Comments on: Diminishing Returns for Scoring - Usage vs. Efficiency

By: Question

Question — Mon, 02 Apr 2012 23:16:07 +0000

Hello, I don’t understand the way you are interpretating the r squared value. This term means that the amount of varion explained by the predictor. Based on your r squared terms, this shows there is basically no correlation between the predictor and response. Correct me of I’m wrong, but I think you are mistaking the r squared term with the p value

By: shiceasia

shiceasia — Sun, 11 Oct 2009 09:37:47 +0000

Hello, it really interesting, thanks

By: Mountain

Mountain — Sat, 08 Mar 2008 06:11:23 +0000

To tie in better with what you’ve done (or at least try) I shoud say (take two)…
The coefficent of the study with all lineups was bigger than for just 50+ possessions.
But what does the chart for all lineups look like and compared to the 50+ possesion chart you presented? How do sections of the curves compare? Does team level detail show the greater volatility of small possession lineups fairly evenly across the league or do good teams / coaches do much better with these 5 man lineups than weak teams / coaches? Is it easier for teams to do ok with their standard lineups and harder to separate yourself from the rest with them or not? Is the war won more in the complicated details of the rotation game or in the main matchups?

By: Mountain

Mountain — Sat, 08 Mar 2008 05:58:31 +0000

Thanks for the research and sharing it to move the debate and understanding forward.

The results make sense to me. High usage lineups maintain or get a small boost. Low usage lineups see a larger decline that accelerates as historical usage declines.

The higher usage 5 man lineups tend to have more players with high usage obviously but it might be good to quantify and frame this. How does the rate of presence of players over a given usage level(s) change across the curve of all 5 man lineups? Is that changing more or less or roughly equally to change in total usage of the 5 man lineups?

Perhaps the data could be broken out for top usage guy and rest of team or top 2 and rest of team? The more detail, the more chance to anchor interpretations about what it going on in support from the data.

I also took interest in the the team average # of different 5 man lineups- about 270 and the number that played 50+ possessions about 18.5. How much total time does the later group have compared to the group below 50 possessions? I understand how the vast array of lineups occurs but I want to know more about the effects of this. Is the efficency of lineups played with 50+ possessions better than that of lineups with fewer than 50 at league level? What does the data look like at team level? Who is best and worst at using “standard ” lineups and alternatively “transition” or “situational” or “random” lineups?

Look forward to hearing what you can say about these things.

By: Eli

Eli — Fri, 07 Mar 2008 20:54:37 +0000

See here for more discussion relating to this post:

http://sonicscentral.com/apbrmetrics/viewtopic.php?t=1679

By: Eli

Eli — Fri, 07 Mar 2008 14:15:21 +0000

Thanks, Ed. I don’t claim too much originality though. I’m just building off of ideas from you, Ben F., Cherokee_ACB, and Guy (and Dean Oliver of course).

I agree that the BasketballValue lineup data is a gold mine that’s only beginning to be explored.

By: edk

edk — Fri, 07 Mar 2008 06:03:21 +0000

Once again, impressive work. I would’ve never thought to use a lineup-based approach in a million years. Regardless of your results here (or even the rebounding thing you did before), that method is going to be your enduring contribution to the field. Thanks to Aaron at BasketballValue for providing the data (criminally underused, in my opinion) and to you for going through all this work.