RealGM

Most people know that the majority of the league decreases efficiency when they get the ball more (usage), because the defense pays more attention to them, they have to stay focused for longer, ect, ect. There are some obvious cases like when 1 player shoots 2.5 FGA and has a 65 TS%. He would most likely not be as good of a scorer than a player puts up 65 TS% while putting up 15 shots a game. The second player's a lot more legit, and it's clear the 2nd guy is a better scorer than the previous player.

What I'm looking to do is separate the smaller cases. Is a guy like Manu a better scorer (both production and efficiency) than a guy like Westbrook? Manu posted 12.9 PPG on 67 TS%, 22.7 USG%. Westbrook put up 23.6 PPG on 54 TS%, 32.7 USG%. The differences are obvious, Manu has a lesser output, lesser load, but at much higher efficiency, whereas Westbrook sees the ball more, puts up more shots and more points, but is less efficient in doing so. So who's the better scorer? Is Westbrook's downgrade in efficiency justified correctly by how much more usage he gets?

I'm guessing the answer lies in taking a huge sample of players and their usage and TS%, averaging them out and finding a normal curve and standard deviation, and seeing which one is more standard deviations from the center at whatever output they produce. However, I'm not exactly sure how to get the actual number that would be the center.

Has anyone done anything like this, or can anyone help to figure out how to come up with something like this? I think it would be very useful.

Before going through any of that, my concern is separating scoring from creating. They go together, and I've never understood the obsession with scoring only (well, once one starts to follow the game closely), so in this case you'd be doing a lot of statistical manipulation to try and control for something and still be left with a giant, giant confound (which is part of the overall important package of "offensive player").

From a pure efficiency to volume standpoint, my statistic (EV) does it automatically. It also incorporates TOV's, which I never understand why people leave out when analyze offensive numbers. More importantly, it incorporates how much a player creates offense and what player's *benefit* from the creation of offense. Until we have a good grasp on that, we can't really tackle this question well because of the giant differences in context.

In general, you're trying to tackle a hard problem from a strange (potentially useless) angle.

Neil took a crack at balancing efficiency/usage a couple years ago with a "points produced per team possession used" metric:

http://www.basketball-reference.com/blog/?p=8359