Author: Pete R.
Date: 11:16:14 07/26/00
Go up one level in this thread
On July 26, 2000 at 13:14:18, blass uri wrote: >On July 26, 2000 at 12:52:48, Pete R. wrote: > >>On July 26, 2000 at 01:44:57, blass uri wrote: >> >>>I think that if you give GM's many position and ask them to evaluate what is the >>>expected result for white after 1 second and you give computers the same >>>positions and ask them to evaluate the expected result for white by translating >>>the number of their static evaluation function to the expected result then you >>>are going to find that the evaluation of computers is better because humans >>>cannnot calculate in one second all the things that computers calculate in one >>>evaluation. >> >>Well of course but this is a silly restriction. > >This is not a silly restriction because we are talking about comparing static >evaluation. > >If you give GM's to look at the position for a long time they will use search to >evaluate the position and learn from the search about the evaluation. > >Uri It may be impossible for humans to totally separate this process. By pattern recognition we may automatically see that one side could eventually create a passed pawn, etc., even without calculating any variations. But this is something an eval function should be able to do as well, i.e. see a pawn majority. Other examples may be more nebulous, but the point is, we want to have "accurate" assessments of the *positional* considerations of the position and who appears to stand better. This assumes the absence of immediate tactics. It will take longer than a second for a human to even count the material balance, let alone decide who has the favorable pawn structure, better minor pieces, better piece coordination, and all the other "positional" factors that you try to duplicate in an evaluation function. But if an eval function had equivalent knowledge to a human GM in this regard the program should play better than any human. The difficulty of course is deciding the relative importance of these factors based on the position, and this is where human intelligence is vastly superior. E.g. whether a bishop is bad depends on more than whether nearby pawns are on blocking squares of its color. A human can see when the rule applies and when other factors are more important.
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.