Re: [Bug-gnubg] None

[Nis]

--On Thursday, April 17, 2003 12:05 -0400 Hanley Hayes 
<address@hidden> wrote:

> Either I stumbled upon a bug or (more likely) my understanding of the
> cube analysis is lacking.  Here is the situation:
>
> 1. I am an running a March 2 build on Windows NT
> 2. gnubg has already borne off two checkers
> 3. gnubg has one checker on the bar, trapped behind my 6-prime extending
> from my 2 pt through my seven point
> 4. The cube is at 2 and I own it.
> 5. I am on roll.

One important piece of info missing: Where are your last three checkers (*)

> Looking at this situation in analysis afterwards, I wanted to see when/if
> I should have doubled.  I clicked on the Cube decision tab.  Clearly I
> think I should double as a) I can't gammon gnubg, b) if I play on, I
> could possibly lose should gnubg start rolling doubles once I open up, c)
> gnubg will clearly pass on a cube at this moment,

Correct.

> d) the possibility that
> my situation will erode sufficiently such that I would both double gnubg
> and gnubg will accept, enabling me to win a game for 4 points I feel are
> infinitesimally small (though not zero)

If gnubg has a take, then you have lost equity by waiting (or gnubg is 
making an error). So the
combination "You both play perfectly, and the cube gets to 4" is not 
possible.

> However, the Cube analysis reports:

Important information missing: Analysis settings

> 1. Double pass +1.000
> 2. Double take +1.460 +0.460
> 3. No double   +1.000 -0.000
>
> Proper cube action: Redouble, pass

The conclusion agrees with you! Well done.

> My question is:  Given that there is still contact possible, why is
> option 3 (No double) given a +1.000.

The answer to this question must be: You can always double next turn, and 
gnubg should drop.

--
Nis Jorgensen
Greenpeace
Amsterdam