Re: [Bug-gnubg] Wrong move in external interface

[Jim Segrave]

On Wed 16 Mar 2005 (07:54 +0100), amarganth wrote:
> Hi all
> 
> I compared both "boards": the one of Joachim, the other from Jim:
> 
> board:you:her:3:2:1:0:1:2:0:1:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:-4:0:1:1
> :2:0:0:1:1:1:0:1:-1:0:25:11:11:0:0:2:2:0:0
> board:you:her:3:2:1:0:1:2:0:1:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:-4:0:1:1
> :2:0:0:1:0:1:0:1:-1:0:25:11:11:0:0:2:2:0:0
> 
> The difference is field number 38 (beginning by 0), which is "1" in the
> first, "0" in the second board structure.
> Field 38 is "may double" of player 0. The "was doubled" flag is number 40.

Yes - you are right - when parsing the sscanf's in drawboard.c, I got
out of step.

My variant command causes the cube ownership to be set rather than
centred.

> I cannot understand, what gives back another move, if that "may double" flag
> is other. It's not a doubling situation. And it's crawford (1-away, 2-away).

And that's the clue - it's not Crawford it's post Crawford.

Try simply setting up the match up as follows in gnubg:

set dice 2 1
set board DwAAoAkAAAAAAA
set score 1 2
set player 1 gnubg
set player 0 human
play
analyse match

Which makes it non-Crawford, but otherwise the same as the original
posting. Gnubg moves 4/1 for me. And the analysis of gnubg's move gives
the following strange numbers

    1. Cubeful 2-ply    4/1                          Eq.:  -1.00000
       0.11574 0.00000 0.00000 - 0.88426 0.00000 0.00000
        2-ply cubeful prune [world class]
    2. Cubeful 2-ply    4/2 1/off                    Eq.:  -1.00000 (+0.00000)
       0.11574 0.00000 0.00000 - 0.88426 0.00000 0.00000
        2-ply cubeful prune [world class]
    3. Cubeful 2-ply    4/3 2/off                    Eq.:  -1.00000 (+0.00000)
       0.11574 0.00000 0.00000 - 0.88426 0.00000 0.00000
        2-ply cubeful prune [world class]
    4. Cubeful 2-ply    2/1 2/off                    Eq.:  -1.00000 (+0.00000)
       0.11574 0.00000 0.00000 - 0.88426 0.00000 0.00000
        2-ply cubeful prune [world class]
    5. Cubeful 0-ply    2/off 1/off                  Eq.:  -4.94842 (-3.94842)
       0.53239 0.00000 0.00000 - 0.46761 0.00000 0.00000
        0-ply cubeful prune [expert]

Then do:

set dice 2 1
set board DwAAoAkAAAAAAA
set score 1 2
set player 1 gnubg
set player 0 human
set crawford on
play
analyse match

Now gnubg does something sensible and plays 2/off 1/off

    1. Cubeful 0-ply    2/off 1/off                  Eq.:  +0.06477
       0.53239 0.00000 0.00000 - 0.46761 0.00000 0.00000
        0-ply cubeful prune [expert]
    2. Cubeful 0-ply    4/1                          Eq.:  -0.76849 (-0.83326)
       0.11575 0.00000 0.00000 - 0.88425 0.00000 0.00000
        0-ply cubeful prune [expert]
    3. Cubeful 0-ply    4/2 1/off                    Eq.:  -0.76849 (-0.83326)
       0.11575 0.00000 0.00000 - 0.88425 0.00000 0.00000
        0-ply cubeful prune [expert]
    4. Cubeful 0-ply    4/3 2/off                    Eq.:  -0.76849 (-0.83326)
       0.11575 0.00000 0.00000 - 0.88425 0.00000 0.00000
        0-ply cubeful prune [expert]
    5. Cubeful 0-ply    2/1 2/off                    Eq.:  -0.76849 (-0.83326)
       0.11575 0.00000 0.00000 - 0.88425 0.00000 0.00000
        0-ply cubeful prune [expert]

The problem isn't in external.c or the FIBs processing - it's doing a
very strange move selection when the cube is live.

(BTW, when setting player 1 to gnubg above, it deosen't help having it
set to Supremo chequer/World Class cube)

Another anomoly - with the position set up as shown - try a zero ply
evaluation:

    1. Cubeful 0-ply    2/off 1/off                  Eq.:  +5.07797
       0.53239 0.00000 0.00000 - 0.46761 0.00000 0.00000
        0-ply cubeful prune [expert]
    2. Cubeful 0-ply    4/1                          Eq.:  -0.76849 (-5.84646)
       0.11575 0.00000 0.00000 - 0.88425 0.00000 0.00000
        0-ply cubeful prune [expert]
    3. Cubeful 0-ply    4/2 1/off                    Eq.:  -0.76849 (-5.84646)
       0.11575 0.00000 0.00000 - 0.88425 0.00000 0.00000
        0-ply cubeful prune [expert]
    4. Cubeful 0-ply    4/3 2/off                    Eq.:  -0.76849 (-5.84646)
       0.11575 0.00000 0.00000 - 0.88425 0.00000 0.00000
        0-ply cubeful prune [expert]
    5. Cubeful 0-ply    2/1 2/off                    Eq.:  -0.76849 (-5.84646)
       0.11575 0.00000 0.00000 - 0.88425 0.00000 0.00000
        0-ply cubeful prune [expert]

An equity of +5.07797 seems slightly on the high side to me

-- 
Jim Segrave           address@hidden