Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Aspiration Window sizes

Author: Mike Byrne

Date: 08:10:35 03/14/05

Go up one level in this thread


On March 14, 2005 at 11:09:23, Mike Byrne wrote:

>On March 14, 2005 at 10:05:24, Robert Hyatt wrote:
>
>>On March 14, 2005 at 09:32:55, Renze Steenhuisen wrote:
>>
>>>Hi all,
>>>
>>>can anyone tell me whether there has ever been a study on the actual needed
>>>aspiration window sizes? The only thing I am only reading is that "it is common
>>>knowledge" and "it is known" concerning Aspiration Windows. And Crafty is using
>>>0.40*PAWN_VALUE because it seemed to perform best. But is there actual data
>>>somewhere?
>>>
>>>Cheers!
>>>
>>>Renze
>>

Oops, my repsonse was for Renze and not for you Bob.  I know I am not telling
you anything new.
>>
>>That is "actual data".  Take your program, and try different values, over a
>>large test set of positions.  Find the range that produces the greatest average
>>search depth for the test set.  It will vary for each program, since the
>>evaluation functions can produce wildly variable scores...
>
>Not only is very program different , but even different versions of the smae
>program.   When a program has a solid a evalualtion, I think you can make it
>tighter -- I have often used .35 with different versions of Crafty. In fact I
>think that may have been the Crafty standard earlier).
>
>Also if you tone down the positional points available, you can also tighten up
>the window.
>
>To answe your question, I don;t beleive there is published data, but I suspect
>some of the professional programmers have their own data and for obvious reason
>are not in a position to share.



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.