Author: Uri Blass
Date: 15:20:10 05/11/04
Go up one level in this thread
On May 11, 2004 at 18:12:48, Dann Corbit wrote: >On May 11, 2004 at 03:01:10, Uri Blass wrote: >[snip] >>I do not understand this evaluation because it seems to contradict the pv. >> >>List plans to allow Crafty repetition of the root position if you change >>Crafty's Nf6e4 that is written in the pv to Qc7b7 but still evaluates the >>position as positive for itself. >> >>I think that the evaluation should be at most 0.00 if the opponent can force >>repetition of the root position against your plan. > >Beyond the first few plies, a pv is a wild guess. Beyond the stated search >depth, a pv is wild speculation (often based on SEE). It is not unusual to see >a pv by a top chess program have a checkmate sign in it, and have the score >nowhere close to a checkmate. It can never happen with movei unless I have a bug. checkmate is always evaluated as checkmate in movei. In the relevant case of List-Crafty the repetition can happen in the 4th ply of the pv and it is not a horizon effect. It is possible that list read the pv from the hash and it is wrong but it is clear that 4 plies should be enough to see a draw for crafty in that line so the fact that list has positive score in the game is clearly wrong. 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.