The Usage argument: a thread

Quick Note all: I’m taking a lesson from Twitter. We added a 1K wordcount limit to comments. Much like the NBA salary cap, there are ways around this. As we did say we wanted a dialog, I am actually going to be a bit more strict that normal. I hope 1K is enough words for you to make your point. If not, take a break, relax and come back to the thread after someone else has had a turn.

Hey all,

We really don’t have a form. Vivek and my recent podcast got a decent amount of comments going on the usage argument. Here’s a bit of background.

In Basketball on Paper, Dean Oliver brought up the concept of the Skill Curve. The idea is that as a player uses more of the offensive possessions, their efficiency will drop. Additionally, as they use less, their efficiency will improve but only to a point (e.g. Allen Iverson taking three shots as opposed to two would not be a marked improvement in his shooting)…or that’s the theory.

There are a few important considerations to take. Oliver admitted that since the data was sparse (it’s not as if players take a wide range of shots over a season for you to compare), he “estimated” in some ranges (in the book he points out Jerry Stackhouse and Allen Iverson refused to take less than 20% of the team’s shots). Additionally, he does not divulge how he makes the curves. Here is the full quote from Oliver (as quoted in Stumbling on Wins)

Skill curves are neat to look at, but you probably don’t want to know the details about how to make them.  Generally, they come from looking at box scores and general trends that players show when using a lot of possessions. How much better do players seem to get if they use fewer possessions? If they seem to get worse when they use fewer possessions, I say, ‘That’s not right.’ It’s just not a sustainable trend. Effectively, I force the curves to be declining.  The details of how I do that would scare someone who doesn’t know formal statistics.  They would scare someone who did know formal statistics for different reasons (p. 239, italics added to original).

The relationship is forced and the details are removed (for a reason I take as a cop out frankly) We should also note he breaks down and spams “All work and no play make Jack a dull boy” in the same chapter. Ian Levy took a crack at skill curves a while back for some members of the Pacers. Brandon Rush had a nice looking skill curve, the more he shot the worse he got. A.J Price was the opposite, more usage meant better production! Josh McRoberts was all over the place. He’d get better then worse, then better then worse! Dahntay Jones had a curve like an age curve. He got better and better and then slowed down and then declined rapidly! Looking at usage vs. efficiency, we see players can be all over the place. In essence, Skill Curves are a nice idea but data doesn’t seem to back them.

I’m going to end with two points. First, we are looking into this more and if we find interesting stuff, you’ll be the first to know! Second, we’ll let you discuss this to your hearts’ content. A few of you asked for an open dialogue. This comes with a warning though. A dialogue implies back and forth and new information. If this turns into a shouting match with no new information, we’ll shut it down.



Comments are closed.