Re: [Bug-gnubg] RE: Monte Carlo

[Joseph Heled]

That's one bug out of the way. I submitted a fix now.

Albert Silver wrote:

> Here I think there may be some bug, possibly some residue from the
> cache. I find that switching back and forth between METs, sometimes
> running a 3-ply, and then switching, etc, leads to discrepancies in the
> results reported at 2-ply. 
>
> I have seen some really big changes, but am not sure the procedure I
> used to produce them. 
>
>                                                 Albert

Now results are not sane more or less. Still no idea which MET was used 
to generate the evaluation in the sgf file. I wonder what you are going 
to do. If you decide to rerun all evaluations, you will have to run over 
the annotations again.

set matchequitytable "met/zadeh.xml"

Cube analysis
2-ply cubeless equity  +0.658 (Money:  +0.653)
   74.0%  18.9%   1.1% -  26.0%   2.6%   0.0%
Cubeful equities:
1. Double, pass         +1.000
2. Double, take         +1.112  ( +0.112)
3. No double            +0.919  ( -0.081)
Proper cube action: Double, pass

set matchequitytable "met/snowie.xml"

Cube analysis
2-ply cubeless equity  +0.654 (Money:  +0.651)
   74.1%  18.5%   1.0% -  25.9%   2.6%   0.0%
Cubeful equities:
1. Double, pass         +1.000
2. Double, take         +1.084  ( +0.084)
3. No double            +0.914  ( -0.086)
Proper cube action: Double, pass

set matchequitytable "met/jacobs.xml"

Cube analysis
2-ply cubeless equity  +0.657 (Money:  +0.653)
   74.0%  18.9%   1.1% -  26.0%   2.6%   0.0%
Cubeful equities:
1. Double, pass         +1.000
2. Double, take         +1.105  ( +0.105)
3. No double            +0.917  ( -0.083)
Proper cube action: Double, pass