Re: [Bug-gnubg] Bug in analysis - position enclosed - more details

[Ned Cross]

Hi,
    I don't know the definition of "crashed" position.  I used the term
because "evaluate" prior to the roll gave the output below.

I'm wondering if there are two separate issues here:

1) That GNU doesn't understand this position.

2) That the reported 2-ply win %  for the candidate move ("after the roll")
does NOT equal the reported 2-ply win %  for the position resulting from the
candidate move ("before the next roll"). Right or wrong, shouldn't those two
numbers be the same?

Ned

__________________________

GNU "Evaluate" output

Evaluator:  CRASHED

Input           Value               dE/dI
OFF1            0.000              -0.168
OFF2            0.000              -0.489
OFF3            0.000              -0.188
BREAK_CONTACT   1.000              -0.597
BACK_CHEQUER    0.000              -0.177
BACK_ANCHOR     0.000              -0.664
FORWARD_ANCHOR  0.000              -0.156
PIPLOSS         0.000 (0.000 avg) -0.632
P1              0.000 (0.000/36)  -0.150
P2              0.000 (0.000/36)  -0.630
BACKESCAPES     0.000 (0.000/36)  -0.150
ACONTAIN        0.000 (0.000/36)  -0.533
CONTAIN         0.000 (0.000/36)  -0.441
MOBILITY        0.000              -0.281
MOMENT2         0.000              -0.074
ENTER           0.000 (0.000/12)  -0.207
ENTER2          0.000              -0.055
TIMING          0.000              -0.202
BACKBONE        0.000              -0.052
BACKGAME        0.000              -0.138
BACKGAME1       0.000              -0.066
FREEPIP         0.000              -0.174

        Win    W(g)   W(bg)  L(g)   L(bg)  Equity  (cubeful)
static: 0.707 0.000 0.000 0.054 0.003 (+0.359  (+0.502))
 1 ply: 0.692 0.000 0.000 0.075 0.004 (+0.310  (+0.425))
 2 ply: 0.659 0.000 0.000 0.080 0.003 (+0.239  (+0.358))


No double           : +0.358
Double, pass        : +1.000   (+0.642)
Double, take        : +0.104   (-0.254)

Correct cube action: No double, take



----- Original Message -----
From: "Joseph Heled" <address@hidden>
To: "Ned Cross" <address@hidden>
Cc: "gnubg" <address@hidden>
Sent: Monday, October 28, 2002 3:54 PM
Subject: Re: [Bug-gnubg] Bug in analysis - position enclosed - more details


>
> Hi,
>
> This is not a "crashed" position. GNU confusion stems from mis-evaluating
the
> position after 6/1 5/1, even at 2ply (not entirely surprising in this case
since
> you are averaging 441 mis-understood positions, which is unlikely to
improve
> things).
>
> 2p:           73+0  22.6+6.7
>
> 1296 rollout: 81.7+0 18.3+0
>
> If a position and most of it's continuation are mis-evaluated, then the
cubefull
> numbers are not reliable too, sometime even plain contradictory.
>
> -Joseph
>
>
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/bug-gnubg