Variation in Supporting Casts

[ElGee]

Not all Supporting Casts are created equally.

Interestingly, the range in supporting cast quality appears to be quite enormous. This jibes with the range of NBA teams that we see in simple SRS observations, scaling from about 15 points worse than opponents to 13 points better.

Supporting casts get better as teams get better -- teams who miss the playoffs have the worse supporting casts, whereas teams who reach the Final 4 have the best supporting casts. The average supporting cast in the NBA (2001-2012) is -5.3/48 minutes. Here's the breakdown:



And in graphical form. Note that finalists have the smallest range in variation between Supporting Casts while teams that reach the PS but are eliminated the 1st round have the largest range in variation:



Without their best players, the worst teams in the NBA have been as bad as -19/48 minutes. This is right in line with what we'd expect based on SRS, as the team would have to only improve to -13/48 with their star (not much of improvement) to finish as a -14.5 SRS team. On the flip side, the best teams in the league have supporting casts that are over +5.0/48 without the best player. This represents a different of about 25 points -- in other words, the difference between the best and worst supporting cast is greater than the impact of the best individual player we've ever observed since +/- has been tracked.

Methodology:

Spoiler:

All results are derived from on/off data and are pts/48 minute based.

"Best player" was defined as the player on a team with over 1500 MP in a season who produced the best net on/off per 48 minutes.

"Supporting Cast" was then defined as the performance of the team per/48 when that player was off the court.

EDIT: There is a typo in the chart -- the mean Finals cast is -2.5

[colts18]

This goes hand in hand with my past thread where i argued that the cast of the average champion is average and maybe below average.

How did you determine the best player?

[B_Creamy]

Got stats for particular teams? Garnett's supporting casts would be good.

[rrravenred]

Interesting (if an obvious point).

An interesting question is what you do with offensively multi-polar teams (Shaq/Kobe being the obvious one) where "best" is relatively narrow margin. I'd also assume that (in general) the "off" number of minutes for the later rounds was relatively small, especially given that the stars tend to play themselves into the ground in the Finals.

Were there any outliers (in either direction) that stood out for you?

Edit: Oh, and excellent work!

[rrravenred]

This goes hand in hand with my past thread where i argued that the cast of the average champion is average and maybe below average.

How did you determine the best player?

Hmmmm... wasn't what I got out of that data at all. The negative average is WITHOUT BEST PLAYER for all compared team i.e. every team is at a net deficit without their best player

You could, however, make an argument that championship teams are optimally specialised to maximise the contributions of their best player so suffer a bigger dropoff when they're not in their set role...

[Texas Chuck]

Got stats for particular teams? Garnett's supporting casts would be good.

considering that's the genesis of the OP Im assuming they are forthcoming.

I dont see whats surprising that the very best supporting cast exceeds the very worst supporting cast by an amount greater than the best individual we've ever seen. Some teams like the 90s Mavs or the recent Bobcats teams simply have no talent at all.

Having the specifics for KG vs say David Robinson would be telling so hopefully he comes thru with it.

[ardee]

Got stats for particular teams? Garnett's supporting casts would be good.

considering that's the genesis of the OP Im assuming they are forthcoming.

I dont see whats surprising that the very best supporting cast exceeds the very worst supporting cast by an amount greater than the best individual we've ever seen. Some teams like the 90s Mavs or the recent Bobcats teams simply have no talent at all.

Having the specifics for KG vs say David Robinson would be telling so hopefully he comes thru with it.

D-Rob would be next to impossible because play by play doesn't exist before 2001.

[An Unbiased Fan]

Got stats for particular teams? Garnett's supporting casts would be good.

I like to look at the efficiency differential of a player's teammates. This data is just an estimate because it's based on position, but still fairly accurate. http://www.hoopsstats.com/basketball/fa ... /1/diffeff

KG's support
2000 = +5.1
2001 = -7.4
2002 = -1.2
2003 = need to isolate KG at SF/PF
2004 = +1.6
2005 = -2.9
2006 = -5.8
2007 = -12.4
2008 = +12.5
2009 = +15.5
2010 = +10.7

[ElGee]

This goes hand in hand with my past thread where i argued that the cast of the average champion is average and maybe below average.

How did you determine the best player?

Inspired by your thread, where I argued that you didn't look at all other supporting casts to make the necessary conclusion about Finalists.

Got stats for particular teams? Garnett's supporting casts would be good.

Yes, and they are available on basketball-reference (2001-) and NBA.com (97-). I guess others asked so here are the particulars:

Kevin Garnett
1996-97 -13.0
1997-98 -5.9
1998-99 -4.4
1999-00 -7.5
2000-01 -5.8
2001-02 -4.9
2002-03 -19.0
2003-04 -11.0
2004-05 -1.3
2005-06 -9.1
2006-07 -11.7
2007-08 3.1
2008-09 3.0
2009-10 -0.3
2010-11 -3.1
2011-12 -3.7

Kobe Bryant
1996-97 4.9
1997-98 9.3
1998-99 5.5
1999-00 5.7
2000-01 -3.7
2001-02 7.5
2002-03 -5.7
2003-04 0.3
2004-05 -4.8
2005-06 -8.0
2006-07 -5.8
2007-08 -0.2
2008-09 -0.4
2009-10 -3.4
2010-11 3.2
2011-12 -2.1

Dirk Nowitzki
1998-99 -3.4
1999-00 -6.6
2000-01 -10.4
2001-02 -2.7
2002-03 -8.0
2003-04 -1.9
2004-05 -6.0
2005-06 -0.6
2006-07 -1.3
2007-08 -6.3
2008-09 -4.4
2009-10 -6.5
2010-11 -6.5
2011-12 -8.4

You can see the PS figures for key players here: viewtopic.php?f=344&t=1248645

By this measure, ITO of career, LeBron and Dirk have had the worst casts of this generations stars. Manu and Kobe have had the best. For 97-98, the Bulls were +1.3 with Jordan out of the game.

