Author: Chessfun
Date: 11:55:57 10/18/00
Go up one level in this thread
On October 18, 2000 at 11:02:08, Uri Blass wrote: >On October 18, 2000 at 10:54:22, Tony Hedlund wrote: > >>On October 18, 2000 at 05:51:55, Uri Blass wrote: >> >>>On October 18, 2000 at 05:33:06, Thorsten Czub wrote: >>> >>>>:-))) >>>> >>>>don´t get a heart-attache steen. I am sure they will get it right. >>>> >>>>sooner or later. or maybe later... >>>> >>>>maybe next year. :-))) the most important thing is of course that the chessbase >>>>programs run accurate :-)) >>> >>>My opinion is that it should be the responsibility of the testers to stop the >>>match when they suspect that the learning of a program that is supposed to have >>>a learning function is destroyed and to ask the programmer for the reason for >>>it. >>> >>>It is not interesting to continue matches when you see that the learning is >>>destoyed and if it is possible to test the program without this bug then it is >>>better to do it. >>> >>>I do not suggest to stop matches after seeing bugs that are not learning bugs >>>because this kind of bug is different. >>> >>>People may complain that tiger12 also lost the same opening against Junior many >>>times in a row but this case was different because I understood that the reason >>>was not a bug but a bad design decision. >>> >>>Uri >> >>And how was I supposed to now that it was a bug and not a bad design decision? >> >>Tony > >You can ask the programmer. > >In the case of tiger the programmer admitted that it was a bad design decision. >Tiger learned but the learning did not help it to avoid the same opening and >changing the moves later did not help. > >In the case of gandalf it is clear that it is a bug because the learning is not >destroyed when you do not stop the games and it is not logical to have a design >decision to destoy the learning file. > >Uri I think myself it's a bit much to expect the SSDF to have noticed anything wrong only a couple of games were played. Some programs may learn slower than others and that assumes that they even know in advance book learning exists in a given program. As is known ChessTiger 12.0e didn't have it neither did CM6K nor will CM8K. I understand that the operator needs to know as much as possible about the program but don't expect the SSDF to trouble shoot programs. Sarah.
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.