Author: Charles Roberson
Date: 16:41:21 01/04/06
Go up one level in this thread
On January 04, 2006 at 18:36:53, Stuart Cracraft wrote: >On January 04, 2006 at 18:35:12, Charles Roberson wrote: > >>On January 04, 2006 at 18:22:29, Roman Hartmann wrote: >> >>>On January 04, 2006 at 17:41:54, Charles Roberson wrote: >>> >>>> >>>There is certainly the danger of 'seeing too much'. That applies also to simple >>>tactics. As an example the engine sees suddenly that it's getting mated in 20 >>>moves and starts to throw away material to delay the mate while the opponent >>>didn't even saw/calculate the mate. So the the engine might play worse in some >>>games even though it is searching deeper than the opponent. >>> >>>regards >>>Roman >> >> Yes! Seems to me your position evaluator would have led you to that position. >>The PE gives strategic guidance while the search procedure(s) keep the engine >>from tactical blunders. Then the engine winds up in the situation that you >>descirbed and assumes that the opponent sees what it sees and then starts >>tossing its pieces! The opponent may never see the mate and the game may >>continue for 40 more moves instead of 20, but your engine is now down material >>because it saw deeper than the opponent. > >I'd like to see an example from reality, a real chess program example, >before I could believe it! It's easy to forsee when you think of extensions. Lets say you get into the unsound position and your engine has check and single-reply extensions. Your opponents engine only has check extensions. Thus, you could search much deeper in the loosely described position than your opponent. My own experiements with single-reply extensions show a definite increase in deep mate detection abilities as long as it is paired with check extensions.
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.