Author: Uri Blass
Date: 15:12:59 02/13/02
Go up one level in this thread
On February 13, 2002 at 17:56:39, Gian-Carlo Pascutto wrote: >On February 13, 2002 at 17:42:08, Gian-Carlo Pascutto wrote: > >>On February 13, 2002 at 17:35:57, Uri Blass wrote: >> >>>Even if I suppose that we know to evaluate only checkmate in chess more time can >>>help because knowing the final result of the game one ply earlier can help in >>>part of the cases to find the right move. >> >>Go has no checkmate. The game is over when both players agree it is over. > >I know what you are going to ask next, so I'll already answer: > >in that case, the game would keep going until you have filled all but one of >your liberties. at that point, the opponent will capture all your stones in one >move. you can continue playing at that point, again playing on until you have >one liberty left, the opponent again capturing all your stones. you can keep >doing this until there are only single-eyed groups of your opponent left, at >which point you have no more legal moves and have to pass. > >This could take quite a while. But I guarantee you your opponent will have >fallen asleep of boredom or killed you out of annoyance before it gets so far. > >-- >GCP I understand In other words you say that if there is a mate in go that programs can prove it is only a long mate when it is practically impossible to do errors and miss the mate. I still believe that faster hardware can help but in this case you must do selective search. One possible way to use faster hardware is to give the program to play against itself. If it sees based on the games against itself that position A is good for it when position B is bad for it then it can understand that position A is probably better than position B when if it has faster hardware it may get more information based on games against itself. Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.