Author: James T. Walker
Date: 17:05:29 01/25/06
Go up one level in this thread
On January 25, 2006 at 19:22:20, Uri Blass wrote: >On January 25, 2006 at 19:07:51, James T. Walker wrote: > >>On January 25, 2006 at 13:03:02, Uri Blass wrote: >> >>>On January 25, 2006 at 12:37:30, Roger Brown wrote: >>> >>>>>Hello Uri, >>>>>You are playing word games. Vasik is trying to be a "purest". I'm sure you can >>>>>add "knowledge" which if not implemented correctly could lower a programs >>>>>rating. I'm pretty sure Vasik will not do that. >>>>>What would you call endgame knowledge that does not win games but only allows >>>>>the program to draw instead of losing??? (food for thought) >>>>>Jim >>>> >>>> >>>> >>>>Hello James, >>>> >>>>By any definition, endgame knowledge that allows an engine to obtain a draw >>>>instead of a loss must add to a program's rating. >>>> >>>>Or am I missing something here? >>>> >>>>Later. >>> >>>The point is that vasik in the readme wrote the following words (that I >>>mentioned in the post that James replied to): >>> >>>"chess knowledge wins chess games. If it doesn't, it isn't knowledge." >>> >>>I guess that he meant that it increase the result of the engine but if we look >>>at what he said and not in what he meant then his words mean that only drawing >>>games instead of losing games is not knowledge. >>> >>>Note that I do not believe that there is knowledge in evaluation that can help >>>only to draw games instead of losing games because if a program knows by >>>evaluation that some position is a draw it can help it to avoid the draw in case >>>that it has better position and if a program knows by evaluation that some >>>position is a win for one side then it can help it to get it and not choose an >>>alternative that is not clear. >>> >>>Uri >> >>Uri you are starting to post nonsense. Are you saying that there are no >>positions where the best move is only a draw? I know you know better. For >>instance there are positions where a kqkp are a draw. If a computer has this >>knowledge it can trade into this position instead of playing into a loss. The >>bishop pawn comes to mind. To my way of thinking this is knowledge that will >>help save a draw instead of losing. Do you not believe this? >>Jim > >If a computer has not the knowledge of drawn KQ vs KP then it can let the >opponent escape to this draw from a lost position so it is going to have less >wins relative to the case of having the knowledge. > >Uri Suppose the "knowledge" is implemented incorrectly. Suppose it can save the draw but not the win. Suppose it works only when it has black and not when it has white. Are these not possibilities? Jim
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.