Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Much of Deep Blue's parallel searching was wasted >>>

Author: Vincent Diepeveen

Date: 08:09:16 10/04/02

Go up one level in this thread


On October 04, 2002 at 09:37:34, Uri Blass wrote:

the difference is that DB searched everything like a minimax
program does and nowadays we are a bit smarter than that :)

still getting 12 ply with that isn't that bad, of course its
evaluation is from 1997, so it would get annihilated now.

>On October 04, 2002 at 08:18:19, Shep wrote:
>
>>On October 04, 2002 at 06:55:44, Gian-Carlo Pascutto wrote:
>>
>>>On October 04, 2002 at 06:53:03, Uri Blass wrote:
>>>
>>>>If you take a program with bad search rules and give it 200M
>>>>nodes per second then it can beat Deep Fritz at blitz but not
>>>>at tournament time control.
>>>
>>>I would say, maybe not even in blitz.
>>
>>I agree. Don't forget the best blitz players are slower searchers like Tiger and
>>Hiarcs.
>>
>>---
>>Shep
>
>The difference in nodes per second is not 200M against 3M
>
>It is possible to test it with public weak programs that use inferior
>algorithms.
>
>Take for example the public movei.
>
>If I give it 20M nodes per move against 0.3M nodes per move of Fritz then I
>expect Movei to beat Fritz but if I give it 1000M nodes per move against 15M
>nodes per move of Deep Fritz then I expect Deep Fritz to beat movei.
>
>This is only a guess and I did not check it.
>If people want to check it then they can do it.
>
>Public Movei does not support analysis but people can use setup position and
>give it long time control of 72 hours per game so it is going to use more than
>an hour per move and it has no problem to search 1000M nodes in an hour on fast
>hardware(not on mine).
>
>They can use the option move now after time that is eqvivalent to the relevant
>number of nodes or use the debug file to decide about the move of movei after
>number of nodes that is big enough.
>
>Uri



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.