Re: Fwd: [Bug-gnubg] Playing the same position over and over for analysis.

[Jim Segrave]

On Mon 20 Oct 2003 (03:21 -0400), address@hidden wrote:
>  

> Return-path: <address@hidden>
> From: address@hidden
> Full-name: Michaeldepreli
> Message-ID: <address@hidden>
> Date: Mon, 20 Oct 2003 03:20:57 EDT
> Subject: Re: [Bug-gnubg] Playing the same position over and over for analysis.
> To: address@hidden
> MIME-Version: 1.0
> Content-Type: multipart/alternative;
>       boundary="part2_b4.23c18b32.2cc4e6d9_boundary"
> X-Mailer: 7.0 for Windows sub 10502
> 
> OK
> I believe someone at the GNU team did such an experiment on one of Neil 
> Kazaross's back game positions.
> Can they send me the batch program please?

I did some testing on a position Neil K flagged as being erroneous
(although I think someone else first raised it), but I used the Unix
syslog facility to get the output, which won't help you very
much. It's not too hard to get gnubg to output moves as it makes them,
getting cube actions or marking the end of a game where truncation is
being done, either as a truncated rollout or truncating at the rollout
database is not feasible - the game record would simply end.

If there's just a small number of positions you want rolled out, the
easiest way is for me to hack back in syslogging and for you to send
me the position(s) you want rolled out and the settings you need for
this. I'll send you back a collection of .sgf files representing the
rollouts. 


-- 
Jim Segrave           address@hidden