Author: Jon Dart
Date: 07:58:44 12/17/01
Go up one level in this thread
On December 17, 2001 at 07:08:30, Uri Blass wrote: >I think that it may be a good idea to build a test suite based on games when the >target is simply to avoid tactical mistakes in the games. > >It is possible to take the ssdf games for this purpose. > >The program can analyze games at 1 minute per move and when it finds a big jump >in the score to the right direction of at least a pawn it can suspect that the >last move was a tactical mistake. > Sometimes the move immediately before the one with the big score is not a mistake - the side to move may be already lost, and have no good alternatives. But neither side sees this earlier, due to limitations in the depth of the search. So you really need to back up a few moves and analyze what happened in the positions leading up to the one that seems decisive. Another factor to consider is just how long you want to analyze, both for the discovery of possible tactical mistakes and the confirmation of them. Something that looks good (or bad) at depth 10 may look different at depth 15. You can't usually be 100% certain that you have a unique best move. --Jon
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.