[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-gnubg] Strange analysis output
From: |
Joern Thyssen |
Subject: |
Re: [Bug-gnubg] Strange analysis output |
Date: |
Thu, 5 Sep 2002 20:00:23 +0000 |
User-agent: |
Mutt/1.4i |
On Thu, Sep 05, 2002 at 11:03:17AM +0200, Jim Segrave wrote
> On Thu 05 Sep 2002 (10:12 +0200), address@hidden wrote:
> >
> >
> > --On 04 Sep 2002 19:24 +0200 Jim Segrave <address@hidden> wrote:
> >
> > >>From the Annotation window:
> > >
> > > 1. Cubeful 2-ply 5/4 5/1 Eq.: +1.033
> > > 0.728 0.371 0.009 - 0.272 0.027 0.000
> > > 2-ply cubeful 100% speed, 8 cand., 0.16 tol. [world class]
> > >* 2. Cubeful 2-ply 8/7 8/4 Eq.: +0.950 (
> > > -0.083) 0.730 0.413 0.016 - 0.270 0.027 0.000
> > > 2-ply cubeful 100% speed, 8 cand., 0.16 tol. [world class]
> > >
> > >The move I made 8/7 8/4 has a
> > >higher probability of a win 0.730 vs 0.728
> > >higher probability of a gammon 0.413 vs 0.371
> > >higher probability of a backgammon 0.016 vs 0.009
> > >and the
> > >same probabilities of being gammoned or backgammoned. So why is the
> > >equity lower? What am I missing or misunderstanding?
> >
> > The easy answer:because you are doing cubeful evaluation. The numbers
> > listed are the cubeless winning chances.
>
> Doh. I knew there was a simple explanation. Dank u wel.
>
> > Understanding WHY the cubeful equity is so much different in a situation
> > loke this of course another matter.
>
> Hmm - I'm trying to work that out for myself now. It may be some time.
I can't offer any explanation beyond Nis'. I can confirm the 2-ply
evaluation by doing a 1296 trial rollout without variance reduction
truncated after 2-ply. So the rollout code and eval code is consistent.
One explanation could be volatility: after hitting, blue has a
relatively good position, 5/4 5/1 often leads to "double, pass" unless
blue rolls something horribly bad (i.e., 66 or 44).
Jørn
--
Joern Thyssen, PhD
Vendsysselgade 3, 3., DK-9000 Aalborg, Denmark
+45 9813 2791 (private) / +45 2077 2689 (mobile) / +45 9633 7036 (work)