Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Selectivity

Author: Robert Hyatt

Date: 11:58:15 01/14/01

Go up one level in this thread


On January 14, 2001 at 13:57:14, Uri Blass wrote:

>On January 14, 2001 at 13:12:38, Christophe Theron wrote:
>
>>On January 14, 2001 at 00:33:56, Robert Hyatt wrote:
>>
>>>On January 13, 2001 at 23:58:41, Christophe Theron wrote:
>>>
>>>>On January 13, 2001 at 15:03:01, Uri Blass wrote:
>>>>
>>>>>On January 13, 2001 at 13:38:48, Christophe Theron wrote:
>>>>>
>>>>>>On January 13, 2001 at 03:17:25, Uri Blass wrote:
>>>>>>
>>>>>>>On January 12, 2001 at 23:05:23, Christophe Theron wrote:
>>>>>>><snipped>
>>>>>>>>However I can tell you in advance that you will (probably) discover that Chess
>>>>>>>>Tiger and Gambit Tiger are extremely selective. And the next versions will be
>>>>>>>>even more selective.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>    Christophe
>>>>>>>
>>>>>>>
>>>>>>>Does it mean that it is not going to find a4 at smaller depth?
>>>>>>>
>>>>>>>Uri
>>>>>>
>>>>>>
>>>>>>
>>>>>>I don't care about finding a4. I don't care about solving test suites faster.
>>>>>>
>>>>>>I do care about playing stronger, and it's a different story.
>>>>>>
>>>>>>
>>>>>>
>>>>>>    Christophe
>>>>>
>>>>>Basically the stories are similiar and in most of the cases the better program
>>>>>in playing games is also better in solving test suites.
>>>>>
>>>>>Uri
>>>>
>>>>
>>>>I think you are right. But if my program gets better at solving test suites, it
>>>>will be because I'm trying to make it stronger in real games.
>>>>
>>>>That's why I don't care about finding a4 in this position.
>>>>
>>>>
>>>>
>>>>    Christophe
>>>
>>>
>>>This is a point I have made many times...  "cause" and "effect".  Is doing
>>>better on test suites an effect of playing better?  (I think so).  Or is
>>>playing better a result of doing better on test suites?
>>>
>>>There are other similar cause and effect questions about other things like
>>>space, mobility, etc...
>>
>>
>>
>>In the case in question I agree with you that doing better on test suites is a
>>by product of playing better overall.
>>
>>I have already noticed it, very clearly, with my own programs.
>>
>>I have never noticed that it worked in the opposite direction, and I can even
>>say that I have noticed that I could easily weaken my program by trying to tune
>>it to test suites.
>
>The question is if you cannot improve your program by using the weaker engine
>that you generate only in part of the cases because it is clear that the weaker
>engine is sometimes better(otherwise it could not be better in test suites).
>
>Uri

A classic example is "chest".  Great at finding mates.  But try to wrap a
front-end around it and play real games.  It will get totally mashed.  Because
it is designed to find tactical things and not positional things.



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.