu is #2 and it's not close, in some playoff series he might even have been #1.

I'm rating over the whole season.

14 - Duncan
03 - D-Rob
05 - Manu or Duncan

99: D-Rob
03: D-Rob (pretty close)
05: Manu (close, Manu had ridiculous +- that year)
07: Manu
14: Duncan (pretty close)

In 2003 the Spurs went like 15-3 in games D Rob missed. Then the next season their defence improved without D.Rob by a large margin. Hard for me to take him, even though there really is no obvious choice. 2003 was Duncan and a bunch of nothing.

One_and_Done wrote:In 2003 the Spurs went like 15-3 in games D Rob missed. Then the next season their defence improved without D.Rob by a large margin. Hard for me to take him, even though there really is no obvious choice. 2003 was Duncan and a bunch of nothing.

I don't know who I'd choose otoh for 2003.

But the arguments here don't seem fair on Robinson.
He's clearly not the Robinson of 1998-2001.
He's playing circa 26mpg (and this limits an already reduced rate impact level in terms of contribution to net impact).

That said I'm not sure of the virture of W-L WoWY when we have plus-minus, on-off and other more granular stuff rooted in that, which will be more accurate. When Robinson's +16 and they lose by one ... well thats a loss for Robinson.

Ditto with the (one way) improvement (net RS improvement smaller). They were a 95.3 (RS) Drtg team with Robinson on the floor in '03. Now as I say the minutes is a legit criticism. But to use their relative weakness in the minutes he's off to minimize his value seems misleading to me. Could one not just as well note Duncan played substantially less in '04 and therefore suggest he was in some way not valuable in '03.

yeah it's probably old DRob for 2003 despite minutes restriction. It's not easy to pick a Robin here but he still brought a reliable elite skill while he was playing. All other players were severely flawed somehow

Owly wrote:

One_and_Done wrote:
But the arguments here don't seem fair on Robinson.
He's clearly not the Robinson of 1998-2001.
He's playing circa 26mpg (and this limits an already reduced rate impact level in terms of contribution to net impact).

That said I'm not sure of the virture of W-L WoWY when we have plus-minus, on-off and other more granular stuff rooted in that, which will be more accurate. When Robinson's +16 and they lose by one ... well thats a loss for Robinson..

The virtue is that its a much larger sample per-game and much less prone to lineup-effects(like Duncan playing alot more with robinson's "backups")

Granularity is not the end-all be-all. Seeing what teams look like entirely without a player as opposed to simply a few minutes without is a great way to avoid dying on dumb hills.

Looking at full season.

'99: Robinson rather easily (Duncan #1)
'03: Manu, can see arguments for Parker, don't really see Robinson's case (Duncan #1)
'05: Duncan, this time it's a tough decision for #1 (Manu #1)
'07: Manu rather easily (Duncan #1)
'14: Most unique championship team ever imo. I'm unsure if every player was individually necessary or if none of them were. I'll go with Manu. Duncan for #3. (Kawhi #1)

eminence wrote:Looking at full season.

'99: Robinson rather easily (Duncan #1)
'03: Manu, can see arguments for Parker, don't really see Robinson's case (Duncan #1)
'05: Duncan, this time it's a tough decision for #1 (Manu #1)
'07: Manu rather easily (Duncan #1)
'14: Most unique championship team ever imo. I'm unsure if every player was individually necessary or if none of them were. I'll go with Manu. Duncan for #3. (Kawhi #1)

What makes you say Manu in 2005 but Duncan in 2007?

One_and_Done wrote:In 2003 the Spurs went like 15-3 in games D Rob missed. Then the next season their defence improved without D.Rob by a large margin. Hard for me to take him, even though there really is no obvious choice. 2003 was Duncan and a bunch of nothing.

If it's not d-rob then who?

OhayoKD wrote:

eminence wrote:Looking at full season.

