Author: Komputer Korner
Date: 12:29:53 08/15/98
Go up one level in this thread
On August 04, 1998 at 11:02:44, Christophe Theron wrote: >On August 03, 1998 at 18:24:25, Komputer Korner wrote: > >(snip) > >>We are getting hung up on the null move issue. Crafty uses null move but is >>still a full width program. CM 5000/5500 and Rebel 9 are selective searchers. > >Really? May we hear from you the evidences you have that CM5000 and R9 don't use >null move??? > > > Christophe > > >(snip) it would be simple to test for CM5500 and Rebel 9 plus all the others. Do the same as the following by Moritz Berger. "From computer vs. computer games, I choose positions with eval-jumps, either immediately or fast increasing/decreasing evaluations by both programs. Sometimes a program sees a few moves earlier the importance of a particular move. Among equally strong programs, there often are single key moves that eventually decide the outcome of a game. It's quite easy to spot them by looking at the analysis recorded during the game, a very important feature in Rebel 8, Genius 5 and Hiarcs (of course Crafty with its full log really shines in this respect, also W-Chess offers nice insights with its similar capability). After finding such positions, I have a perfect test suite for different relevant positions (i.e. they are taken from real computer vs. computer games, so both the problem and its solution are likely to be identified by means of a chess program). Here's one example (1st 2 lines are in .EPD format, you will have to join them into one line to make it readable by some programs). The theme is king safety at shorter time controls, after the name of the program I give the solution time in seconds, in brackets the search depth needed for the solution, in some cases alternative moves that include the solution in the PV: r1r1n1k1/5p2/p1b1p1p1/4P2Q/PpNq4/6R1/1PB3PP/R6K b - - id=Human - REBEL 8.0 P5/133; bm=e8g7; Hiarcs 4 (Fritz 4 engine): 1"(3): Ne8g7 Genius 3: 1"(4): Ne8g7 Hiarcs 5: 3"(3): Ne8g7 Genius 5: 4"(3): Ne8g7 Hiarcs 3: 6"(4): Ne8g7 Fritz 3.1 (Fritz 4 engine): 6"(8): Ne8g7 Fritz 4.01: 7"(8): Ne8g7 Mchess 6: 17"(5): Ne8g7 Mchess 5: 35"(5): Bc6xg2 53"(5): Ne8g7 Rebel 7: 57"(7): Ne8g7 Rebel 8: 89"(7): Bc6xg2 104"(7): Ne8g7 Fritz 1.2: 404"(10): Ne8g7 Please note that the programs without Null-move (Rebel 7,8 and Fritz 1.2) have more problems with this position ... M-Chess 6 is mucho faster than M-Chess 5, this shows the huge progress that Marty Hirsch has made with his latest version ... Moritz" Perhaps Thorsten has done some tests on this but with zugswang positions and can add to this list of null movers. Nimzo 98, Fritz 5, Crafty, Now, Zarkov, Ferret. Zugswang positions are null movers achilles heel. -- Komputer Korner
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.