Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Odyssey2001-tournament: round 10 over, pairings round 11, games

Author: Uri Blass

Date: 07:44:17 12/09/01

Go up one level in this thread


On December 09, 2001 at 10:32:20, Thorsten Czub wrote:

>On December 09, 2001 at 08:52:59, Mogens Larsen wrote:
>
>>I don't have any prejudices.
>
>any human beeing has prejudices. only you don't tell them in public.
>nobody can live and judge anything without having prejudices. only
>god has no prejudices.
>
>>That's the advantage of fact over opinion.
>
>the fact that you talk about games you haven't seen.
>
>>Well, it's a fair assumption that no less than two new versions within a short
>>timespan indicate some sort of dissatisfaction. And it's a certainty that you're
>>unable to evaluate strength by your methods. So a crystal ball would be more
>>effective.
>
>when my methods are wrong, why was i succesful in finding out that hiarcs would
>be world chess champion BEFORE the championship was played out.
>when it is wrong, why was i able to see that chess tiger is very very strong,
>without even the programmer (christophe) knowing and believing it.
>when my method is wrong, how can i see in forward that century is very strong,
>and century makes 2nd rank, and how can i find out about cstal ?
>and about nimzo?
>all i see is that you claim that you do not replay games, hear about strength
>from hearsaid, and call other methods lousy. when in fact you don't know about
>the programs.
>
>instead of finding out yourself by researching, you call other people idiots and
>their methods wrong.
>
>of course it is nonsense to test version x versus version x+1. this does not
>lead you to a better chessprogram.

I disagree here
I think that testing version x against x+1 is one of the tests that can be
productive.
I agree that it is not the only test but if x+1 is clearly better than x then I
expect it also to beat x.

The question is also what is the change between x+1 and x.
There are cases when comparing x+1 and x is enough.

If you find a way to do x+1 10% faster than x with no change in the evaluation
function or the search rules then comparing x with x+1 only to verify that you
have no new bugs is enough.

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.