On Sat, 2008-10-11 at 17:01 -0400, Michael Williams wrote: > Don Dailey wrote: > >> Speaking of which, wouldn't it be better to limit consecutive Kos instead > >> of limiting consecutive 1-stone captures? > > > > Wouldn't this definitely slow down most if not all light implementations > > and require code changes to existing programs? My own programs do not > > really know that a ko exists until it decides to make that specific > > move. Then when it tries to play that move the make() routine discovers > > that it is was a ko and reject it, returning an illegal move value. > > > > To fix this I would have to test every move just to see if there is a ko > > situation (although there are no doubt some shortcuts.) > > > I detect simple Ko at the time of the legal capture. I'm curious how you do > it at the time of the illegal capture attempt. How do you tell a legal > 1-stone > capture from an illegal one?
An illegal one stone capture is a capture of the same stone that was just used to make a 1 stone capture itself. I think I got that right. Whenever there is a 1 stone capture I flag that and remember it's location (in the move list stack) but a one stone capture does not imply a KO and that takes time to test - so I don't test. So I don't know that a capture move is a KO and most of the time I don't need to know. The next player selects a random move, which MIGHT be a capture attempt which is illegal due to simple KO. If it is, only then do I know that the previous move was a KO. I'm slightly confused by the language "ko" what is a ko? To me a ko is illegal so it never happens but I suppose my definition is wrong and it's any board position where an illegal repetition is possible due to whatever ko rule is in effect. - Don > _______________________________________________ > computer-go mailing list > [email protected] > http://www.computer-go.org/mailman/listinfo/computer-go/
signature.asc
Description: This is a digitally signed message part
_______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
