Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: negascout vs pvs

Author: Dave Gomboc

Date: 20:51:43 06/04/99

Go up one level in this thread


On June 04, 1999 at 22:33:48, Will Singleton wrote:

>On June 04, 1999 at 22:08:18, Dave Gomboc wrote:
>
>>On June 04, 1999 at 21:54:16, vitor wrote:
>>
>>>theyre basically the same except negascout has the extra depth>2 condition.
>>>their performance numbers are near equal too. negascout searches maybe 0.1%
>>>fewer nodes. another thing is ,if you're using extensions, doesnt it get more
>>>complicated to properly detect depth>2? so is it worth it to include the extra
>>>condition?
>>
>>PVS can be enhanced with the depth_remaining<=2 test too.  PVS and NegaScout are
>>different, but this is not the difference.
>>
>>Using the depth_remaining<=2 early exit is bad if you use extensions.  It is a
>>subtle bug, and to my knowledge, this has not been mentioned in the literature.
>>Jonathan Schaeffer found this one the hard way many years ago, debugging a
>>search that gave incorrect results. :-)  Considering that the complete special
>>condition for early termination of the search line occurs next to never, it is
>>probably faster to not do the check at all, even if you don't extend, that is,
>>the depth_remaining<=2 refinement is not really worthwhile.
>>
>>Dave
>
>I could never get that to work properly, always got bad results.  Don't know why
>it affected the search negatively.
>
>Will

It affected your search negatively because the assumption that the score won't
change is extremely likely to be false if an extension would have occured if you
had kept searching.

Dave



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.