Re: [Bug-gnubg] Holding games

[Joseph Heled]

 - No doubt gnuBG 2ply gets this wrong. Not for the first time nor for 
the last either.

- We all want GNU to be correct every time on each position, however 
this is not realistic.

- In order to make a case we need more than one position. Probably in 
the order of hundreds or thousands.

- Even with the above, there is no guarantee we can do something about 
it. ( In the sense of improving it and not degrading the rest).

- It is best if someone keeps all those positions for future reference. 
With the new bug system, can we have a category for it (or maybe one 
already exists?)

-Joseph


Albert Silver wrote:
> The following position from Depreli's series came up in which GNU 0.14
> chose one of the weakest replies:
>
> GNU (O, 0 pts) vs. Albert (X, 0 pts) (money game)
> Move number 6: O to play 63
>
>     GNU Backgammon  Position ID: Ms/gBwBYd4PBAA
>                     Match ID   : MAEPAAAAAAAA
>     +-1--2--3--4--5--6-------7--8--9-10-11-12-+  O: GNU
>     |          O  O  O |   | O  O           X |  0 points
>     |          O     O |   | O  O           X |  Rolled 63
>     |                O |   | O              X |  
>     |                  |   |                X |  
>     |                  |   |                6 |  
>     |                  |BAR|                  |^ (Cube: 1)
>     |                  |   |                  |  
>     |                X |   |                  |  
>     |                X |   |                  |  
>     |          X     X |   | O  X           O |  
>     |    X     X     X |   | O  X           O |  0 points
>     +24-23-22-21-20-19------18-17-16-15-14-13-+  X: Albert
> Pip counts: O 130, X 128
>
> * GNU moves 13/10 13/7
>
> Rolled 63 (+0.019):
> *1. Cubeful 2-ply    13/10 13/7           Eq.:  +0.214
>     56.7%   2.3%   0.0% -  43.3%   0.7%   0.0%
>  2. Cubeful 2-ply    8/5 8/2              Eq.:  +0.203 ( -0.012)
>     56.1%   3.2%   0.0% -  43.9%   1.0%   0.0%
>  3. Cubeful 2-ply    8/5 7/1              Eq.:  +0.202 ( -0.012)
>     56.1%   3.1%   0.0% -  43.9%   1.0%   0.0%
>  4. Cubeful 2-ply    8/2 7/4              Eq.:  +0.189 ( -0.025)
>     55.8%   2.9%   0.0% -  44.2%   1.0%   0.0%
>  5. Cubeful 2-ply    8/2 6/3              Eq.:  +0.188 ( -0.026)
>     55.7%   3.0%   0.0% -  44.3%   1.0%   0.0%
>
> MOVE            8/2 8/5         8/2 7/4         8/2 5/2         8/5 7/1
> 13/7 13/10 
> BOTS' CHOICE S3, G13 	  G12   			  S4
> G14, J 
> ROLLOUTS 	 0.098 	 0.084 	 0.077 	 0.074 	 0.058
> Rollouts by SW3.2 1296 trials 2ply huge 3ply cube with bearoff database.
>
> BTW, G12 is Gnu 0.12, G13, Gnu 0.13 etc.
>
>
> Neil Kazaross, who had chosen the correct move unsurprisingly, added:
> We see an example of why I am hoping more training in holding games will
> improve GNU even more.
>
>
>                                                         Albert
>
>
>
>
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/bug-gnubg