Let me fixthat last sentance: That is not a rare situation. Plus, tree search will find more strange situations than you will find in a game. And playouts will find more strage situations than you will find in the tree.
On Sun, Mar 27, 2011 at 2:27 PM, Michael Williams < michaelwilliam...@gmail.com> wrote: > It doesn't seem as rare as you are saying. Maybe I don't understand the > exact definition of "living by two false eyes". Does this count? > ________ > |.XXO. > |X.XO. > |XXOO. > |OOO.. > |..... > > That is not a rare situation. Plus, tree search will find more strange > situations that you will find in a game. And playouts will find more strage > situations that you will find in the tree. > > > > On Sun, Mar 27, 2011 at 12:25 PM, Aja <ajahu...@gmail.com> wrote: > > As attached, if you are considering such case of "living by two false > eyes", > > it might be good to solve it by some special rules. But I doubt it ever > > appeared in any game in the history. Otherwise, I think "don't fill real > > eye" and "filling false eye is allowed" should be promising for the MC > > playouts. > > > > BTW, I am Aja, not the famous MC program Aya. :) > > > > Aja > > > > ----- Original Message ----- > > From: Daniel Shawul > > To: Aja ; computer-go@dvandva.org > > Sent: Sunday, March 27, 2011 8:19 PM > > Subject: Re: [Computer-go] UCT parameters and application to other games > > Hi Joona > > Thanks for the description. If a simple eye detection such as that > > works , then it should accelerate the MC playouts very much. > > I will try it. > > To Aya : I am afraid not. I am majorly a chess programmer trying my hands > > on Go since a couple of months. > > regards, > > Daniel > > > > On Sun, Mar 27, 2011 at 3:38 AM, Aja <ajahu...@gmail.com> wrote: > >> > >> Hi Daniel, > >>  > >> Are you the one that talked with me (ajahuang) in KGS? If yes, then I > can > >> understand what you meant about "double eyes", but I doubt it ever > >> happened in any 9x9 game in the history.... > >>  > >> Aja > >> > >> ----- Original Message ----- > >> From: Daniel Shawul > >> To: computer-go@dvandva.org > >> Sent: Sunday, March 27, 2011 1:44 AM > >> Subject: Re: [Computer-go] UCT parameters and application to other games > >> Ok I will do that. It has already started performing really well after I > >> allowed > >> the full lenght playouts though. This was I think a major problem > earlier. > >> Especially in the endgame it is outplaying the alpha-beta searcher. > >> Thank you all for your suggestions. > >> > >> On Sat, Mar 26, 2011 at 1:19 PM, Michael Williams > >> <michaelwilliam...@gmail.com> wrote: > >>> > >>> You need to prevent it from filling any real eye.  Regardless of how > >>> many eyes the given chain has.  I cannot tell if that is what you are > >>> doing or not. > >>> > >>> > >>> On Sat, Mar 26, 2011 at 1:05 PM, Daniel Shawul <dsha...@gmail.com> > wrote: > >>> > Well you said single eye which make it a suicide move. It will not > >>> > fill its > >>> > single eye! > >>> > But to prevent filling one of the double eyes, I need to have to code > >>> > to > >>> > detect real double eyes from false eyes. > >>> > I do not do that right now. May be I will just exclude any move which > >>> > reduces us to a single eye. > >>> > That should take care of it I guess. > >>> > > >>> > On Sat, Mar 26, 2011 at 12:57 PM, Christoph Birk > >>> > <b...@obs.carnegiescience.edu> wrote: > >>> >> > >>> >> On Mar 26, 2011, at 9:54 AM, Daniel Shawul wrote: > >>> >> > >>> >>> As I already said before, I _do_ prevent filling of single eyes > >>> >>> (suicide > >>> >>> moves). I thought you guys were talking about > >>> >> > >>> >> We are talking about your own eyes (not suicide) > >>> >> > >>> >> Christoph > >>> >> > >>> >> _______________________________________________ > >>> >> Computer-go mailing list > >>> >> Computer-go@dvandva.org > >>> >> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go > >>> > > >>> > > >>> > _______________________________________________ > >>> > Computer-go mailing list > >>> > Computer-go@dvandva.org > >>> > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go > >>> > > >>> _______________________________________________ > >>> Computer-go mailing list > >>> Computer-go@dvandva.org > >>> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go > >> > >> ________________________________ > >> > >> _______________________________________________ > >> Computer-go mailing list > >> Computer-go@dvandva.org > >> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go > >> > >> _______________________________________________ > >> Computer-go mailing list > >> Computer-go@dvandva.org > >> http://dvandva.org/cgi-bin/mailman/listinfo/computer-go > > > > > > _______________________________________________ > > Computer-go mailing list > > Computer-go@dvandva.org > > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go > > > > >
_______________________________________________ Computer-go mailing list Computer-go@dvandva.org http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
