web analytics

Visualization: 2012 VORP Treemap

January 27, 2012

Inspired by Evan Zamir’s use of a treemap to visualize WARP over at The City, I created a treemap to visualize ASPM VORP for the 2012 season-to-date.  I used Google Docs and the treemap-gviz widget to construct the visualization.  I used the latest iteration of the ASPM framework to compile the data, and selected for display only those players who had played a significant amount of minutes (technically, based on regressed MPG).

The size of each tile is based on the player’s Value over Replacement Player, a total-value-to-the-team measure.  The color of each tile is based on each player’s Advanced Statistical Plus/Minus, a per-minute rate.  Really bad players (below -2.5 ASPM, replacement level) will not be on the chart at all–Charlotte has a number of these, and barely any over replacement level.


  • Manu was dominant when he was playing–when he returns, the Spurs will likely be among the league’s elite once again.
  • Who is the big 2 in New York again?
  • Ryan Anderson: good player
  • Carrying otherwise poor teams: Millsap, Lowry, Jennings
  • Philly, the Spurs, and Indy are very balanced

Click on the tiles to see who the player is and what his VORP and ASPM numbers are.

Tags: , , , ,

2 Responses to Visualization: 2012 VORP Treemap

  1. JJ on February 12, 2012 at 7:33 pm

    Hey, I’m a non-math person that’s into NBA stats. I’ve been reading some of Alex’s analysis, which indicates that this stat tends to be pretty good at both predicting and explaining. I downloaded the spreadsheet that goes back to 78, but I’m confused as to how you have the +/- from back then–do you use RAPM to find the values of box score events and then simply use those weights for games that predate plus minus? If so,how does ASPM address defense that the box score can’t account for? It would be great to get a methods article for your particular variant of ASPM. The treemap’s great!

    • DanielM on February 14, 2012 at 8:56 am

      ASPM is purely a box-score stat. I used 8 years of RAPM to develop the weights, and then calculated ASPM back to 1978. Likely, the accuracy won’t be as great at that distance from the regression.

      Yes, there is a significant lack of accuracy due to limited box-score data for defense. The r^2 is much higher for offense than defense when attempting to predict RAPM.

Leave a Reply

Your email address will not be published. Required fields are marked *

Current day month ye@r *

DSMok1 on Twitter

To-Do List

  1. Salary and contract value discussions and charts
  2. Multi-year APM/RAPM with aging incorporated
  3. Revise ASPM based on multi-year RAPM with aging
  4. ASPM within-year stability/cross validation
  5. Historical ASPM Tableau visualizations
  6. Create Excel VBA recursive web scraping tutorial
  7. Comparison of residual exponents for rankings
  8. Comparison of various "value metrics" ability to "explain" wins
  9. Publication of spreadsheets used
  10. Work on using Bayesian priors in Adjusted +/-
  11. Work on K-Means clustering for player categorization
  12. Learn ridge regression
  13. Temporally locally-weighted rankings
  14. WOWY as validation of replacement level
  15. Revise ASPM with latest RAPM data
  16. Conversion of ASPM to" wins"
  17. Lineup Bayesian APM
  18. Lineup RAPM
  19. Learn SQL