Author: Uri Blass
Date: 14:56:59 06/22/04
Go up one level in this thread
On June 22, 2004 at 14:08:57, Rolf Tueschen wrote: >a) WM Test of CSS is not a *test* > >b) running 260 programs through these 100 positions is not *testing* > >What you report here is playing with chess engines in chess positions. Just >because you start it, look at displays and report the results this can't be >called *testing* I agree that it is not testing the strength of the engines and the evidence is that gambit shredder that is not the best shredder for games is the best shredder in the list. I do not think that it is not testing and it is testing how the engines perform in specific positions. Getting information about positions that the program did not solve can be productive for programmers to improve the program by thinking if there is some missing knowledge that caused the program not to solve the position and trying to implemnt that knowledge in a productive way(I am not talking about tuning for test suite that is done by giving some knowledge that is productive for the specific positions but counter productive for other positions and I am talking about fixing holes that prevent the program to solve some positions without doing demage). Note that one of the ways that I use test suites is by thinking what is the reason that my program did not solve some position or does not solve it faster and thinking if I can fix it in a productive way that does not hurt playing in other positions but help in the specific position. I have enough problems in other test suites and I do not think to do it for this test suite in the near future but results of test suite clearly can be productive and I am not talking about number of solutions but about the question what positions the program did not solve and what is the reason for it. 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.