Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Verified Null-Move Pruning, ICGA 25(3)

Author: Omid David Tabibi

Date: 21:43:35 11/26/02

Go up one level in this thread


On November 27, 2002 at 00:12:19, Robert Hyatt wrote:

>On November 26, 2002 at 21:47:00, Omid David Tabibi wrote:
>
>>On November 26, 2002 at 20:02:06, Robert Hyatt wrote:
>>
>>>On November 26, 2002 at 16:21:00, Omid David Tabibi wrote:
>>>
>>>>On November 26, 2002 at 15:58:06, Gian-Carlo Pascutto wrote:
>>>>
>>>>>On November 26, 2002 at 15:55:56, Omid David Tabibi wrote:
>>>>>
>>>>>>So it is reasonable that on every program starting from a certain depth >adaptive null-move pruning will always construct a smaller search tree.
>>>>>
>>>>>Don't you mean the other way around?
>>>>>
>>>>
>>>>Yes :-)
>>>>
>>>>Starting from a certain depth, verified null-move pruning will always construct
>>>>a smaller search tree than the adaptive one.
>>>>
>>>>
>>>>
>>>>>--
>>>>>GCP
>>>
>>>
>>> I am doing some testing now.  First thing I noticed is that for WAC, the
>>>time-squared
>>>measurement went down very significantly for your algorithm.  And I have not
>>>modified
>>>anything such as turning null-move off when low material happens, since your
>>>idea will
>>>catch the zug problems.
>>>
>>>I am playing a match between old and new, and at the moment the result is
>>>
>>>5 wins, 2 losses, 13 draws, new version is ahead.  Not enough games, but it is
>>>interesting...
>>>
>>>More as I get more info...
>>
>>Definitely interesting...
>>
>>BTW, since Crafty has strong extensions, I suggest that you also try reducing
>>the depth by 2 plies (instead of 1) on fail-high reports. In the "Conclusions"
>>section I reported that in this case the tactical strength was lower than std
>>R=2 on occasions; but considering the fact that its search tree was also far
>>smaller, it is still an interesting option.
>>
>>In the paper, my goal was to present an algorithm tactically stronger than std
>>R=2, and with smaller search tree. In practice however, other values might work
>>even better.
>
>Intuition tells me that 1 or 2 is wrong.  That is should be some sort of
>proportional value insteadk.  I've been fiddling with the adaptive depth
>and the extension limit as well, playing with scaling it proportionally rather
>than lopping things off in a static way...  I think that reducing the depth by
>some percentage might make more sense...

I also believe that an adaptive value should exist here. For more than a month
now, I have been testing different adaptive values; I have found some
improvements, but no breathtaking ones yet...!

>
>But I will fiddle and test a while and report back if I find anything that
>improves it further, since it is your idea...

Maybe we can write together a paper as a sequel to adaptive/verified null-move
pruning.



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.