[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-gnubg] never too good?
From: |
Joseph Heled |
Subject: |
Re: [Bug-gnubg] never too good? |
Date: |
Sat, 03 Aug 2002 08:08:09 +1200 |
Joern Thyssen wrote:
>
> On Fri, Aug 02, 2002 at 07:20:51AM +0000, Joern Thyssen wrote
> > On Fri, Aug 02, 2002 at 07:07:00PM +1200, Joseph Heled wrote
> > > GNU Backgammon Position ID: BgAC2HZ7AAAEAA
> > > Match ID : MAGgAAAAAAAA
> [snip]
> > > My cubefull rollout gave
> > > ND 59.428%
> > > D/T 59.754%
> > >
> > > With D/D 57.5%, this is strange. Without gammons/bg's, ND should be <=
> > > D/D. I think this means X is making a great number of bad takes, or
> > > in other words, this position (and others like it) are a good source
> > > for finding positions where GNUbg makes cube errors.
> >
> > I agree.
> >
> > When I get home, I'll try doing a rollout with gnubg and see if I can
> > reproduce the result, so eliminate possible bugs.
>
> With a 1296 games 0-ply rollout (no truncation, with var.redn.) I get:
>
> ND 59.96%
> DT 60.09%
> DP 57.76%
>
> --> too good to double, pass.
which we agreed can't be the right one. My code, BTW, has been checking this for
a long time now -
(line 724 in analyze/danalyze.cc)
if( probs[1] == 0.0 || /* no gammons */
matchProbDoubleDrop > matchProbNoDouble ) {
actionDouble = true;
} else {
actionDouble = false;
tooGood = true;
}
>
> Jørn
>
> --
> Joern Thyssen, PhD
> Vendsysselgade 3, 3., DK-9000 Aalborg, Denmark
> +45 9813 2791 (private) / +45 2077 2689 (mobile) / +45 9633 7036 (work)
>
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/bug-gnubg