BGonline.org Forums

Why we need actual statistics: A worked example

Posted By: Timothy Chow
Date: Sunday, 11 September 2011, at 9:16 p.m.

In Response To: Why we need actual statistics: A worked example (Daniel Murphy)

Daniel Murphy wrote:

whether using actual statistics means more time-consuming rollouts

Yes. This is something I mentioned at the end of my "concrete proposal."

It's not inconceivable that someone could figure out a clever way to do the equivalent of variance reduction for actual statistics, but I don't see offhand how to do it.

One small thing that could be done is to update the actual statistics in an appropriate way when one reaches the exact bearoff database, rather than forcing the bot to play it out to the bitter end. For most positions this won't have that big an effect, though, I think.

Messages In This Thread

• Why we need actual statistics: A worked example
  Timothy Chow -- Sunday, 11 September 2011, at 2:22 a.m.
  • Minor correction
    Timothy Chow -- Sunday, 11 September 2011, at 2:47 a.m.
  • Why we need actual statistics: A worked example
    Daniel Murphy -- Sunday, 11 September 2011, at 3:17 a.m.
    • Why we need actual statistics: A worked example
      Timothy Chow -- Sunday, 11 September 2011, at 5:22 p.m.
      • Why we need actual statistics: A worked example
        Daniel Murphy -- Sunday, 11 September 2011, at 5:41 p.m.
        • Why we need actual statistics: A worked example
          Timothy Chow -- Sunday, 11 September 2011, at 9:16 p.m.
          • Why we need actual statistics: A worked example
            Maik Stiebler -- Sunday, 11 September 2011, at 10:23 p.m.