Author: Albert Silver
Date: 14:09:12 01/30/02
Go up one level in this thread
On January 30, 2002 at 15:57:04, Vincent Diepeveen wrote:
>On January 30, 2002 at 14:20:36, David Rasmussen wrote:
>
>>On January 30, 2002 at 11:15:03, Gian-Carlo Pascutto wrote:
>>
>>>>
>>>>In 1998 and 1999 it was mentionned by direct postings from Hsu
>>>>and others that they only did a fullwidth search not a single form
>>>>of pruning as they disbelieved this, Bob has quoted that zillions of
>>>>times in these years.
>>>
>>>Bob'll have to be the reference on this, but I always understood they
>>>didn't use any form of pruning except for quiescent futility pruning,
>>>which is what I tested with.
>>>
>>
>>I don't think DB used FP in quiescence.
>
>it is pretty trivial thing to do futility pruning in qsearch,
>though incorrect. When DB was a beancounter in the true sense
>(deep thought) i am sure for a psq program with very small values
>for king safety (near to none) and doing little in eval in general
>there futility is definitely a smart thing to do.
>
>When then later the thing got some more speculative eval, i bet
>he didn't even test without FP for correctness. How can a no-rating
>guy measure the difference anyway, all they did were a few blitzgames
>against old rebel version short before it played kasparov.
First of all, I thought that was why they had some grandmasters on the team as
consultants such as Benjamin and Illescas. As to testing against a program, I
heard the story about it beating the micros which isn't terribly hard to believe
as this was in 1997, so I would imagine that it's best and strongest pratice
partner would be itself. In which case, Hsu could just have it play games
against itself to test one version with a given function/algorithm, and another
without.
Albert
>
>>/David
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.