Is it posted somewhere?

Thanks
John

Oddly enough (or maybe not oddly) the term actually behaved backwards. What I found was that a term of AST%*USG% works far better in an SPM regression! What does this mean? Apparently, the player's offensive impact is related to his versatility in BOTH using possessions and dishing assists. The best term for assists is actually of the form AST%*USG%*PPP, where PPP is points per possession for the player (accounting for turnovers)--otherwise, inefficient players are benefited at Steve Nash's expense.

Out of interest did the (100-USG%)*(AST%/100) term have a positive coefficient (or if modelled in a non-linear way did it kick upward for high values)? I imagine that the term holds at least some information content on "importance" and how much a team might suffer when the type of player with a high rating is off the floor.

Player           USAGE   Adjusted Usage
LeBron James     33.5%     39.6%
Dwyane Wade      34.9%     40.1%
Kevin Durant     32.0%     29.6%
Chris Paul       22.2%     34.0%
Chris Bosh       28.7%     26.8%
Carmelo Anthony  33.3%     32.2%
Kobe Bryant      32.3%     33.4%
Dirk Nowitzki    28.8%     25.6%
Brandon Roy      26.8%     29.5%
Dwight Howard    23.9%     23.1%
Pau Gasol        21.4%     21.3%
Steve Nash       22.9%     36.9%
Deron Williams   23.8%     33.0%
David Lee        23.8%     22.6%
Zach Randolph    24.6%     22.5%
Amare Stoudemire 27.4%     23.0%
Gerald Wallace   20.3%     18.7%
Joe Johnson      26.4%     28.3%
Tim Duncan       26.1%     25.5%
Tyreke Evans     26.2%     31.1%
Rajon Rondo      20.2%     30.7%
Gilbert Arenas   32.0%     38.2%
Monta Ellis      29.4%     31.2%
Derrick Rose     27.2%     31.4%
Danny Granger    28.7%     26.5%
Josh Smith       22.2%     23.0%
Brook Lopez      23.6%     21.9%
Chauncey Billups 24.3%     28.7%
Andre Iguodala   21.8%     25.0%
Carlos Boozer    24.8%     22.6%
Andrew Bogut     23.2%     21.3%
Manu Ginobili    25.8%     29.6%
John Salmons     22.7%     22.9%
Marc Gasol       16.9%     16.4%
Russell Westbrook25.7%     34.4%
David West       24.0%     22.7%
LaMarcus Aldridge22.9%     20.4%
Baron Davis      24.3%     33.1%
Paul Pierce      23.8%     23.8%

Player USAGE Adjusted Usage
LeBron James 33.5% 39.6%
Kevin Durant 32.0% 29.6%
Dwyane Wade 34.9% 40.1%
Dirk Nowitzki 28.8% 25.6%
Dwight Howard 23.9% 23.1%
Steve Nash 22.9% 36.9%
Kobe Bryant 32.3% 33.4%
Chris Bosh 28.7% 26.8%
Amare Stoudemire 27.4% 23.0%
David Lee 23.8% 22.6%
Zach Randolph 24.6% 22.5%
Carmelo Anthony 33.3% 32.2%
Deron Williams 23.8% 33.0%
Rajon Rondo 20.2% 30.7%
Brook Lopez 23.6% 21.9%
Tim Duncan 26.1% 25.5%
Josh Smith 22.2% 23.0%
Gerald Wallace 20.3% 18.7%
Joe Johnson 26.4% 28.3%
Andre Iguodala 21.8% 25.0%
Derrick Rose 27.2% 31.4%
Russell Westbrook 25.7% 34.4%
Brandon Roy 26.8% 29.5%
David West 24.0% 22.7%
Carlos Boozer 24.8% 22.6%
Pau Gasol 21.4% 21.3%
Tyreke Evans 26.2% 31.1%
Chauncey Billups 24.3% 28.7%
Al Horford 17.6% 16.5%
LaMarcus Aldridge 22.9% 20.4%
Manu Ginobili 25.8% 29.6%
Rudy Gay 22.5% 21.2%
Andre Miller 23.8% 29.9%
Nene Hilario 16.4% 15.6%
Baron Davis 24.3% 33.1%
Aaron Brooks 25.7% 28.6%
Stephen Curry 21.8% 25.3%
Jason Kidd 14.5% 22.7%
Monta Ellis 29.4% 31.2%

Nice work on realizing the term should be (100-USG%)*(AST%/100). I used that term in one of my latest SPM regressions.

