Author: blass uri
Date: 04:11:30 12/13/99
Go up one level in this thread
On December 13, 1999 at 05:52:59, Enrique Irazoqui wrote: >On December 12, 1999 at 18:44:00, Amir Ban wrote: > >>There are no two schools of thought. That's the fallacy. There are tradeoffs >>everywhere, and every programmer tries to strike the right balance. The better >>ones manage to find the right compromises, or better yet, to find ingenious ways >>to avoid making compromises. > >By reading and listening to programmers it seems clear that there are indeed >these 2 schools of thought: speed vs. knowledge, optimizing for comp-comp vs. >optimizing for human-comp. Ed posted several times about this, including in the >last few days. Others did too. > >>Do I understand that your secret test suite is a tactical one ? If so, I don't >>believe in it. > >And nevertheless it worked and the test believes in you. :) The rating of J5 in >my test was within 5 points of the rating given by the SSDF. I do not believe that a test of only finding the right tactical move can be a good test for all the programs. I thought that there are tactical positions but also some positional tests. positional is of course tactics if you see deep enough but I saw cases when computers do mistake of taking pawns because they are too greedy and no program can prove that taking the pawn is a mistake by tactics so in this case avoiding taking the pawn is a positional test. Junior5.9 is better in tests like this because the value of pawn of Junior5 was reduced. Junior5.9 can avoid the famous mistake of zungzwang against a human Qxg7 after some minutes. It does not understand that Qxg7 is losing and give it a positive score but prefer another queen move. Another example for positional test is Junior's sacrifice against nimzo from wccc. It is a positional test because I believe that no program can see by tactics more than 3 pawns for the piece(it is possible to prove by a tree that there is more than 3 pawns but programs(at least programs that I know do not know to build the right tree by themselves). Junior5 can solve this positional test. Junior5.9 cannot solve it because in this case reducing the value of pawn was counter productive. 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.