Author: Robert Hyatt
Date: 18:49:45 08/04/05
Go up one level in this thread
On August 04, 2005 at 16:00:01, Engin Üstün wrote: >i have a question to Robert Hyatt too. > >some days ago i read in your crafty source about "singular.c" to extend a move >if this scores better then all other moves, i haved experemented with this >singular extension in my program Tornado, but i don't see any changes or efects >or any speed up with this extension and remove it from my source. Not sure what you are reading. singular.c has not been a part of any normal distribution, as I have never found satisfactory settings that made it work effectively. Also the "singular.c" approach is a real "poor-man's" singular extension approach, far removed from what Hsu/Campbell described in their JICCA paper on the topic... > >The search down more and more if the engine searching deeper. >Why using this idea, if this cost more time in search and get nothing for this. >if we know that a move is better then other ones, we don't searching any other >moves. > >And i don't see in other sources about singular using this idea. the simple idea is to see if one move is better than all moves by some margin "s". If the answer is "yes" then you search that move a ply deeper. Proving that one move is better is both expensive in terms of search time required, and tricky... > >The other question is, why you don't using in crafty the history pruning, this >brings in search near 3-4 ply deeper in my program. > >The idea i have taken a little from Fruit :) and implemented with my own ideas. I really have not seen any info on "history pruning". I'll try to see what I can find unless you can provide an explanation...
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.