Author: Tord Romstad
Date: 08:43:17 05/04/04
Go up one level in this thread
On May 04, 2004 at 10:57:34, Tony Werten wrote: >On May 04, 2004 at 10:55:20, martin fierz wrote: > >>On May 04, 2004 at 10:53:44, martin fierz wrote: >> >>>On May 04, 2004 at 10:44:43, Tord Romstad wrote: >>> >>>>When there is little thinking time left, my engine annoyingly often fails to win >>>>KQKR endgames. Are there any simple but effective eval tricks which can help to >>>>play such endgames well without searching very deeply? >>>> >>>>Tord >>> >>>none really that i know of. of course there are two, but those are obvious, the >>>first is easy, the second is hard: get the defending king to the edge of the >>>board, and if possible, separate king and rook of the defender. but it's tough >>>to win this on knowldege, i would say - humans at least can hardly win against a >>>computer defence, and they are nice examples of chess-playing entities not >>>searching very deeply :-) >>> >>>cheers >>> martin >> >>umm, forgot one: of course your king has to be close to the defender's king too. >>and i really can't think of anything else, but then i never looked at the theory >>of this endgame. i had to play it once, but for a human player it's pretty >>irrelevant to know about this... Thanks for the ideas, Martin! Unfortunately I already use mostly the same heuristics as you describe above, but perhaps it would work better with some tuning of the weights. Experimenting with null move tricks, like Gerd suggested, is another possibility. >You forgot 1 more: Use tablebases .... No, they are too big. The KQKR tablebases alone consume more than 1 MB. :-( Tord
This page took 0.01 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.