Re: computer-go: Insight of a human continued

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

At 03:20 PM 9/1/00 +0200, Christian Nentwich wrote:
>
>My question is then:
>  - Does anyone have a cutoff mechanism for reading which detects a
>"natural
>    point for tenuki" instead of reaching a tree-depth cutoff ? (I'm 
>    not talking about 5-liberties-is-safe either)

Yes.  When the local temperature falls to the whole board temperature.
Many Faces has a notion of the value of sente, and at each ply gives
the player to move the options of continuing in the local fight or taking
sente.  If the change in value due to the local move is less than the
value of sente, then it is time to tenuki.

>  - Does anyone have any ideas about how to judge a thickness/territory 
>    tradeoff using an algorithm ?

All programs have some way of doing this, but none of them are ideal.  I
radiate influence from groups to empty points, based on the strength of
the groups and distance.  Partial points indicate thickness.  I've adjusted
the algorithm so it gives about 2 points total of influence for each outside
facing stone in a strong group.

>
>By the way, as for the discussion about patterns: I don't think patterns
>are obsolete in a good go-playing program, at least not if you are
>trying to play like a human player. Human players use patterns to
>consider whole candidate sequences instead of single moves, e.g. this
>white crosscut happens a lot

Many Faces also associates sequences of moves with patterns.

-David

>
>..........
>.....ox...
>...x.xo...
>..........
>
>Here's a well know candidate sequence that everybody knows in this
>pattern (atari-connect, threatening a ladder and giving a good fight for
>black):
>
>..........
>.....1....
>....2ox...
>...x3xo...
>..........
>
>
>Christian Nentwich
>(2d)
>
>