BGonline.org Forums

41S-54 at dmp

Posted By: Nack Ballard
Date: Friday, 28 May 2010, at 12:28 a.m.

In Response To: 41S-54 at dmp (Stick)

Basically, what Daniel said. CIs are based on a square route function.

If the CIs for the dmp rollouts, rounded, are 5k = +/- .004, 10k = +/- .003, and 31k = +/- .002, then the highest they could be (i.e., what XG thinks they are) is

\\5k is +/- .00450
10k is +/- .00318
31k is +/- .00184

...and the lowest they could be is

\\5k is +/- .00367
10k is +/- .00260
31k is +/- .00150

If you know the rounded CIs for other numbers of trials, then we can use those bits of information to deduce the CIs (to five or however many decimal places) within a tighter range. For example, if 1296 dmp trials is +/- .007 then we can use the above ranges we've already deduced to further deduce that 5k trials is between .00367 and .00375; or if 1296 trials is +/-.009 then 5k trials must be between .00425 and .00450.

That's not exactly true because not all rollouts of the same number of trials seem to always give the same CIs, but that's the general idea.

Now that computers and software have gotten so fast, I think it would be nice if CIs below some threshold (.005, or maybe even .01) were reported to a fourth decimal place, at least optionally. There is a meaningful difference between a fat .002 (which might be .0025) and a thin .002 (which might be .0015). Or you might see one play at .002 and another play at .001 (in the same rollout or in a different rollout) and you don't know if that's about what they really are or if they're both around .0015 or even if one is .0025 and the other .0005.

Nack

Messages In This Thread

• 41S-54 at four scores
  David Rockwell -- Thursday, 27 May 2010, at 12:36 p.m.
  • Attn Xavier
    Chase -- Thursday, 27 May 2010, at 3:36 p.m.
    • Attn Xavier
      Nack Ballard -- Thursday, 27 May 2010, at 4:33 p.m.
      • Rollout Seed part deux
        eXtreme Gammon -- Thursday, 27 May 2010, at 6:12 p.m.
        • Rollout Seed part deux
          Nack Ballard -- Thursday, 27 May 2010, at 6:30 p.m.
          • Rollout Seed part deux
            David Rockwell -- Thursday, 27 May 2010, at 6:53 p.m.
        • Rollout Seed part deux
          Chase -- Thursday, 27 May 2010, at 7:19 p.m.
          • Rollout Seed part deux
            Stick -- Thursday, 27 May 2010, at 7:24 p.m.
            • Rollout Seed part deux
              Chase -- Thursday, 27 May 2010, at 9:03 p.m.
          • Rollout Seed part deux
            eXtreme Gammon -- Friday, 28 May 2010, at 1:03 a.m.
  • 41S-54 at four scores
    Nack Ballard -- Thursday, 27 May 2010, at 5:32 p.m.
  • 41S-54 at dmp
    Stick -- Thursday, 27 May 2010, at 8:40 p.m.
    • 41S-54 at dmp
      Daniel Murphy -- Thursday, 27 May 2010, at 9:53 p.m.
    • 41S-54 at dmp
      Nack Ballard -- Friday, 28 May 2010, at 12:28 a.m.