Books on Designing Metrics...

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • bookiebust
    SBR Rookie
    • 07-11-10
    • 8

    #1
    Books on Designing Metrics...
    Hello, I have been looking at some of the methods that have been suggested on here for designing metrics to create quantitative data from existing raw data that can be used as modeling data, suitable for, say, a neural network. However, I am not just interested in this from the perspective of designing a good neural network.

    I am not exactly sure how to phrase what I'm looking for. Perhaps I need to become more familiar with the topic and come back asking questions later. However, like I said, I believe what I'm looking for is a complete reference or references to creating metrics and transforming data (putting data in alternate formats useful in sizing up performance).

    Looking for a book on metrics is a fruitless venture. It seems no one is 100% clear on metrics these days.

    Sabermetrics develops many supplemental ways of measuring baseball performance. What are the mathematical tricks used to create these metrics? And moreover, is there a book that covers them?
  • Flying Dutchman
    SBR MVP
    • 05-17-09
    • 2467

    #2
    Bookie, you are on a long road. First you need to build a solid db or neural will sub-optimize and it needs to be long enough or it will overfit. On metrics you need to experiment, you can use the usual suspects, but then everyone uses them who models. BTW, neural is an unusually slow method that many times does not work.

    PM me if you want more...

    Comment
    • roasthawg
      SBR MVP
      • 11-09-07
      • 2990

      #3
      I'd start off with good old fashioned linear regressions if I were you.
      Comment
      • bookiebust
        SBR Rookie
        • 07-11-10
        • 8

        #4
        Originally posted by roasthawg
        I'd start off with good old fashioned linear regressions if I were you.
        Well, I do have a degree in the sciences, but I don't have much experience using modeling in any truly rewarding way, such as perhaps with handicapping. I think I can jump onto more complex techniques faster.

        --
        Comment
        • Maverick22
          SBR Wise Guy
          • 04-10-10
          • 807

          #5
          Which "sciences"? Natural sciences? Social sciences? Computer Sciences?
          Comment
          SBR Contests
          Collapse
          Top-Rated US Sportsbooks
          Collapse
          Working...