I've always wondered if you could used boxscore statistics to qualify a player's position. It's a meaningless problem to solve in this era of positionless basketball, but it's always interested me. Recently, I acquired access to a robust but easy to use statistical package (Spotfire), and my time spent trying to figure out how to use ggplot in R has dropped significantly. I threw together a small script to scrape some NBA.com data and threw it into Spotfire to see what I could come up with.
I was playing with treemaps to see what I could see, when suddenly I had apparently created a chart that defined positions. Granted, this was not basic boxscore data, it was SportsVu data, but my mind was happy with what I saw. Apparently, average time per touch is a decent position identifier. Check it out.
Comments
In previous posts I've mentioned that part of my motivation for the work being done here is to learn. Learn about basketball, learn about data acquisition, analysis, and the assorted tools that can or should be used to do the work. Insofar, I have solely used Microsoft Excel and VBA to perform work that I have done. Its what I know best and by far has the lowest knowledge entry point. What you'll see in this post is charts created in an open source statistical package called R. R is the premium statistical analysis package in use today for one paramount reason, it's free. I've done work in SPSS and Spotfire before. They are both far easier in terms of usability, but don't hold a candle to the accessibility to your run of the mill NBA fan looking to do work completely over their head.
The Thunder defense has been terrible this year. Everyone knows this. 104.9 DRtg (bottom half of the league) 8th worst in the league in opponents FTA/FGA. It has been ugly and lazy (looking at you Russ).
Let's look at some numbers.
|
AuthorArchives
November 2017
Categories
All
|