We don't have information before 97, so Robinson is just 98-03, aka the Duncan Years, with an average cast of +1.8.

[ElGee]

Interesting (if an obvious point).

An interesting question is what you do with offensively multi-polar teams (Shaq/Kobe being the obvious one) where "best" is relatively narrow margin. I'd also assume that (in general) the "off" number of minutes for the later rounds was relatively small, especially given that the stars tend to play themselves into the ground in the Finals.

Were there any outliers (in either direction) that stood out for you?

Edit: Oh, and excellent work!

There's no PS numbers included. I'm looking at RS performance of the teams and then stratifying by how far they went in the postseason. Perhaps I should note that in the methodology!

There were some outliers: 2 of the 3 best teams by this measure were the 2011 and 2012 Chicago Bulls. Just a super balanced team. The 2012 Bulls (with Luol Deng out) were +7.4! 3.1 standard deviations away from the mean and half a deviation better than the 02 Kings (+5.3).

On the negative side there were two teams that stood out at over 3 standard deviations. The 02 Grizzlies (-18.0 without Shane Battier) and the 03 Wolves (-18.9 with Garnett).

Another interesting note is the 2011 Cavs were -12.6 but the 2010 Cavs -5.6. Now, I'm not saying this metric is a perfect indicator of a true "Supporting Cast" by any measure, but I think people go too far with comparing the 10 and 11 teams with/without James, and that's a huge data point; Basically, the 2010 Cavs had "average" play without LeBron and the 2011 Cavs nearly 2 STDs worse without their best player.

Both the 2003 Spurs and the 2011 Mavs were just below average by this measure with Dirk and Timmy.

Also worth noting is that -1.7 mark we see for Final Four teams. Only 19.7% of teams in the sample were -1.7 or better. I see that as a pretty large piece of evidence that it's hard to make a Conference Finals without a really good team.

[colts18]

How did you determine the best player?

[Dipper 13]

Yes, and they are available on basketball-reference (2001-) and NBA.com (97-). I guess others asked so here are the particulars:

Could you also do LeBron, McGrady, & Iverson? Thanks.

[rrravenred]

How did you determine the best player?

"Best player" was defined as the player on a team with over 1500 MP in a season who produced the best net on/off per 48 minutes.

[colts18]

D-Rob would be next to impossible because play by play doesn't exist before 2001.

Play by play data does exist from 1997-2000. Robinson was strong in those years. In 1997, the Spurs were +21 with him and -8.43 per 48 minutes without him despite having a lot of the same core from the previous season.

[colts18]

Could you also do LeBron, McGrady, & Iverson? Thanks.

This is from 2001-onward:

LeBron: -4.3 per 100
McGrady: -1.5 (-9.8 in 03 and -11.5 in 04)
Iverson: -1.3 (+1.6 in 01)

[An Unbiased Fan]

Kevin Garnett
1999-00 -7.5
2000-01 -5.8
2001-02 -4.9
2002-03 -19.0
2003-04 -11.0
2004-05 -1.3
2005-06 -9.1
2006-07 -11.7
2007-08 3.1
2008-09 3.0
2009-10 -0.3
2010-11 -3.1
2011-12 -3.7

I'm bothered by these results. KG's 2004 supporting cast >>> his 2007 support. Yet both are nearly even in +/- metrics. His 2005 cast better than the 11 & 12 squads? And comparable to the 10 team? 2000 was arguably his best cast in Minny, yet it rated below 02, 03, and 05.

[ElGee]

Supporting Cast by RAPM

Here's another look at Supporting Casts using non box-score data, this time, Jeremiah Englemann's (prior-informed) RAPM from 2002-2012. RAPM has slightly less variance than on/off and the numbers are also weighted by minutes played, which narrows the distribution slightly.

Nevertheless, the average Supporting Cast in the NBA from 2002-2012 based on this methodology was -2.1. The best way to interpret this number is that an average team would be about a -2 SRS without the player contributing most to its SRS gains.



Methodoloy

Spoiler:

"Best Player" is defined by the player with best minute-weighted RAPM for a team for a season. For example, if Player A has 5 RAPM and Player B a 4 RAPM but Player B plays 100% of the minutes and Player A 50% of the team minutes, Player B is credited with a "4" lift to the team and is thus considered their best player. (B would have a +2.5 lift).

"Supporting Cast" is defined by the minute-weighted totals of all other players on the team, summed together.

**For traded players, the total season RAPM value was used

**For minute-weighted totals, only players with 300 MP were considered

[colts18]

Elgee, do you have a list of the best supporting casts overall in that span? Also a list of the worst supporting casts of good teams?

[ElGee]

Elgee, do you have a list of the best supporting casts overall in that span? Also a list of the worst supporting casts of good teams?

Absolutely.

Best:
2008-09 BOS 5.30
2002-03 SAC 4.87
2008-09 LAL 4.51
2002-03 PHI 4.33
2011-12 CHI 4.28
2006-07 PHO 4.17
2010-11 LAL 4.11
2011-12 OKC 3.98
2007-08 BOS 3.93
2008-09 ORL 3.72

Worst of the 6+ MOV teams:
2009-10 CLE -0.43
2005-06 DAL 1.22
2001-02 SAC 1.23
2007-08 LAL 1.31
2001-02 LAL 1.42
2004-05 SAS 1.66
2004-05 PHO 1.84
2005-06 SAS 1.85
2011-12 MIA 2.40
2010-11 CHI 2.51


EDIT: Fixed trade issue

[colts18]

Interesting how the 02 kings went from one of the worst casts to being one of the best the next year