In my own experience in StoneGrid, in the playout phase, when it
encountered a ko, it knows it cannot immediately take it. It then put it
into a stack. And plays elsewhere. When the other side plays, it either
answer last move, or connect the ko. And if it answered the last move,
then the other side can take it back. It goes back and forth for quite a
while before it settles. It is a little bit more complicate in the
detail. But that's the overall approach.
It can get the test case right in mostly for 50K simulations. But still
sometimes it picks the wrong one.
Interesting.
So that a consequence of automatic replies in semeai would be to make
the losing side ko master.
Jonas
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
