[Bug-gnubg] RE: GNU Help Please

[Øystein Johansen]

>===== Original Message From address@hidden =====
>Hi
>Here's another one.
>I'm pretty sure the GNU settings are set to 2ply. I think if this wasn't the
>case there would be many more differences.
>In this 10 game money session this is the only discrepancy between play
>chosen and analysis.
>Same kind of bear in move.
>Position ID 620DAAy/Y2MAAA
>Match ID QYkaAEAACAAA
> 1. Cubeful 2-ply    6/1      Eq.:  -0.335
>   0.442 0.016 0.000 - 0.558 0.103 0.000
>   2-ply cubeful 100% speed, 8 cand., 0.16 tol. [world class]
> 2. Cubeful 2-ply    9/4      Eq.:  -0.532 ( -0.197)
>   0.377 0.025 0.000 - 0.623 0.192 0.003
>   2-ply cubeful 100% speed, 8 cand., 0.16 tol. [world class]

Yes, obviously moving 9/4 is a blunder, and gnu moves this
horrible move at 0-ply.

I'm not sure why this move was selected during play. I have two theories:

1. The acctual setting was 0-ply. (Due to a bug in Dueller)
2. The candidate pruning didn't have enough tolerance to make a deeper search.
  (I can see that 0-ply actually evaluate 9/4 as 0.431 (!!) better than 6/5.)

Theory number two is very likly, I believe.

>I'm using GNU no gui exe 14 Aug 2002 to play and
>GNU windows version 1 oct 2002 to analyse.

Is this runned from Dueller?

>I've just realised there is a newer no gui exe released.

A newer gnubg-no-gui.exe will be released later today.

>Could this be the problem?

No, I don't think so. I think this is one of the above problems. There's no
difference in the evaluation engine from the two release dates.

>Also I'm using World Class to play Snowie against GNU do you think this is an
>equivalent playing strength versus  Snowie 3 Ply Huge 100%?

I'm not sure how Snowie makes the move selection pruning. I see Albert's post
in the thread.

>Thanks

-Øystein