Author: Sune Fischer
Date: 01:37:04 11/28/02
Go up one level in this thread
On November 28, 2002 at 04:30:15, Uri Blass wrote: >On November 28, 2002 at 03:49:28, Sune Fischer wrote: > >>On November 28, 2002 at 03:43:14, scott farrell wrote: >> >>>On November 27, 2002 at 14:02:29, Stan Arts wrote: >>> >>>>Hello! >>>> >>>>I am wondering how strong you can go by not using pruning methods like null-move >>>>or futilitypruning etc. but basicly just use a minimax-type algorithm with >>>>some form of alpha-beta pruning. I mean to keep the "full-width" (depth before >>>>going into Q-search) depth 100% sound. >>>> >>>>How well would the strong programs of today do when they dis-able (in theory, >>>>might not be possible ofcourse) all types of pruning except alpha-beta? (keeping >>>>their Q-search and move-ordening as it is now) >>>> >>>>Would be interesting to find out, I mean to see the effect of very good move- >>>>ordening of advanced programs etc. Also it would probably show which programs >>>>depend most heavelly on their Q-search that probably makes a lot of difference >>>>in strength in this case too? >>>> >>>>PS. Are there any strong engines/programs that do NOT use null-move, or a whole >>>>lot of futility-pruning/razoring etc? And if so, which ones? >>>> >>>>Thanks! >>>> >>>>Stan >>> >>>Read my recent post titled "pruning vs extensions vs qsearch - are these all >>>effectively the same?" >>> >>>My argument is that you dont need to forward prune at all, if you extend enough, >>>it is virtually the same thing. >>> >>>Are you aware that deep blue did NOT forward prune OR null move ??? just lots >>>and lots of brute force and extensions. >> >>I don't think it is the same entirely. >>One difference is that with pruning you can "undo" the extension you did >>earlier. Ie. in a long variant you may extend at ply 2, but at ply 3 there are >>some moves with should be pruned rather than extended. >> >>I see no way of doing this only with (positive) extensions. >> >>-S. > >You can do it by not extending at ply 2 but remembering a flag and extend at ply >3 only if the rules of the original program tell you not to prune at ply 3. The book keeping would be enormous, but "theoreticly"... >I think that the pilosopical point that you can avoid pruning and do only >extensions is not important. > >I do not need to look for some strange ways to define extensions only to avoid >prunings. > >Uri Yes exactly, it is a duality principle, why not take advantage of that :) -s
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.