Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Evaluation Autotuning

Author: Uri Blass

Date: 13:49:14 06/29/04

Go up one level in this thread


On June 29, 2004 at 16:10:14, Vincent Diepeveen wrote:

>On June 29, 2004 at 15:53:26, Uri Blass wrote:
>
>>On June 29, 2004 at 15:34:31, Vincent Diepeveen wrote:
>>
>>>On June 29, 2004 at 13:51:51, Uri Blass wrote:
>>>
>>>>On June 29, 2004 at 13:13:19, Dan Honeycutt wrote:
>>>>
>>>>>On June 29, 2004 at 12:34:40, Vincent Diepeveen wrote:
>>>>>
>>>>>>On June 28, 2004 at 18:16:39, Dan Honeycutt wrote:
>>>>>>
>>>>>>>On June 28, 2004 at 16:44:33, Vincent Diepeveen wrote:
>>>>>>>
>>>>>>>>On June 28, 2004 at 12:37:42, Dan Honeycutt wrote:
>>>>>>>>
>>>>>>>>>On June 28, 2004 at 08:54:00, Anthony Cozzie wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>>settings, and then N games with the new settings.  I am only really interested
>>>>>>>>>>in longer timecontrols: 20 min + on an Athlon 2.0G or so (70 min on P-650, etc),
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>Why long time controls?  I thought you could test evaluation with shorter time
>>>>>>>>>controls, search needed longer (or varied) time controls.  Am I out in left
>>>>>>>>>field?
>>>>>>>>>
>>>>>>>>>Dan H.
>>>>>>>>
>>>>>>>>Do you believe the commercial authors at home only test at 1 0 level?
>>>>>>>
>>>>>>>No, but their situation is different than mine.  They have mature engines with
>>>>>>>evaluation values that are somewhere near right.  I've got a green engine with
>>>>>>>evaluation values that are junk.  I'm looking at short time control tests to get
>>>>>>>me headed in the right direction.  Got to walk before you run. :)
>>>>>>>
>>>>>>>Dan H.
>>>>>>
>>>>>>Why do you suggest it works for crafty to play 1 0?
>>>>>
>>>>>I never mentioned crafty. But same as the commercials - it's evaluation values
>>>>>are a lot more refined than mine.
>>>>>
>>>>>What do you recommend for someone like me starting from scratch?  Will short
>>>>>time control testing better my evaluation or do you think it's a waste of time?
>>>>>
>>>>>Dan H.
>>>>
>>>>I think that short time control testing is good for everybody.
>>>>
>>>>The question of Vincent was:
>>>>
>>>>"Do you believe the commercial authors at home only test at 1 0 level?"
>>>>
>>>>My answer is that I do not believe that they only test at 1 0 level but I also
>>>>do not believe that they do not use 1 0 level as part of their testings.
>>>
>>>Why do you not believe that they do not use 1 0 time control as part of their
>>>testing?
>>
>>because I think that testing at 1 0 can reveal some information very fast(for
>>example if there is a bug and the program can crash because of some change)
>>
>>I also think that watching fast games can also help to detect if the program has
>>bugs in the evaluation because you can look not only in the moves but also in
>>the evaluation.
>>
>>
>>Uri
>
>Why do you think that the commercial authors will use 1 0 testing if the only
>convincing argument is that you feel that it might be useful because they can
>look in the evaluation which they cannot do at say 2 0 ?
>
>What is your argument why 1 0 is more useful than 2 0?

1 0 and 2 0 are the same principle so it is not very important if they use 1 0
games or 2 0 games.

It is also possible that they do not use 1 0 games but 55 seconds per side time
control.

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.