[Bug-gnubg] bug in 3 ply equities

[Hugh Sconyers]

 

Hi

 

There is something strange going in this position.

 

side on roll has:

1-24, 1-23, 1-13, 2 each 9 8 7 6 5 4 pc = 138

 

side not on roll has 1- 22

 

it is dmp.

 

if you ask gnu for an evaluation, it give 9.69% mwc.  which is a little high, but close to the right answer of 8%.

 

if you now ask gnu how to play 2-1 it gives these results:

 

    1. Cubeful 3-ply    23/22 13/11                  MWC:  31.18%
       0.312 0.000 0.000 - 0.688 0.024 0.018
        3-ply cubeful [grandmaster]
    2. Cubeful 3-ply    24/22 23/22                  MWC:  31.13% ( -0.05%)
       0.311 0.000 0.000 - 0.689 0.031 0.017
        3-ply cubeful [grandmaster]
    3. Cubeful 3-ply    23/21 13/12                  MWC:  30.96% ( -0.22%)
       0.310 0.000 0.000 - 0.690 0.023 0.019
        3-ply cubeful [grandmaster]
    4. Cubeful 3-ply    13/10                        MWC:  30.86% ( -0.32%)
       0.309 0.000 0.000 - 0.691 0.025 0.019
        3-ply cubeful [grandmaster]
    5. Cubeful 3-ply    24/21                        MWC:  30.74% ( -0.44%)
       0.307 0.000 0.000 - 0.693 0.030 0.017
        3-ply cubeful [grandmaster]
    6. Cubeful 3-ply    24/22 13/12                  MWC:  30.68% ( -0.50%)
       0.307 0.000 0.000 - 0.693 0.031 0.017
        3-ply cubeful [grandmaster]
    7. Cubeful 3-ply    23/20                        MWC:  30.61% ( -0.57%)
       0.306 0.000 0.000 - 0.694 0.024 0.019
        3-ply cubeful [grandmaster]
    8. Cubeful 0-ply    24/23 13/11                  MWC:   7.42% (-23.76%)
       0.074 0.000 0.000 - 0.926 0.018 0.010
        0-ply cubeful [expert]

 

 

the 0 ply numbers are very close to right and all the 3 ply number are way off and most likely all the moves are about the same.  this looks like a bug to me.

 

regards

hugh