Author: Uri Blass
Date: 11:00:18 04/19/03
Go up one level in this thread
On April 19, 2003 at 13:39:28, Drexel,Michael wrote: >On April 19, 2003 at 08:37:14, Sune Larsson wrote: > >> >> [D]r2r2k1/ppq1bpp1/2bppn1p/6B1/3NPP1P/2N5/PPPQ2P1/1K1RR3 b - h3 0 15 >> >> This is from a game Larsson-Fritz 7. Black played 15.-hxg5? and lost quickly. >> Test is avoid 15.-hxg5. >> >> /s > >Shredder is incredible strong, it would not have played 15...hxg5. >It even wouldnt play ...hxg5 in this position: > >[D] 6k1/ppq1bpp1/3ppn1p/6B1/4PP1P/2N5/PPP1Q1P1/1K6 b - - 0 5 > > >Stupid Shredder gives a penatly <-3 for this: > >1. > >[D] k1K5/8/8/8/7p/6n1/7P/8 w - - 0 5 > >2. > >King on g(b)-file or h(a)-file > >3. > >Material for both sides > X > > >If someone intends to do it in the same way, I suggest to check if the opponents >qeen and at least one rook is still on the board. > >Michael I do not plan to do it in the same way as shredder. I think that it only does shredder weaker and shredder could probably be stronger(at least against computers) without that stupid code. I agree that the position is bad for testing king safety if you do not give in the same test suite also positions when capturing is the only good move. I do not see a reason to do a program weaker against computers only for doing it stronger against weak humans so weak humans will be able to beat it only in 0.1% of their games and not in 0.2% of their games. You already proved that there are practical cases when shredder will avoid taking the gift and lose because of that and it seems that search does not help when deep search could usually help to avoid the cases of wrong capture in case that the sacrifice is correct. A better code may be for example to tell shredder to use more time if it plans to capture that kind of capture but not telling it that capture is always wrong. Uri
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.