On 6/13/05, Paul Pogonyshev <[EMAIL PROTECTED]> wrote: > Gunnar Farneb�k wrote: > > I wrote: > > > For the fuseki testing against people and other programs is > > > particularly important. It's easy to overadapt and find something that > > > is effective against GNU Go but possibly not against anyone else. I > > > recently saw a player on KGS devastate GNU Go with a parallel 5-5 > > > opening, for example. So some testing on KGS should be worthwhile. > > > > Here is how to secure more than 200 points of territory before move > > 20. > > Ouch. > > I think it highlights the weakness of relying on patterns too much.
Agreed. However, some more general patterns could improve things. I'm thinking things like |..??? |...?? |..*.? |..... |..... +----- as a sort of generalized 3-3 invasion. Similar things on the edges, etc. But I like the algorithmic fuseki / moyo moves a lot better, assuming they can be made to work reasonably in actual play. Evan
_______________________________________________ gnugo-devel mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnugo-devel
