Author: Bernhard Bauer
Date: 06:07:00 10/12/98
Go up one level in this thread
On October 12, 1998 at 05:53:21, Ernst A. Heinz wrote: >On October 12, 1998 at 05:02:09, Bernhard Bauer wrote: > >>On October 11, 1998 at 01:37:52, Bruce Moreland wrote: >> >>> >>>On October 10, 1998 at 21:25:25, James T. Walker wrote: >>> >>>>I tested Crafty15.20 tonight and it solves the problem instantly. I also tested >>>>the following computers/programs. ALL solved the mate instantly. >>>>Fritz5.0,Rebel Decade 2.0,Rex2.3,Fidelity Mach3,Fidelity Mach IV,M-Chess Pro 3.5 >>>>M-Chess Pro 5.0,Genius 4.0, Genius 2.0,Kasparov's Gambit,CM4000,CM5000,CM5500 >>>>and my handheld Mephisto Marco Polo(USCF about 1800). I do not have a single >>>>program/computer that does not solve this instantly. Maybe the creator of JR. >>>>should take a look at this problem. >>>>Jim Walker >>> >>>After you check to see if they find the mate, please also check to see if you >>>can play out the main line by making each move. Also, check to see whether it >>>works with colors reversed. I bet that there are cases where programs will call >>>the game a draw, without executing the mate, or without allowing you to execute >>>the mate. >>> >>>bruce >> >>Yes, we should be carefull in this type of position. >>For the following position: >>White: Kc8, Bg8 >>Black: Ka8, Ba7 >>White to move >>FEN: k1K3B/b/8/8/8/8/8/8 w >>Crafty 15.20 will move Bh7 instead of mating with Be5# and it will tell us >>game is a draw due to insufficient material. The learning is wrong too. >>Crafty says: learning position, wtm=1 value=-66 >> >>By the way, the "insufficient material feature" has further consequences. >>For the following position, which is a mate in 6 Crafty has no clue. >>FEN: 8/1k1K4/8/8/1pN5/1N6/8/8 w > >"DarkThought" knows about insufficient material but *also* about the >exceptions. It solves your above examples within fractions of a second. > >As I have already pointed out before, the efficient implementation framework >for interior-node recognition allows us to incorporate such knowledge with >hardly any losse of search speed. > >=Ernst= Fully agreed! Its usually wrong to say that something "wrong" has to be done for this and that reason. The better aproach is to first make it right and than fast. Glad to hear you have allready implemented it that way. Kind regards Bernhard
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.