Author: Uri Blass
Date: 14:19:51 11/09/04
Go up one level in this thread
On November 09, 2004 at 11:33:08, chandler yergin wrote: >I think we are talking at cross purposes here Gentlemen. > >As a Player, not a Programmer, I use Chess engines as a 'Tool', >to learn about a position. >The Analysis Module can generate a "Variation" >OR, many Main Lines, Simultaneously. >You, are 'locking' it to a single 'variation'. >As such, it will stay with that variation, and go as deeply as the time you >allow, or until it recognizes it is not as good as another line. > >You imply, and IMO mislead, by stating that it doesn't 'see' or find >another move, or may never find a certain move. >This is not correct! Ed is not misleading The subject here is testing engines and not the use of them by chess players. You are misleading by claiming that the engine find a move when it is using previous learning. The engine is slower when you give it to search some lines so with some lines obviously it cannot find the correct move as best faster. The engine always search also other lines even when you tell it only to find the best move but it simply does not calculate exact evaluation for them and it is enough for it to know that they are inferior so in the default mode it shows you only a single line. > >The Engine is searching many of the most promising "Candidate" moves >simultaneously. No The engine does not search them simulataneously. You can see simultaneously many lines but they were not searched at the same time and one line may be searched 5 seconds ago when another line was searched 5.7 seconds ago. > This is exactly what a Player wants to know. No the question that a player wants to know is if the engine can find similiar moves in games when it has no previous learning. The question if you can convince the engine that the move is correct by analysis is a different question and this was not the question that was asked. >Increase the lines to 3, or 4 and you'll see the Best continuations, and how >the evals change with time. By this definition every engine that support multi variation can find every move because you only need to increase the number of lines to the number of legal moves in the worst case. > >It matters not a whit if Crafty or Shredder when locked to a single 'variation' >takes hours to search to a Depth of 16 ply, and then changes it's mind, >when by Opening it to 3 Main Lines, you see that it DOES find and eval >11. h5 in less than 2 minutes, then promotes it to the PV. I do not believe that it can find 11.h5 as best in less than 2 minutes when it has 3 lines. With 3 lines it should be slower relative to the case that it has 1 line(otherwise it could be better for it to use 3 lines and choose the best also in regular game) > >Now what is important here? >To sit and wait for hours, or search and find the most promising >Candidate moves almost immediately? We are not discussing about the ability of the team human+computer to find the best move but only about the ability of the computer with no help to find the best move. >Does it make any sense to 'test' an engines response time, or to gain >information about the position? > >Hmmmmm? >CY If the question is which engine helps to get better information about the position then the question is different but this was not the question that Ed asked. 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.