Author: Uri Blass
Date: 04:35:55 09/20/01
Go up one level in this thread
On September 20, 2001 at 07:10:56, Gian-Carlo Pascutto wrote: >On September 19, 2001 at 18:05:43, Joshua Lee wrote: > >>Nice try at a test set but i will let you know which ECM problems are not solved >>and you should use those. Some have errors like ECM 390 ...i will post what i >>find in a few months > >I expressly want problems that _can_ be solved with varying difficulty. > >A testset isn't going to get help in debugging and regression testing >if you go from 0 solved to 0 solved. > >The idea isn't to make something to compare programs on. Testsets >are useless for that. I wan't something to test improvements on. > >-- >GCP The main problem is that it is possible that an improvement in solving test suites is counter productive in games. I think that it is better to use a positional test suite based on games and not a tactical test suite to test improvements. I can give few examples of cases from my correspondence games when I am almost sure that a positional move is the best move. Note that Be2 has a tactical idea but the difference between the evaluation of Be2 and the next best move is a positional difference from computer's point of view. r2qkb1r/5pp1/p3p2p/4n3/1p1NP1P1/4B3/PPPQ3P/2KRR3 b kq - 0 1 bm Bc5 2r2rk1/p3q1pp/bp6/2bp1p2/4n3/1P2P1P1/PB3PBP/1QRR1NK1 b - - 0 1 bm Be2 3r1r1k/p4qpp/8/2p1Rp2/4n3/1P2PPP1/PB5P/4QBK1 b - -bm Rd2 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.