Author: Uri Zlatnik
Date: 14:06:51 12/25/04
Go up one level in this thread
On December 25, 2004 at 00:24:31, Uri Blass wrote: >On December 24, 2004 at 18:25:40, Uri Zlatnik wrote: > >>On December 24, 2004 at 14:17:36, Uri Blass wrote: >> >>>On December 24, 2004 at 12:52:55, Pallav Nawani wrote: >>> >>>>On December 23, 2004 at 14:03:34, Uri Blass wrote: >>>> >>>>>On December 23, 2004 at 13:12:45, Pallav Nawani wrote: >>>>> >>>>>>On December 23, 2004 at 06:25:20, Uri Zlatnik wrote: >>>>>> >>>>>>>How is the search extension of a software improved? Does this have to do with >>>>>>>alpha-beta pruning methods? >>>>>> >>>>>>I am not sure what you mean here, so I'll pass this one. >>>>>> >>>>>>>For example CM 10th edition (the King engine) has longer search extension than >>>>>>>Fritz 8 but it doesn't always find the best move. >>>>>> >>>>>>Search extensions are mainy good for tactical moves. If a positional move is the >>>>>>best one, then evaluation is also of importance. Also, no engine can always find >>>>>>the best move! >>>>> >>>>>The position that Uri Zlatnik posted was only about tactics and not about >>>>>positional moves. >>>>> >>>>>Uri >>>> >>>>Well, maybe he was talking about some tactical position in some other thread, >>>>but I haven't read that one, and he did not specify any position here. >>> >>>He talked about tactical position in the following thread >>> >>>http://www.talkchess.com/forums/1/message.html?402041 >>> >>>He claimed that it is a positional problem but when you look at the position it >>>is clearly about tactics and not about positional evaluation. >>> >>>Uri >> >>True the variant is mostly tactical but some basic understanding of the pawn >>structures is also required for this. >> >>The problem with chess softwares is that they do not play a move until the full >>variant is calculated. The problem is that some variants might even be as long >>as 15 moves (30 plies). > >30 plies is no problem for programs with the right extensions. >The problem is not having the right extensions. > >I guess singular extensions to help here. >Deep Sjeng has singular extensions so it may be interesting to try it. > >Uri There is so much of a difference between how humans and chess softwares think. The average human brain is estimated to perform 10,000 trillion operations per second. The problem with search extension is that you can't tell if a move is going to be good but you can certainly tell if it looks as if it might become interesting. We can only guess and hope that we don't overlook something. This is the whole principle of search extensions. The danger if used carelessly is an explosion in tree size.
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.