Author: Uri Blass
Date: 23:53:32 08/11/03
Go up one level in this thread
On August 12, 2003 at 02:48:48, Uri Blass wrote: >On August 12, 2003 at 01:08:47, Anson T J wrote: > >>>Not only SSDF do that. Companies are selling incomplete CDs sets, users download >>>only some TBs (I doubt many will download e.g.. kqqkb), etc. That is common >>>practice. >>> >>>If program cannot properly handle such practice I'd call that program's fault, >>>especially when ways to fix such a behavior are well-understood and discussed >>>here (and in other forums) lot of times. Such program will lost half a point >>>here and there, and in *my* opinion that's fair. >>> >>>If programmer still don't want to include proper fix, (s)he can modify the >>>program, so (for example) it will not start with incomplete set, or will not use >>>it, or will give large warning on startup, etc. >>> >>>BTW, those fixes will also lower product support costs... >>> >>>Thanks, >>>Eugene >> >>I believe the Chessbase Fritz GUI warns you if you have an incomplete set of >>tablebases and requires confirmation before proceding with a tournament game. I >>don't think its fair to penalise an engine because the tester set it up >>incorrectly. >> >>I can see why an incomplete set would cause problems but not how to solve it >>easily. If a position yields a score of mate in 8 due to tablebases, but the >>mate requires a transposition into a tablebase set that isn't present, the >>program will be choosing between a mate in 8 and a score of say 9.00 if it >>hasn't yet found the mate for itself. >> >>From Dr. Hyatts posts, I gather Crafty handles incomplete TBs, but how? > >If the score does not go down from mate in X to mate in X-1 then the engine can >stop using the tablebases, forget everything and consider only moves that lead >to positions that are not in the tablebases set that it has. > >Uri Note that the last part is not needed and it may be possible that the score does not go down because of qsearch that leads to non tablebase position so there is no direct move that leads to it. I believe that crafty simply stop using tablebases if the distance to mate is not reduced(it can be reduced by more than 1 because of mistake of the opponent but if it is not reduced then it is better to stop using the tablebases). 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.