Author: Robert Hyatt
Date: 09:44:32 05/18/00
Go up one level in this thread
On May 18, 2000 at 06:58:44, Jan Pernicka wrote: > Hi, >I have to face this problem: > When using aspiration search at the root with window <eval-100,eval+100> I >occasionally get fail high - for example when I have found a mate. Problem >is when the returned score (let's call it S)is actually the exact value of my >position - in this case I do a research with window <S,+infinity> but >I'm unable to set PV properly at that ply - and if it's last ply of the search >PV is wrong... (e.g. showing PV not leading to mate even though returned score >says it will be mate...). > I set the PV in the tree when: > 1) I got cutoff > 2) eval()>alfa >Thank you in advance for your suggestions. > Jan You might be missing one place. Normally you fill in the PV right where you call evaluate... as that score goes with the current pathway of moves from the root to the current ply. With mate, you have a special case. You have an exact evaluation. Where you assign the value of MATE-ply to the current score, you also have to take the current path of moves from the root to the previous ply and stuff them into the PV. Then back up the score _and_ the PV as normal.
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.