BGonline.org Forums

Rockwell/Kazaross MET available for XG users

Posted By: Timothy Chow
Date: Saturday, 27 February 2010, at 9:20 p.m.

In Response To: Rockwell/Kazaross MET available for XG users (David Rockwell)

Thank you very much, David and Neil, for performing this valuable service to the BG community.

David Rockwell wrote:

The key parameter in the table construction is the MWC (match winning chances) calculated for each rollout. I don't understand how GNU calculates this number very well, but we used it anyway. An improvement would be to capture all possible results (single point drops, gammon wins, double out on a four cube, etc. etc.) and apply each to the appropriate MET points. What we did is supposed to accomplish that. However, if approximations are involved, as I suspect there are, then there is room for improvement.

You can get the numbers I described from GNU. And, I experimented with it. But, you can't store them anywhere and there are other practical problems. Furthermore, they are compiled without variance reduction. So the number of rollouts needed would increase to some prohibitive amount.

If I understand you correctly, you are implicitly asking for the same thing that I've been beating the drum for, which is for GNU to provide more support for its "View Statistics" feature.

Although I certainly agree with this, I don't believe that it would actually have any effect as far as your project is concerned. I haven't actually checked the source code, but I'm pretty sure that the MWC that GNU produces is exactly what you're hoping for—namely, an average of all the actual match results, with variance reduction applied. That is, it does not, despite appearances, take the cubeless figures for the rollout and apply Janowski's formula. You can partially verify this yourself by turning variance reduction off, computing the MWC from the "View Statistics" numbers by hand, and checking that it matches the cubeful rollout equity.

Perhaps someone from the GNU team can confirm this.

Messages In This Thread

• Rockwell/Kazaross MET available for XG users
  neilkaz -- Friday, 26 February 2010, at 3:00 a.m.
  • Will be available for GNU soon - nm
    neilkaz -- Friday, 26 February 2010, at 3:11 a.m.
    • Will be available for GNU soon - nm
      mfic -- Saturday, 27 February 2010, at 6:38 p.m.
      • Thank you Neil
        David Rockwell -- Saturday, 27 February 2010, at 6:52 p.m.
        • My pleasure
          neilkaz -- Saturday, 27 February 2010, at 7:39 p.m.
      • DoubleTakeCalculator by David Hart
        Mislav Radica -- Saturday, 27 February 2010, at 7:40 p.m.
  • Rockwell/Kazaross MET available for XG users
    Timothy Chow -- Saturday, 27 February 2010, at 3:14 a.m.
  • Rockwell/Kazaross MET available for XG users
    Phil Simborg -- Saturday, 27 February 2010, at 1:31 p.m.
    • Rockwell/Kazaross MET available for XG users
      Matt Cohn-Geier -- Saturday, 27 February 2010, at 1:57 p.m.
    • Rockwell/Kazaross MET available for XG users
      David Rockwell -- Saturday, 27 February 2010, at 6:46 p.m.
  • Rockwell/Kazaross MET available for XG users
    Phil Simborg -- Saturday, 27 February 2010, at 1:51 p.m.
    • Rockwell/Kazaross MET available for XG users
      neilkaz -- Saturday, 27 February 2010, at 4:46 p.m.
      • Rockwell/Kazaross MET available for XG users
        David Rockwell -- Saturday, 27 February 2010, at 7:21 p.m.
        • Rockwell/Kazaross MET available for XG users
          Timothy Chow -- Saturday, 27 February 2010, at 9:20 p.m.
  • Rockwell/Kazaross MET available for XG users
    Chase -- Saturday, 27 February 2010, at 10:33 p.m.
  • Rockwell/Kazaross MET now available for GNU users
    neilkaz -- Monday, 1 March 2010, at 1:46 a.m.