4 Comments

  1. J

    “This is more useful for predicting the team future performance than quantifying the impact of individual players.”
    Did you actually test this or is it just an assumption? How does the weighing work, exactly?

    • DanielM

      The weighting is the same I used to maximize out-of-sample prediction when I constructed my league-wide Bayesian rating system, here: On Bayesian Predictive Efficiency Ratings. Example: Most recent game weighted 1. 25 games ago, weighted 0.50. 50 games ago, weighted 0.30. 75 games ago, weighted 0.20. Etc.

      Because the more recent performance is weighted more highly, but the regression to the mean portion of the equation still remains the same, players who had most of their time off court more recently vs. more distantly would probably not be handled correctly. Do you have any more insight on this?

  2. […] DStats, Daniel M looks at the WOWY metrics for OKC’s bigs. Link Posted on Wednesday, April 13th, 2011 at 5:34 pm, Category: Basketball, Tags: dstats, nba, […]

Leave a Reply

Your email address will not be published.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