Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: About testing and tests suits

Author: Uri Blass

Date: 08:11:00 12/21/02

Go up one level in this thread


On December 21, 2002 at 10:54:58, Rolf Tueschen wrote:

>On December 21, 2002 at 10:10:38, Uri Blass wrote:
>
>>On December 21, 2002 at 09:58:10, Rolf Tueschen wrote:
>>
>>>This time no murdering intended, assured!
>>>
>>>Let me please ask you a couple of very basic questions. Perhaps they are so
>>>basic that nobody really thought about it before.
>>>
>>>
>>>In a little thread about the question if Fritz 8 is weker than Fritz 7 I read
>>>certain remarks without further explanation and I was really surprised.
>>>
>>>Uri is asking Prof. Irazoqui, if he had already tested so and so in his "secret
>>>test suite".
>>>
>>>What is a secret test suite??
>>
>>This is a tactical test suite but the positions are not known.
>>The last fact is important because when the positions are known
>>programmers may be influenced by the positions in implementing tactical
>>knowledge.
>>
>>Enrique found that in most of the cases result in his secret tactical test suite
>>gave good prediction for the results in the ssdf list.
>>
>>I guess that the reason is that in most of the cases rhere are no big difference
>>in positional strength between programs and the program that is better in
>>tactics is also the better program.
>>
>>>
>>>Then the addressed answered I'll be glad to "test" your Movei.
>>>
>>>What is meant with "I'll test"??
>>>
>>>Further.
>>
>>It means that he will give it a lot of positions to solve and use the times of
>>the solutions to calculate a score for it.
>>
>>Uri
>
>Tanks, but what is a "score"? I hope that you don't have the feeling as if I
>wanted to play games here. I swear that I ask questions with a very reasonable
>and basic importance. I hope you will believe me.

the score is an elo number that is calculated based on the times to solve the
problems.

Enrique did not publish the formula that is used.

>
>And IF the programmers had the possibility to "cheat" when they know the
>positions, what could a "tester" do in his position?

If they know the positions then the positions need to be changed.
The positions are basically positions when there is only one best move and
programs cannot find the moves for the wrong reason.

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.