Author: José Carlos
Date: 01:18:12 02/05/00
Go up one level in this thread
On February 05, 2000 at 04:04:06, Dann Corbit wrote:
>On February 05, 2000 at 03:51:24, Gian-Carlo Pascutto wrote:
>
>>On February 05, 2000 at 01:05:13, Dann Corbit wrote:
>>
>>>It uses NegaScout() rather than NegaMax() like {for instance} Crafty.
>>
>>
>>Hahaha...Crafty (and every other half-decent chessprogram) uses
>>Principal Variantion Search, which is essentially identical to
>>NegaScout.
>>
>>If it were to use pure NegaMax, you'd see the search depth nearly
>>halved.
>
>Shows you what I know, which is not much. I was just going by the comments in
>search.c that say this:
>
>********************************************************************************
>* *
>* Search() is the recursive routine used to implement the alpha/beta *
>* negamax search (similar to minimax but simpler to code.) Search() is *
>* called whenever there is "depth" remaining so that all moves are subject *
>* to searching, or when the side to move is in check, to make sure that this *
>* side isn't mated. Search() recursively calls itself until depth is ex- *
>* hausted, at which time it calls Quiesce() instead. *
>* *
>********************************************************************************
>
>What (exactly) is the modification that turns a negamax search into a Principal
>Variantion Search?
In PVS, you search the first move (PV move) with the full AlfaBeta window (or
the aspiration window, if you use AS), and the rest of the moves with a null
window Alfa,Alfa+1 that can only fail (high or low). If your move ordering is
good, none of those moves will fail high, and you'll get to the next ply very
fast.
José C.
>In what way does this cause identical search performance?
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.