[computer-go] Standard references on CGOS
Jason House
jason.james.house at gmail.com
Thu Nov 1 07:26:14 PDT 2007
I think I've found a bug in your bot(s). It seems to resign at
inappropriate times. In both cases, your bot was playing as white and mine
as black.
myCtest-V-0008 resigned incorrectly in
http://cgos.boardspace.net/9x9/SGF/2007/11/01/179849.sgf
(I think optimal endgame is J5 leading to a 36 to 52.5 victory for you)
myCtest-UCT-10k resigned incorrectly in
http://cgos.boardspace.net/9x9/SGF/2007/11/01/179826.sgf
(White F9 seems to be the obvious move to keep the top. Black is also
vulnerable at D2)
On 10/29/07, Christoph Birk <birk at ociw.edu> wrote:
>
> On Mon, 29 Oct 2007, Christoph Birk wrote:
> > myCtest-10k-UCT: 10000 random playouts guided by a UCT search (1350 ELO)
> > * nodes are expanded after 50 runs through them
> > * UCT_score = win_ratio + 0.5 * sqrt(log(N)/n)
>
> I added variants with different expansion thresholds
> 'myCtest-V-0006': 25
> 'myCtest-V-0007': 10
>
> Christoph
>
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://computer-go.org/pipermail/computer-go/attachments/20071101/fca4705a/attachment.htm
More information about the computer-go
mailing list