'99: Robinson rather easily (Duncan #1)
'03: Manu, can see arguments for Parker, don't really see Robinson's case (Duncan #1)
'05: Duncan, this time it's a tough decision for #1 (Manu #1)
'07: Manu rather easily (Duncan #1)
'14: Most unique championship team ever imo. I'm unsure if every player was individually necessary or if none of them were. I'll go with Manu. Duncan for #3. (Kawhi #1)

What makes you say Manu in 2005 but Duncan in 2007?

Both are great that entire stretch, but Duncan normally has a pretty clear minutes edge. In '05 in particular that edge wasn't really there (3.6% more minutes total). In '07 it was back (30.0% lead).

OhayoKD wrote:

Owly wrote:But the arguments here don't seem fair on Robinson.
He's clearly not the Robinson of 1998-2001.
He's playing circa 26mpg (and this limits an already reduced rate impact level in terms of contribution to net impact).

That said I'm not sure of the virture of W-L WoWY when we have plus-minus, on-off and other more granular stuff rooted in that, which will be more accurate. When Robinson's +16 and they lose by one ... well thats a loss for Robinson..

The virtue is that its a much larger sample per-game and much less prone to lineup-effects(like Duncan playing alot more with robinson's "backups")

Granularity is not the end-all be-all. Seeing what teams look like entirely without a player as opposed to simply a few minutes without is a great way to avoid dying on dumb hills.

So I'm not sure I'm understanding you correctly but I'll give the response per my present understanding as it is. [edited typo here]

1) "larger sample per-game"
Well sample of what. The input for both if contrasting on and off is 48 minutes. The output is reduced to a binary W-L. If looking only at the "off" sample there's more per game ... but why would the focus be per game? We're looking at the season. Robinson's out sample has all those games. And it has more. And it doesn't include those bits that are out within the "in".

2) "much less prone to lineup-effects(like Duncan playing alot more with robinson's "backups")"
I'm not sure what this means.

But with regard to Duncan specifically ... this isn't an argument about Duncan. It isn't a comparison of the two. He isn't mentioned apart from in manner where I'd implicitly disapprove of diminishing him by a similarly clunky measure.

There's a broader, generally accepted point about collinearity and the term margin being very important but subject to influence from 9 other players when you're on and 10 when you're off. So rotation and backups is absolutely something to be aware of, particularly with the raw numbers. I guess the point could be "he gets to play with Duncan" which is true but he has a big impact signal across his career including prior to Duncan.

Is W-L WoWY less prone to lineup effects ... thinking it through out loud is it not just subject to different factors that are not related to the player who is ostensibly measured. How good the backup to the backup (or whoever fills the vacated minutes) is? How flexible/resourceful the front office and/or coach is? But against a smaller range of opponents. With a smaller sample. And a binary output. And is not all impact side stuff is the result of particular lineups versus particular lineups and therefore noisy and regarding contextual value?

What am I missing?

Owly wrote:

OhayoKD wrote:

Owly wrote:But the arguments here don't seem fair on Robinson.
He's clearly not the Robinson of 1998-2001.
He's playing circa 26mpg (and this limits an already reduced rate impact level in terms of contribution to net impact).

That said I'm not sure of the virture of W-L WoWY when we have plus-minus, on-off and other more granular stuff rooted in that, which will be more accurate. When Robinson's +16 and they lose by one ... well thats a loss for Robinson..

The virtue is that its a much larger sample per-game and much less prone to lineup-effects(like Duncan playing alot more with robinson's "backups")

Granularity is not the end-all be-all. Seeing what teams look like entirely without a player as opposed to simply a few minutes without is a great way to avoid dying on dumb hills.

So I'm not sure I'm understanding you correctly but I'll give the response per my present understanding as it is. [edited typo here]

1) "larger sample per-game"
Well sample of what. The input for both if contrasting on and off is 48 minutes. The output is reduced to a binary W-L. If looking only at the "off" sample there's more per game ... but why would the focus be per game? We're looking at the season. Robinson's out sample has all those games. And it has more. And it doesn't include those bits that are out within the "in".

