Wanted to add a few thoughts on this, I think the points created idea is a fine one, but the methodology used here is slightly flawed.
First, in general linear weighted metrics tend to be more subjective than objective - to quote Premier, "from the results of this statistics, it seems very useful..." and that can be a trap that many linear weighted models fall into. Try to be as objective as possible in your selection of coefficients and statistics.
Second, the use of FGM is problematic. As an example, suppose two players, player A and player B both take 10 shots. Player A takes all 3 point shots and makes 4, and player B takes all 2 point shots and makes 6.
Player A's PC(points created) is 12 + 1.03(-.71*6) = 7.6122
Player B's PC(points created) is 12 + 1.03(-.71*4) = 9.16
Yet for all intents and purposes in a game, the player's value should be identical. Even more to the point, there is no quantification of free throws taken or missed except under the catch-all of points. A player could go 1 of 100 from the free throw line with no other productive statistics and have a positive points created value.
Next, while I agree with the basic idea of using possessions created and multiplying it by the value of an average possession (also I'm curious how you derived 1.03 for your coefficient - the last time I calculated it was two years ago and I got 1.046, but I'm still curious as to what method you used to get 1.03), quantifying possessions created at the player level can be difficult. I wanted to test this, and here's what I came up with:
In a game, two teams should start with a base number of possessions that is equal. Their possessions created would then add to their total number of possessions, and the difference between the two's possessions should be equal to the difference between their possessions created.
(A + b) - (A + c) = b - c where A is the base possessions and b and c are the possessions created of each team.
In the formula above, (A + b) is equal to FGA + (FTA/2.5) and (A + c) is equal to FGA allowed + (FTA allowed/2.5). I grabbed all of the latest data and plugged it into a spreadsheet. Using these formulas and plugging in your (.75*OReb + .25DReb + Stl + .5Blk - TO - .71*FGMissed) metric, I calculated both sides of the above formula for each team. Here are the results: (apologies for formatting)
Atlanta 0.08 -2.959
Boston (2.30) -2.894
Charlotte 4.48 0.506
Chicago (0.16) -2.425
Cleveland 1.08 1.583
Dallas 1.74 3.989
Denver (0.02) 3.044
Detroit 1.92 4.492
Golden State 1.46 -0.617
Houston (1.12) -0.337
Indiana (1.34) -1.302
LA Clippers (2.48) 2.698
LA Lakers 0.96 1.203
Memphis 0.32 -0.574
Miami (2.06) 3.437
Milwaukee 1.78 -1.466
Minnesota (2.28) 0.591
New Jersey 1.40 -0.435
New Orleans 1.98 -4.219
New York (0.66) 0.887
Orlando (2.22) -1.305
Philadelphia (1.04) 1.252
Phoenix (2.92) 1.322
Portland (2.52) -3.821
Sacramento (1.80) -1.726
San Antonio (1.16) 2.861
Seattle 1.16 -2.919
Toronto 0.78 -2.413
Utah 2.22 -1.725
Washington 2.80 2.205
The first column is the actual possession differential, and the second column is the possessions created differential. I then plugged the points into a scatter plot and used a linear regression to get an R-squared value. Here's a link to the graph:
Graph
As you can see, the R-squared value is .002, which definitively shows there is no correlation between the actual possession difference and the possessions created difference.
Finally, I just want to say that I think the idea of points and possessions created at the player level are great ideas, but this method I think still has some work to be done on it.
NBA Points Created
