Re: [computer-go] Pattern matching - example play

["Frank de Groot" <frank@xxxxxxxxxxxxx>]

----- Original Message ----- 
Subject: Re: [computer-go] Pattern matching - example play


> In going from 2 dan to 3 dan, one of the
> main faults which I corrected in my play was
> not answering the stronger player and taking
> gote so much.  That is, my move was the correct
> one in the local area, but I should have taken
> sente and played somewhere else.
>
> I am wondering whether a pattern matching approach
> such as the one Frank has been discussing can
> deal with this issue, or whether it might not
> have the same problem.



The pattern system sorts moves in order of urgency.
The urgency is based on statistical data.

It does not care about what has just been played, every move it looks at all
legal moves and sorts them in order of statistical urgency.

During a game it often switches attention (temporarily) to another area just
like a human player would do.
I think a pro-prediction of 45% is around the theoretical maximum,
considering that pro's don't alway play optimal (or there are equally good
moves), and that a pattern-based system has many, and severe, limitations.


_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/