Author: Uri Blass
Date: 22:40:39 08/10/00
Go up one level in this thread
On August 10, 2000 at 20:39:24, James Robertson wrote: >On August 10, 2000 at 20:06:15, Adrien Regimbald wrote: > >>Hello, >> >>Fritz is designed to play well in a "normal" game of chess. Fritz would have >>absolutely slaughtered the opponent a LOOOONG time before that position arose >>had the opponent been playing as per the sample line leading to that position. >>The chances of a position like that ever coming up in a real game of chess are >>about nil. If I were Frans, I wouldn't even give this position a second of my >>thought (well, beyond possible implications that Fritz' capture extensions are a >>bit too greedy..). >> >> >>Regards, >>Adrien. > >Of course. This position means nothing to a real player or program. But it is >interesting and fun to find positions that "dazzle" programs because it gives us >insight into what they are thinking. > >James I disagree that it means nothing. People use programs also to analyze strange positions and they should know that it is better to buy Hiarcs or Junior and not to buy Fritz in this case. I think that it is easy to prevent the problem by having an upper bound for the number of plies that are searched with extensions(Qsearch is one kind of extension). Junior does it and my experience is that it never searches more than 2x plies at iteration x that mean seeing everything at depth x/2. I am not sure about the numbers and they may be slightly different (2x+1 instaead of 2x or (x+1)/2 instead of x/2 when x is odd) but the idea is clear. 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.