Archetypal Ballers and Ternary Plots - Visualizing NBA Skills for Fun and Profit

*
Proposal
Short Form
Intermediate

Excerpt

Basketball is second only to baseball in its rich variety of detailed data and analysis techniques. This project uses two techniques to reduce this complexity. The first, archetypal analysis, is an unsupervised learning technique that reduces the 18-dimensional box scores to a three-dimensional vector. The second, ternary plots, provides an elegant visualization for comparing players and teams. Using these techniques, I'll review the 2016 - 2017 NBA season.

Description

Basketball is second only to baseball in its rich variety of detailed data and analysis techniques. From this richness, though, comes complexity. For example, the conventional box score line for a player in a game has 18 variables!

This project uses two techniques to reduce this complexity. The first, archetypal analysis, is an unsupervised learning technique that reduces the box scores to a three-dimensional vector, rating a player by

  • their front-court skill relative to the best front-court player in the league,
  • their back-court skill relative to the best back-court player in the league, and
  • their lack of overall skill, relative to the worst player in the league.

The second technique is visualizing these vectors on a ternary plot. A ternary plot maps these three-dimensional vectors onto a triangle, with the best front-court player at one apex, the best back-court player at another, and the worst overall player at the third.

Using these techniques, I’ll review the 2016 – 2017 NBA season. You’ll learn who the best players and teams were and the impact of some highly-publicized trades. #RipCity

Tags

basketball analytics, NBA, ternary plots, archetypal analysis, data visualization

Speaking experience

I've been speaking since I was a child. ;-) But seriously, I used to give sales presentations and demos for high-performance computing, I was on Jeopardy! and I was in 25 plays when I was a graduate student - in applied mathematics.

Oh, yeah - and I was a speaker last year at Open Source Bridge: http://opensourcebridge.org/sessions/1786

Speaker