Author: Matthew Herman
Date: 05:18:03 03/25/99
Go up one level in this thread
On March 24, 1999 at 14:33:34, Dann Corbit wrote: >On March 24, 1999 at 08:56:14, Matthew Herman wrote: > >>I have been doing some research (i.e. comp vs comp games at high time controls) >>and I have seen that there ARE some openings that are TACTICAL, yet very bad for >>computers (this relates to dann corbit's posts). For instance the Velirimovic >>attack, white will often sacrifice a N on d5 or f5 to remove the valuable e6 >>pawn (without which the kingside usually crumbles), yet computers snap this up >>instantly. Also in some variations of the yugoslav, computers don't see white's >>powerful attack before it is too late. Also if playing against computers w/ out >>books, sometimes the dragon (as black) is strong, as without the books some >>programs DON't find the h4-h5 idea. >> >>This is very relevant material, maybe it will be knowledge added to programs >>that will solve this, maybe not. >> >>Still computers dont "know" the opening, so they don't play it as well as the >>normal GM would (in some cases). >Could you post some EPD positions where computers make the wrong choice? I am >*very* interested in this. 1st position : velirimovic attack r1b2rk1/2q1bppp/p1npp3/1pnB1N11/4P1P1/2N1B3/PPP1QP1P/2KR2R w - - here computer will take on d5 or f5 (both lose). 2nd position : Yugoslav Dragon 3q1rk1/pp1bppbp/3p1np1/8/2rNP1P1/2N1BP2/PPPQ3P/1K1R3R b - - Here most computers play the slow a6, b5 and white should develop a strong attack with h4-h5, yet programs like junior after a6 play qd3?!. Only Fritz has found h4-h5 so far.. I havent tested crafty yet. There are other positions like this, but these are the 2 main ones I am looking at.
This page took 0.01 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.