Author: James T. Walker
Date: 15:08:45 11/09/98
Go up one level in this thread
On November 09, 1998 at 16:29:23, Dr. Gregor Overney wrote: >Last weekend, I was running some tests with Crafty 16.0 and Crafty 15.20. After >10 games, Crafty 16.0 is leading 1:0. This is indeed only statistical noise. > >Of course, I made sure that 15.20 and 16.0 are compiled with the SAME >optimization levels using exactly the same compiler (VC++ 6.0). > >Gregor I have been running test also. I put in a given position and compare the results after equal search times. This I think is the best comparison. There is a definite difference in the two programs. I'm getting about 11 knps faster search with Crafty16.0. In the following test position: 1rr5/p1qn1k2/2n1p1p1/1p1p3p/3P3P/2P5/PP1Q1PP1/1BR1R1K1/ w (If I got the FEN wrong it is from the Nov Chess Life page 36-top right diag.) The data after a 200 second search looks like this: Crafty 16.0-- Nodes 31,763,620 (Search time 3:20) NPS 158,493 Evals 3,592,678 Score= Draw 1.Bxg6,Kxg6 2. Rxe6+ Kf7 3. Qh6 Nf8 4. Rf6+ Kg8 5. Qg5+ Kh8 6.Qxh5+ Kg7 7. Qg5+ Kh8 8. Qh5+ Crafty 15.20- Nodes 29,700,817 (Search time 3:21) NPS 147,713 Evals 4,135,492 Score -0.66 Same variation. Despite the increased speed by Crafty 16.0 it is losing to Crafty 15.20. There were no Probes involved in the search. I also saw crafty with K,Q,PP vs a K sac the Q for nothing just because it has the KPPK tabelbase and knew it was a win. This looks very strange. Jim Walker
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.