Your idea is more in the spirit of MC, I like it.

Another idea is borrowed from my first reasonable MC player.   I looked
at the "futures" of interesting move points  and discouraged self-atari
moves unless the future belonged to the player executing the move.   (A
"future" is the expected percentage of time a given player ended up with
a given point at the end of the random games.)   So some sort of
pre-processed quick all-moves-as-first random play-out can give you a
sense of which self-atari points are interesting.   But it is not
dynamic unfortunately and thus not scalable unless done periodically
during the tree search.

- Don


Jonas Kahn wrote:
>> I think the general outline is that you pre-test groups first to see if
>> a self-atari move is "interesting."    It's worthy of additional
>> consideration if the stones it is touching have limited liberties and
>> the group you self-atari is relatively small.    Then you could go on to
>> other tests which will consume even more time of course.  
>>     
>
>
> Here is an alternative, still in the line of gathering information from
> play-outs.
>
> Do not turn down self-ataris TOO agressively.
> When a self-atari occurs in play-out:
> - notice which (and when).
> - see if the self-ataried position is yours at the end (indeed it should
>   for a nakade).
>   If not: cancel the play-out (or start again from the moment you
>       self-ataried) Add a -1 to self-atarying at this place.
>       After say 10 (-1): strictly forbid this self-atari in future
>       playouts.
>   If yes: Remove all constraints on this self-atari (maybe give a bonus,
>   or study within the tree).
>
> Notice that this would also deal with seki: self-atari for both sides
> getting completely forbidden at one point.
>
> Jonas
> _______________________________________________
> computer-go mailing list
> computer-go@computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/
>
>   
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to