Re: [Bug-gnubg] strange evaluation

[Christian Anthon]

This is a known weakness in the eval engine. Bearing in against many
checkers isn't handled too well.

Christian.

On Wed, Jul 1, 2009 at 6:00 AM, Masanori Itikawa<address@hidden> wrote:
> Hi.all.
> I find gnubg make strange evaluation in some positions.
>  GNU Backgammon  Position ID: YAMHwHu2bR0EAA
>                  Match ID   : cInsABAAKAAA
>  +13-14-15-16-17-18------19-20-21-22-23-24-+     O: Staf1
>  | X              O | O | O                |     1 point
>  |                O | O | O                |
>  |                  | O |                  |
>  |                  | O |                  |
>  |                  |   |                  |
> v|                  |BAR|                  |     7 point match (Cube: 1)
>  |                  |   |                  |
>  |                  |   |                O |
>  | O           X    |   |                O |
>  | O           X    |   | X  X  X  X  X  O |     Rolled 13
>  | O           X  X |   | X  X  X  X  X  O |     5 points
>  +12-11-10--9--8--7-------6--5--4--3--2--1-+     X: othello
>     1. Cubeful 2-ply    13/10 6/5                    Eq.:  +1.687
>        0.885 0.730 0.414 - 0.115 0.001 0.000
>         2-ply cubeful prune [world class]
>     2. Cubeful 2-ply    8/7 6/3                      Eq.:  +1.686 ( -0.001)
>        0.886 0.727 0.414 - 0.114 0.001 0.000
>         2-ply cubeful prune [world class]
>     3. Cubeful 2-ply    6/5 6/3                      Eq.:  +1.674 ( -0.013)
>        0.884 0.723 0.408 - 0.116 0.000 0.000
>         2-ply cubeful prune [world class]
>    23. Cubeful 2-ply    6/3 5/4                      Eq.:  +1.440 ( -0.246)
>        0.834 0.690 0.336 - 0.166 0.005 0.000
>         2-ply cubeful prune [world class]
>    24. Cubeful 2-ply    8/5 7/6                      Eq.:  +1.418 ( -0.269)
>        0.835 0.647 0.411 - 0.165 0.000 0.000
>         2-ply cubeful prune [world class]
> My choice was 8/5 7/6. I do not insist that my choice is best, but 13/10
> 6/5
> and 8/7 6/3 are clearly not best.
>
>  GNU Backgammon  Position ID: YAMHwHe23RkEAA
>                  Match ID   : cAnzABAAKAAA
>  +13-14-15-16-17-18------19-20-21-22-23-24-+     O: Staf1
>  | X              O | O | O                |     1 point
>  |                O | O | O                |
>  |                  | O |                  |
>  |                  |   |                  |
>  |                  |   |                  |
> v|                  |BAR|                  |     7 point match (Cube: 1)
>  |                  |   |                O |
>  |                  |   |                O |
>  | O                |   | X  X           O |
>  | O           X    |   | X  X  X  X  X  O |     Rolled 64
>  | O           X    |   | X  X  X  X  X  O |     5 points
>  +12-11-10--9--8--7-------6--5--4--3--2--1-+     X: othello
>     1. Cubeful 2-ply    8/2 6/2                      Eq.:  +1.494
>        0.849 0.675 0.351 - 0.151 0.000 0.000
>         2-ply cubeful prune [world class]
>     2. Cubeful 2-ply    13/7 6/2                     Eq.:  +1.485 ( -0.009)
>        0.847 0.669 0.358 - 0.153 0.000 0.000
>         2-ply cubeful prune [world class]
>     3. Cubeful 2-ply    13/3                         Eq.:  +1.315 ( -0.179)
>        0.812 0.619 0.358 - 0.188 0.000 0.000
>         2-ply cubeful prune [world class]
>     4. Cubeful 2-ply    13/9 8/2                     Eq.:  +1.245 ( -0.249)
>        0.798 0.607 0.355 - 0.202 0.000 0.000
>         2-ply cubeful prune [world class]
>     5. Cubeful 2-ply    8/4 8/2                      Eq.:  +1.190 ( -0.304)
>        0.780 0.593 0.345 - 0.220 0.000 0.000
>         2-ply cubeful prune [world class]
>     6. Cubeful 2-ply    13/7 8/4                     Eq.:  +1.071 ( -0.423)
>        0.755 0.556 0.349 - 0.245 0.000 0.000
>         2-ply cubeful prune [world class]
> My choice was 13/7 8/4. Why is this move a super big blunder?
>  GNU Backgammon  Position ID: YAMHwHe2uysAAA
>                  Match ID   : cInpABAAKAAA
>  +13-14-15-16-17-18------19-20-21-22-23-24-+     O: Staf1
>  |                O | O | O                |     1 point
>  |                O | O | O                |
>  |                  | O |                  |
>  |                  |   |                  |
>  |                  |   |                  |
> v|                  |BAR|                  |     7 point match (Cube: 1)
>  |                  |   |                O |
>  |                  |   |                O |
>  | O                |   | X  X  X        O |
>  | O                |   | X  X  X  X  X  O |     Rolled 32
>  | O           X  X |   | X  X  X  X  X  O |     5 points
>  +12-11-10--9--8--7-------6--5--4--3--2--1-+     X: othello
>     1. Cubeful 2-ply    6/4 6/3                      Eq.:  +1.477
>        0.847 0.666 0.350 - 0.153 0.001 0.000
>         2-ply cubeful prune [world class]
>     2. Cubeful 2-ply    5/3 5/2                      Eq.:  +1.430 ( -0.047)
>        0.838 0.659 0.346 - 0.162 0.001 0.000
>         2-ply cubeful prune [world class]
>     3. Cubeful 2-ply    6/3 4/2                      Eq.:  +1.421 ( -0.056)
>        0.836 0.647 0.348 - 0.164 0.000 0.000
>         2-ply cubeful prune [world class]
>    15. Cubeful 2-ply    8/6 5/2                      Eq.:  +1.071 ( -0.406)
>        0.751 0.535 0.335 - 0.249 0.000 0.000
>         2-ply cubeful prune [world class]
>    16. Cubeful 2-ply    8/3                          Eq.:  +0.923 ( -0.554)
>        0.715 0.489 0.329 - 0.285 0.000 0.000
>         2-ply cubeful prune [world class]
> My choice was 8/3. It may be not best, but It never be a super big blunder.
> Why is 6/4 6/3 best???
> I analyzed my match by GNU Backgammon 0.90-mingw 20090625 on Windows xp sp3.
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-gnubg
>
>