Author: Rolf Tueschen
Date: 05:57:33 06/19/04
This is a translation of an article I published in the German forum of hagra. [ Computerschach Stellungstest Schaubude ] At the end of a discussion that went also into international circles, I want to give a general and possibly widely understandable solution to the basic debate. This is out of solidarity for hagra and his fun forum and in contradiction to the CSS forum, where truth has not the chance and critics get censored and then banned. Hagra, who was also banned yesterday, please don't let you be demoralized. There is a problem, that a test lay can't understand without appropriate education: ** Why a test, like the CSS-WM-Test of Mikhail Gurevich, shouldn't bring something good, if it is clearly proven, that one gets exactly the same results, people do also get who are making tests with tournament games results? ** Why above all, all these programmers don't use this good test in their work to see very quickly , what a specific version of their creation can - objectively! - do and perform? These questions are so difficult, because to lays it must appear as if all the experts were dick-headed somehow. The test author MG mentioned that therefore all these people should simply do his test, because only then they could see how valuable and precise the CSS-WM-Test really is. - This is also the reason for this solution now! Solution ======== The whole debate, that has been fought, fierecly in special from the side of the CSS people, see M. Scheidl, can be reduced on a simple misunderstanding. To be able to understand one needs testtheoretical knowledge. Otherwise on dosn't find this (simple) solution. a) veritable tests are made because one doesn't know something; one has suspicions, but now the statistical proof must be found; validity and reliability... The programmers are in that situation a), when they are working on their babies. They don't know the answers on their questions. If they would do the WM-position test and trust the results, then surprises at the next tournament were programmed. Only therefore the programmers say, we can't use these position tests at all. b) people who run test-suites and also the test founders, in case of position tests at least, are testing end products, ready made and already made optimal for being used in tournaments and such test suites. :) Their strength is already optimized towards playing strength. Forcedly the testers get similar results like the tourament games comp vs comp testers. And by force they then think that it makes no difference if one does test playing strength or the analytical strength to solve these smartly chosen test positions. This alleged lack of differences which does surprise all the testers in b) is the reason for the misunderstandings in the so called WM-debate. Because the ready made final product can well be tested with these position tests, but - this is Bob Hyatt speaking - we only get what we, the experts, did already know about our creations. Nit at point zero theses position tests don't show us anything. Of course we would get results, but these results are not reliable, as Sandro Necchi, book author of Shredder, told us today here in CCC. To all the fans and users: the results of the position tests are ok, but empty for the experts. The expert can't profit because he already has optimized his product. But testing in its real meaning is meant for unknown results. Results that could be good or bad for your plans. The whole debate about the Elo numbers for such test results in the CSS-WM-test is a different question. But it's a fact that such Elos, however been presented, don't have a meaning or a power in itself. I hope that it's clear by now that all the computerchessfans can well run their test suites like before. They can enjoy the results. It makes likewise much fun to let one's own games be analysed by the programs. This is one of my little hobbies. To everyone what pleases ... Kind regards, Rolf
This page took 0.02 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.