Because supporting casts/teammates that hold up for a few minutes a game can collapse when forced to sustain performance for the whole game or large stretches of whole games(see: 2018 Cavs). Alternately, supporting casts/teammates that look hapless without their star can adjust with some time to prepare for said absence. That's why it's notable when what on/off suggests about teammates and stars is not reflected in what WOWY suggests.

There is also the matter of when players affect their team(or the opposing one) in a way which their teammates can benefit from when they're on the bench. Like say a floor-general yelling instructions from the bench(ex: Draymond, Lebron) or someone putting defenders in foul-trouble/wearing them out(ex: Shaq, Giannis).

And then there is the matter of platooning and staggering. Things which stats like rapm or whatever can mitigate but are not directly measuring and have to artificially drag down data-points to try and account for.

It's good to look at and test things against inclusive data, and it's hard to get more inclusive than WOWY(especially if it's full-season). With adjustments, you lose that inclusivity. When you choose things to exclude, you create the potential for bias.

2) "much less prone to lineup-effects(like Duncan playing alot more with robinson's "backups")"
I'm not sure what this means.

But with regard to Duncan specifically ... this isn't an argument about Duncan. It isn't a comparison of the two. He isn't mentioned apart from in manner where I'd implicitly disapprove of diminishing him by a similarly clunky measure.

There's a broader, generally accepted point about collinearity and the term margin being very important but subject to influence from 9 other players when you're on and 10 when you're off. So rotation and backups is absolutely something to be aware of, particularly with the raw numbers. I guess the point could be "he gets to play with Duncan" which is true but he has a big impact signal across his career including prior to Duncan.

He has a big impact signal prior to getting injured and seeing his conventional numbers and minutes go down in the aftermath. Then the Spurs go from being greatly perturbed by his absence from games...to completely fine without him.

Is W-L WoWY less prone to lineup effects ... thinking it through out loud is it not just subject to different factors that are not related to the player who is ostensibly measured. How good the backup to the backup (or whoever fills the vacated minutes) is?

Yes. How good the backup is remains a factor, but that is not a lineup-effect, that is a roster-effect; something which was still there when you were using on/off but is harder to identify or curve for. Moreover, when it's a matter of the game as opposed to a set of lineups, you also get to see more of how other players besides that backup perform without you. Perhaps someone you share the court with a bunch can actually replace a lot of what you do. Perhaps your fellow starters cannot. Maybe that is something that would be true league-wide, maybe it isn't; but it becomes more apparent when the team does not have access to your skills for a longer duration.

This also helps for when players see their effectivness limited contextually; for example when they are moved a postion(or even two) up due to a lack of backup paint-protection. When players stagger star teammates, the on/off generally suffers. But take one of them out and you can see how much of what the player offers is actually replaceable by the star working against their lineups.

How flexible/resourceful the front office and/or coach is? But against a smaller range of opponents. With a smaller sample. And a binary output. And is not all impact side stuff is the result of particular lineups versus particular lineups and therefore noisy and regarding contextual value?

[/quote][/quote]
Not neccesarily. WOWY(or wowy-adjacent) samples are potentially larger than on/off ones. You quite literally cannot get larger than the season before or after a player's arrival for example. But there is more to sampling than just accumulative whatever. Time is an important dimension too. A smaller total sample may yearn unique insight if it is more concentrated.

99- Robinson
03-Robinson
05: Manu
07:Manu
14:Duncan

1999
Regular Season: David Robinson Playoffs: David Robinson

2003
Regular Season: Tony Parker Playoffs: Stephen Jackson

2005
Regular Season: Manu Ginobili Playoffs: Tim Duncan

2007
Regular Season: Manu Ginobili Playoffs: Manu Ginobili

2014
Regular Season: Tim Duncan Playoffs: Any 4 of these players(Manu Ginobili, Tim Duncan, Kawhi Leonard, Tony Parker) No player was particularly better than the other this season.

2005: Tim Duncan

This thread just makes me realise how optimised that 2014 Spurs team was come Playoffs/Finals.