Re: [Bug-gnubg] Predefined move filters

[Gary Wong]

On Thu, Jan 02, 2003 at 02:47:34PM -0000, Ian Shaw wrote:
> I get the same result both ways round. Am I doing something wrong?
> 
> # 0.5122      0.1417  0.0073  0.1231  0.0055  +0.0449 n/a
> # 0.0218      0.0067  0.0010  0.0067  0.0008  0.0508  n/a
> 
> # 0.5122      0.1417  0.0073  0.1231  0.0055  +0.0449 n/a
> # 0.0218      0.0067  0.0010  0.0067  0.0008  0.0508  n/a

Very strange!  I don't understand what could be going wrong.  When I
use your script (with two modifications -- "set confirm new off" at
the top, to run non-interactively, and "set rollout trials 36" instead
of 1296), I get:

Chequer play in rollouts (for player 0) will use 0 ply evaluation.
Chequer play in rollouts (for player 1) will use 1 ply evaluation.
...
                               Win  W(g) W(bg)  L(g) L(bg) Equity   Cube E    n
            Current position 0.719 0.125 0.001 0.134 0.005 ( 0.424)    n/a   36
              Standard error 0.120 0.037 0.004 0.030 0.004 ( 0.262)    n/a
...
Chequer play in rollouts (for player 1) will use 0 ply evaluation.
Chequer play in rollouts (for player 0) will use 1 ply evaluation.
...
                               Win  W(g) W(bg)  L(g) L(bg) Equity   Cube E    n
            Current position 0.549 0.142 0.002 0.121 0.005 ( 0.117)    n/a   36
              Standard error 0.128 0.037 0.003 0.042 0.004 ( 0.290)    n/a

which looks pretty sensible.  Doesn't make any sense to me...

Cheers,
Gary.
-- 
   Gary Wong           address@hidden           http://www.cs.arizona.edu/~gary/