Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is Deep Blue still considered better than Deep Junior ?

Author: Robert Hyatt

Date: 14:58:05 08/20/02

Go up one level in this thread


On August 20, 2002 at 13:41:13, Uri Blass wrote:

>On August 20, 2002 at 13:34:10, Robert Hyatt wrote:
>
>>On August 20, 2002 at 10:14:57, Uri Blass wrote:
>>
>>>On August 19, 2002 at 11:25:17, Robert Hyatt wrote:
>>>
>>>>On August 18, 2002 at 23:33:02, Uri Blass wrote:
>>>>
>>>>>On August 18, 2002 at 23:24:33, Robert Hyatt wrote:
>>>>>
>>>>>>On August 18, 2002 at 23:05:35, Uri Blass wrote:
>>>>>>
>>>>>>>On August 18, 2002 at 22:41:12, Robert Hyatt wrote:
>>>>>>>
>>>>>>>>On August 18, 2002 at 21:55:06, Uri Blass wrote:
>>>>>>>>
>>>>>>>>>On August 18, 2002 at 21:21:52, Vincent Diepeveen wrote:
>>>>>>>>>
>>>>>>>>>>On August 18, 2002 at 16:17:34, Uri Blass wrote:
>>>>>>>>>>
>>>>>>>>>>>On August 18, 2002 at 11:08:37, Vincent Diepeveen wrote:
>>>>>>>>>>>
>>>>>>>>>>>>On August 18, 2002 at 09:06:02, Jorge Pichard wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>No for many reasons
>>>>>>>>>>>>  a) deep blue is old and how many old programs are still considered
>>>>>>>>>>>>     'good'?
>>>>>>>>>>>>
>>>>>>>>>>>>Even programs that reigned the world for 10 years (genius) completely
>>>>>>>>>>>>get butchered with x-0 scores nowadays.
>>>>>>>>>>>
>>>>>>>>>>>Genius is not losing x-0 unless x is very small.
>>>>>>>>>>>It is clearly weaker than the top programs of today but not so weak.
>>>>>>>>>>>
>>>>>>>>>>>Uri
>>>>>>>>>>
>>>>>>>>>>against movei perhaps. Against DIEP it's 100% simply. 20Mhz preprocessor
>>>>>>>>>>dedicated genius getting 7 ply with 1000 move book
>>>>>>>>>>against 1 million move diep book getting 10-14 ply.
>>>>>>>>>
>>>>>>>>>I thought about genius on the hardware of today otherwise it is not fair to say
>>>>>>>>>"Even programs that reigned the world for 10 years (genius) completely
>>>>>>>>>get butchered with x-0 scores nowadays."
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>It is software+hardware and not only the software.
>>>>>>>>>
>>>>>>>>>I agree that software+hardware of 10 years ago are losing x-0 against
>>>>>>>>>software+hardware of today.
>>>>>>>>>
>>>>>>>>>Uri
>>>>>>>>
>>>>>>>>
>>>>>>>>I don't even agree with that.  I played Crafty vs Cray Blitz, using 1992
>>>>>>>>hardware (a C90) and Cray Blitz won 7-3.  That certainly is _one_ example
>>>>>>>>where Vincent's statement is false...
>>>>>>>
>>>>>>>
>>>>>>>Vincent was talking about genius.
>>>>>>>
>>>>>>>Genius with the hardware of 1992 is losing 10-0 against the top programs of
>>>>>>>today with 2002's hardware.
>>>>>>>
>>>>>>>How many games did Cray blitz with the 1992's hardware play in computer
>>>>>>>tournaments?
>>>>>>>
>>>>>>
>>>>>>at least 15.  We played in 1992, 1993 and 1994 ACM events...
>>>>>>
>>>>>>>
>>>>>>>I know that it did not play in WCCC of 1995 when Fritz3 became the world
>>>>>>>champion.
>>>>>>>
>>>>>>>I do not know about a single tournament that it played.
>>>>>>>
>>>>>>
>>>>>>Now you do.  I just gave you three.  Last one was 1994 in Cape May New
>>>>>>Jersey.  Won by Deep Thought.
>>>>>
>>>>>I understand that it played only a single game against deep thought
>>>>>and a program may lose a single game and be significantly better than the
>>>>>opponent.
>>>>>
>>>>>I believe that it was signifcantly better than deep thought inspite of the
>>>>>single loss.
>>>>>
>>>>>Uri
>>>>
>>>>
>>>>I have _no_ idea what you might be basing that on.  I _know_ the difference
>>>>between the two programs.  DT was better, by a significant margin.
>>>
>>>I do not understand how do you know.
>>>
>>>Cray blitz does not beat Crafty with the same number of nodes per second so if
>>>cray blitz won 7-3 then I guess that Cray searched more nodes per second.
>>
>>Correct.  As I reported, we did 5-7M nodes per second on the last Cray we
>>used.
>>
>>
>>
>>>
>>>Crafty can search 2M on fast hardware so I guess that Cray of 1992 searched more
>>>than it.
>>
>>
>>yes...
>>
>>
>>>
>>>
>>>In other words I understood that
>>>Deep thought searched 2M nodes per second and Cray blitz of 1992 searched more
>>>nodes per second(if we talk about the same cray that beated Crafty 7-3).
>>
>>
>>Again, their nodes and my nodes were not the same, because of their ability
>>to do things in the evaluation in hardware that I could not afford to do on
>>the Cray.  Cray Blitz had the same advantage over Crafty, I could do things on
>>the Cray (using vector hardware) that I can not do on a micro (yet)...
>>
>>Comparing NPS is not a particularly useful comparison if the numbers are
>>fairly close.  IE today's programs have an order of magnitude variance from
>>fastest to slowest on equal hardware.  Once you go outside that box, however,
>>to (say) three or more orders faster, _then_ NPS becomes important in comparing
>>unequal engines.
>>
>>
>>
>>>
>>>I do not think that cray blitz was inferior software relative to deep thought
>>>and if the cray searched more nodes per second then it should be better than
>>>deep thought.
>>
>>
>>I can't answer for the very last versions of Cray Blitz with singular
>>extensions,  It _might_ have been as good as deep thought, or it might still
>>have been behind them.  But my impression from watching them play, thru 1994,
>>was that nobody was "close" to them...  myself included.
>
>My impression based on looking in games of the thing that the thing did tactical
>mistakes that the commercial of today do not do so my impression is different.
>
>If we look at the games they lost points or half points against humans then they
>often did mistakes that the commercial of today do not do.
>
>Uri


That is a poor way of analyzing, however.  How many times do the commercials
of today make horrible moves?  Just because program X won't make the same
mistake as program Y does _not_ mean X is better.  It means X is different.
Y can still be better.



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.