Author: Dann Corbit
Date: 13:01:57 03/07/06
Go up one level in this thread
On March 07, 2006 at 15:42:29, Robert Hyatt wrote: >A note. This is not a good way to test programs that use something like the >late-move reductions. The trees may well be searched in a different order when >you flip or flop the board from side to side, since for a bitboard program, >FirstOne() will find a different bit set, which means moves for pieces will >likely pop out in a different order. And that affects the history values enough >that you could end up pruning in one position but not in another, depending on >how late or early a key move is searched. Fruit uses late move reductions, and yet the search is perfectly symmetrical. >This is a good approach for pure static evaluation testing, but once you get >into the dynamics of the search tree, particularly with forward pruning (or >reductions, however you think about it) you can't search the same tree easily >since the pieces are on different squares, and so the final result can vary. >Trying to debug that is basically chasing your tail... I agree that static evalution is the way to be really sure that you are chasing the right problem.
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.