Re: [Bug-gnubg] Bug: Double/Beaver

[Christopher D. Yep]

At 07:00 AM 8/21/2003 +0000, Jørn Thyssen wrote:
> On Wed, Aug 20, 2003 at 02:14:22AM -0400, Christopher D. Yep wrote
> > Windows XP, GUI version
> > Version 0.14-devel 1.1282 030818 (build Aug 19 2003)
> >
> > gnubg misinterprets the following evaluation:
> >
> >     GNU Backgammon  Position ID: 27ZCFADbtgEDQA
> >                     Match ID   : cAkAAAAAAAAA
> >     +24-23-22-21-20-19------18-17-16-15-14-13-+  O: gnubg
> >     | O  O  O  O  O  O |   | O           O  X |  0 points
> >     | O  O  O  O  O    |   |                X |
> >     |                  |   |                  |
> >     |                  |   |                  |
> >     |                  |   |                  |
> >     |                  |BAR|                  |v (Cube: 1)
> >     |                  |   |                  |
> >     |                  |   |                  |
> >     |                  |   |                  |
> >     | X  X  X  X  X  X |   |                  |  On roll
> >     | X  X  X  X  X  X | X |          O  O    |  0 points
> >     +-1--2--3--4--5--6-------7--8--9-10-11-12-+  X: Chris
> >
> > Cube analysis
> > 2-ply cubeless equity  -0.038
> >    38.3%  28.2%   6.2% -  61.7%  14.7%   0.0%
> > Cubeful equities:
> > 1. Double, take         -0.236
> > 2. Double, pass         +1.000  ( +1.236)
> > 3. No double            -0.389  ( -0.153)
> > Proper cube action: Double, beaver
> >
> > [Jacoby rule used above]
> > Correct cube action, based on the above numbers is actually
> > no-double/beaver.  Double/beaver is worth -.472 to X which is worse than
> > the -.389 he can get from holding the cube.
>
> No, as gnubg writes the value of double, take is -.236. The value
> reported does include the value of the cube! So the logic is correct:
> this is a double, beaver.
>
> Here is a plain evaluation of the position after the double:
>
>         Win     W(g)    W(bg)   L(g)    L(bg)   Equity    Cubeful
>  2 ply: 0.383   0.286   0.071   0.149   0.000    -0.027    -0.107
>
> My version of gnubg running under windows is quite old so you don't see
> the expected value of -0.236/2 = -0.118.
>
> Can you set up the position after the double and run Analysis->Evaluate?
> You should see a cubeful equity (normed to 1) of -0.118. Hence the value
> of double, take before the double is -0.236.

Actually I think the -0.236 does not include the value of the 
beaver.  Although I didn't mention it in my original note I also tested 
this without beavers (set beavers 0), with exactly the same equities reported.

Btw, "help set beavers" says "Usage: set beavers on|off".  Should this be 
changed to "Usage: set beavers [number]" or something similar?

Here is the output from Analyse - Evaluate:

        Win     W(g)    W(bg)   L(g)    L(bg)   Equity    Cubeful
static:  42.0%   34.2%    7.7%   12.8%    0.1%   -0.159    +0.032
 1 ply:  42.6%   28.6%    6.2%   11.1%    0.1%   -0.148    -0.396
 2 ply:  38.3%   28.2%    6.2%   14.7%    0.0%   -0.235    -0.389

Cube analysis
2-ply cubeless equity  -0.038
   38.3%  28.2%   6.2% -  61.7%  14.7%   0.0%
Cubeful equities:
1. Double, take         -0.236
2. Double, pass         +1.000  ( +1.236)
3. No double            -0.389  ( -0.153)
Proper cube action: Double, beaver

So actually I don't see -0.118, but instead see -0.235 (not -0.236)!  Is 
this round-off error?

I get the exact same results if I set beavers 0.

Chris