Author: Roland Pfister
Date: 23:24:50 05/24/98
Go up one level in this thread
On May 21, 1998 at 12:42:13, Ernst A. Heinz wrote: >On May 20, 1998 at 06:58:47, Roland Pfister wrote: > >>Watcom-Patzer on K6-233 DOS Window under WinNT 4.0 (16 MB Hash ?) >> >>It.5 0:02 -M7 Bxa8 >> 0:03 3.35 Ne5 >>It.6 0:06 3.70 g4 >>It.7 0:22 3.62 g4 >>It.8 2:01 3.12 b6 >>It.9 3:28 3.46 g4 >> >>That is the version with lazy evaluation window = 130 ( = two pawns ). >>Djordje, if you want to reproduce it, set the variable "poswin" to 130. >> >>The previous version that had a lazy eval window of 96 (1.5 pawns) >>needs very long to see that Bxa8 looses. So a minor change produces >>very different results. > >Really strange ... but anyway, how does "Patzer" see the mate >(if it really is one, I am still not sure) so quickly? > >Do you use singular extensions or something else special? No > >Or is it maybe a bug in "Patzer" that makes it solve this very >position extremely fast? > You are right. I do some forward pruning that usually gets corrected by deeper searches if it prunes wrong. But if falsely results in a negative mate value it will never be corrected :-( So Patzer pruned Bxa8 and had luck because it is a bad move. If it would have been a good move it would be desastrous... >=Ernst=
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.