2011/6/15 Darren Cook <[email protected]> > > Funny the details note: > > More time would probably favour a MCTS program, > > I think, at current hardware levels at least, Zen's results on KGS has > proven me wrong on that. > > Darren > > Longer times like 40 Secs/move may produce quite optimal play that MC program can reach. But it is the relative differences that matter. Selective searcher like human gain almost linearly on thinking time while program gains logarithmically.
So on short times like 10 Secs move humans do not really have chance. We will lose in two ways: First is that time just wont allow thinking and second is that we need physically move stone/mouse to make a move and lose another 1 second or two there. This is by no means different for chess either. Humans stop having fighting chance in blitz lonf before they lost it in longer games. So computer loses on short times a bit and a human a lot. Petri
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
