Author: Brian Richardson
Date: 10:00:01 05/19/00
Go up one level in this thread
On May 18, 2000 at 18:26:05, Jorge Pichard wrote: >In the game Tiviakov vs Fritz SSS we witnessed how simple it was for any IM to >paralize the tactical capability of a any good program. Since we all know >that open positions increase the tactical chances of most programs. Therefore, >it is recommended to implement a subroutine which will enable the program to >detect when the position is becoming locked in the middlegame, which will >trigger it to sacrifice the key pawn in order to open up the position for the >program to excel tactically. > >Jorge I had tried several tests to avoid blocked pawn positions/files against human opponents (but not piece mobility) in Tinker during April. Its ratings became steadily worse on ICC. This may have been due to the mix of eval function and other changes, or better competition, or both. I have backed out these changes and its ratings have improved somewhat. My best guess at this point is that without pawn hash tables, the extra evaluation terms slowed the search down more than the benefit of the extra "knowledge". I am planning to add pawn hashing next and will experiment again with the extra terms when they don't cost as much.
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.