Author: Uri Blass
Date: 01:37:11 04/28/04
Go up one level in this thread
On April 28, 2004 at 02:48:40, Tony Werten wrote: >On April 28, 2004 at 01:34:21, Jouni Uski wrote: > >>Aren't they totally unnecessary? Why not simple check every line BEFORE putting >>to book and remove all -5 lines! What's the problem? Of course You cannot trust >>any GM lines before checking. >> >>Or/and alternatively engine can analyze already in book and decline all bad >>lines by own move. Hmm. may be it's not so easily to implement... > >How deep do you want to search ? > >Too shallow and you'll miss all the nice positions where the computer thinks >it's bad, but the position is actually quite good. > >Too deep and it will take centuries. > >Maybe with -5 positions the problem is not so big, but with -2 it is. > >Tony It is possible to have a special small book of relevant positions that are good inspite of score of -2. The computer should do a small search of all the moves of the big book(searching only to depth 3) in order to detect moves with bad score(they are small minority) and the person who writes the book will need to decide by himself if to include them in a special small book. In game the computer may use 10% of its usual time to check if the move is not a blunder. Usually in good positions the score improves and does not get down so if you detect in a game bad score based on search to bigger depth and you did not detect the bad score in the first 3 plies the position is usually bad. 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.