Research indicates that assisted shots go in at an 8% higher clip than unassisted. ( http://www.82games.com/assisted.htm ). If we consider that, on an assist, half of the usage is to the player that received the assist, and half is to the player that gave out the assist, we can estimate total usage:

Old Usage: = 100 * ((FGA + 0.44 * FTA + TOV) * (Tm MP / 5)) / (MP * (Tm FGA + 0.44 * Tm FTA + Tm TOV))

New Usage: = 100 * ((FGA + 0.44 * FTA + TOV + 0.5*(AST)/(50%) - 0.5*(AST'D)/(FG%+.04) ) * (Tm MP / 5)) / (MP * (Tm FGA + 0.44 * Tm FTA + Tm TOV))

The concept is that the player's assists are 1/2 of his "potential assists" (League average assisted FG% is ~50%) and his "assisted" FG% is his normal FG% + .04.

This isn't entirely the case, but a good rough estimate.

An interesting thing: I have found that the value "Assisted Close Shots should be credited almost entirely to the one giving out the assists, while the value of "Midrange Assisted Shots" have almost nothing to do with the one giving out the assist.

18 - Jerry Sloan (Stockton x 15, Deron x 3)
11 - Pat Riley (Magic x 5, Tim Hardaway x 3, Wade x 3)
9 - Byron Scott (Baron Davis x 1, Kidd x 3, Marbury x 1, CP3 x 4)
5 - Mike D'Antoni (Marbury x 1, Nash x 4)

Interesting that Riley, Scott and D'Antoni could soon be in position to potentially coach LeBron (and/or Wade) next season.

As a quick-hitter for a crude one-number metric I had a look at USG% + (100-USG%)*(AST%/100). The first term obviously representing the percentage of plays used directly by the player while on the floor and the second term being the remaining percentage of plays used by teammates scaled by the proportion of those teammates baskets the player assisted (Being a rough measure, other key items like teammate turnovers and free-throws haven't been worked into the mix).

Since the Magic/Bird era started the highest measure has been 67.0, Chris Paul in 08-09. 21 player seasons have been higher than 60 and the cut-off for the Top 50 is around 57.5. Chris Paul features again at number 4 with a measure of 64.5 while John Stockton's late 80s to mid 1990s take up 6 of the other spots in the Top 10. The other two players in the Top 10 are Mark Price from 94-95 and Steve Nash's 09-10. That these point guards rank so high is no surprise given the decision making responsibilities and number of opportunities created for their teammates. Here is the breakdown of players with multiple entries in the Top 50:

Stockton - 13
Magic - 5
Nash - 4
Wade - 4
LeBron - 3
Iverson - 3
Chris Paul- 2
Deron Williams - 2
Rod Strickland - 2
Stephon Marbury - 2

Back to the topic at hand, Wade's first entry comes in right at number 11 (61.9 in 08-09) while LeBron's 09-10 season is number 14 at 61.3. Wade's 06-07 (61.1) follows at 15. These guys are both accustomed to dominating the ball and making a huge amount of the decisions that end in a recorded statistic. The reasons for this range from personal style of play and ability to create their own unassisted opportunities, quality and skillset of teammates, status, competitiveness,coaching and offensive systems.

Moreso than just the usage rates, it would (will?) be fascinating to see in real life how a coach could mesh these styles of play and how the decision making and ball control would be distributed. Would both guys see a moderate reduction in this type of index or would one guy take the brunt? There are a few less-extreme examples to draw on in the pairs Neil listed. For example when Kidd and Vince joined forces in 2005 Vince took a slight dip from 51.4 in 03-04 to 49.2 in 04-05. Kidd took a bigger drop from 58.4 to 53.3. When Mo Williams joined Lebron his measure went from 45.5 to 38.8 while LeBron stayed around the same (58.3 up slightly to 59). Also, when Payton moved from his fairly decision dominant style of play in the early 2000s (between 53.8-57.2 from 98-99 to 02-03) to the triangle offense in 03-04 his measure dropped heavily to 41.8. This is an indicator of decentralization of decision making which is key to the triangle offense, with none of Jordan/Pippen/Kobe/Shaq having a rating higher than 53.5 while playing in the triangle. Kobe often falls in the 46-49 range, Jordan in the 47-50 range, Pippen in the 42-45 and Shaq in the 41-44 range.