Author: Uri Blass
Date: 12:30:47 10/12/01
Go up one level in this thread
On October 12, 2001 at 14:47:34, Robert Hyatt wrote: >On October 12, 2001 at 14:34:04, Uri Blass wrote: > >>On October 12, 2001 at 14:26:18, Uri Blass wrote: >> >>>On October 12, 2001 at 13:32:25, Robert Hyatt wrote: >>> >>>>On October 12, 2001 at 11:24:03, Daniel Clausen wrote: >>>> >>>>>Hi >>>>> >>>>>On October 12, 2001 at 10:23:07, Robert Hyatt wrote: >>>>> >>>>>[snip] >>>>> >>>>>>First try the Nolot test positions for 36 hours. Then report back with >>>>>>your "finding". :) >>>>> >>>>>And where exactly is the proof that the so-called solutions of these tests >>>>>really are the best? Or is it more like opening theory where the so-called >>>>>book moves just happen to be considered best/good by a large number of >>>>>people, whereas it's possible that white to move is lost from the beginning >>>>>(due to ZugZwang ;) and the best defense is a3? (maybe followed by h6! hee) >>>>> >>>>>Sargon >>>> >>>> >>>>Some of them are provably correct (almost all in fact). But they take >>>>far longer than 36 hours. >>>> >>>>If you really think 36 hours will solve _any_ problem, then you must think >>>>that computers are unbeatable at correspondence chess? They are _far_ from >>>>it in fact... >>> >>>I do not think that 36 hours are enough for every problem but beating a program >>>at correspondence game is not going to prove the question of the original poster >>>because he asked for as position that no program can solve >> >>It should be of course no top program because it is easy to build a lot of >>programs when always one of them will solve the position that you ask. >> >>You only need to write about 200 program by the following idea if you do not ask >>for top programs >> >>The first program is build to >>play the first move in the list of the legal moves. >> >>The second program is build to play the second move in the list of the legal >>moves or the first move if there is only onelegal move. >> >>The 50th program always play the 50th move in the move list or the first move if >>there are less than 50 legal moves. >> >>I think that the shirov Bh3's sacrifice is a good example for a position that no >>top program of today can solve in 36 hours. >> >>Uri > > >That is a good one, yes. Another interesting set of positions would be >the "key" positions in the old Mike Valvo vs Deep Thought two-game match >played on r.g.c several years ago. There were some positions (one where he >offered a pawn) that I don't think _any_ program will decline, yet it led to >a very deep forced loss. Here I believe that there are chances that at least one of the top program is going to find the right move after 36 hours. There are a number of such positions. I tried the >Bh3 position for a while and managed to make crafty play it by fiddling with >the evaluation. But The resulting "settings" were way beyond non-reasonable >and were only interesting to show how far a program had to be "warped" to think >that Bh3 was winning when the program couldn't see the win tactically (it was >a tactical position of course, and I eventually made crafty solve it >positionally, which was not very interesting for long-range chess skill). Changing the evaluation of the top programs to solve problems is not what is meant. The question is about positions when no top program of today can solve and I guess that after you change the evaluation to find Bh3 Crafty is clearly weaker than the original crafty and stops to be a top program. 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.