Author: Vincent Diepeveen
Date: 15:17:16 04/22/01
Go up one level in this thread
On April 21, 2001 at 22:01:21, Robert Hyatt wrote: >On April 21, 2001 at 16:37:15, Alex Boby wrote: > >> >>My PV search uses the assumption that when the minimal window search fails high, >>then the score returned is at least as high as the true score which will be >>returned when you search again with a wider a/b window. >> >>All was fine and dandy before hashing. Now I'm experiencing problems where this >>assumption is incorrect because the score from the minimal window search was >>returned via a hash table hit which was overwritten before the research takes >>place. So the research doesn't return a score >= min window score. >> >>And so I'm unsure about how to deal with the PV search now when the base >>assumption for that technique is invalid. >> >>Any suggestions are greatly appreciated :) >> >>Alex Boby > >You can get search inconsistencies two ways... I see at least 3 ways :) >1. A key hash entry gets overwritten at a bad time, so that a fail high >search can't resolve the true score. > >2. You store an upper bound and it causes a fail high, but then when you >re-search you can't actually see the reason for the fail high due to depth >reasons or something... so you "smell" that the move is good, but your search >isn't able to prove how much better it is. 3. Bug in program >Both are normal and unsolvable problems... 3 is not abnormal. Happens here way more as 1 and 2 :)
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.