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

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 12:52 1-12-2004 -0500, Chris Fant wrote:
>> I think human Go pro's play Go very sub-optimally so learning from pro
games
>> will never produce anything "better" than a Go pro, *but it will keep pace*
>> with pro patterns, for example.
>
>I disagree with this.  The limiting factor in Go skill if the bad
>moves that one makes.  As many of the pros who are generating your
>learning games are not making the same mistakes, their mistakes will
>disappear from the system's memory due to the law of large numbers.

the largest number of pro's aren't world champion, so the 'law of the large
numbers' only means the system will play real poor. Not that it plays better.

However add a 30 ply search which the pro doesn't have, as he sometimes
makes a 20 ply mistake which the search will not.

Then the pro is beated.

So the first problem in go is to get, no matter how selective, to that 30
ply search tactical and to not make any tactical error within 30 ply.

Of course that doesn't require a 30 ply search in itself. Probably 15 ply
is more than sufficient.

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