Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Kiriakov vs Deep Shredder annotated by Kiriakov.

Author: Chris Carson

Date: 11:15:05 06/05/01

Go up one level in this thread


On June 05, 2001 at 12:51:19, Uri Blass wrote:

>On June 05, 2001 at 11:24:09, Robert Hyatt wrote:
>
>>On June 04, 2001 at 23:16:07, Mark Young wrote:
>>
>>>On June 04, 2001 at 22:37:54, Robert Hyatt wrote:
>>>
>>>>On June 04, 2001 at 11:08:11, Mark Young wrote:
>>>>
>>>>>On June 04, 2001 at 10:34:36, Slater Wold wrote:
>>>>>
>>>>>>This is the annotated game by Petr.  Look like Mark was correct, 10...Bxf3 was
>>>>>>the correct move as well as 10...Be6 or 10...Bh5.  Shredder was lost, bottom
>>>>>>line.
>>>>>
>>>>>I can't be wrong all the time:)
>>>>
>>>>
>>>>I doubt that _one_ move lost the game.  It was the _group_ of bad moves that
>>>>made it hopeless..   Against a good GM you can't toss a tempi here, a tempi
>>>>there, and expect to do anything but get tossed overboard yourself.  :)
>>>>
>>>>Of course, the computers are all super-GM players, so this discussion really
>>>>is _not_ happening.  :)
>>>
>>>Who said "all" computers are super-GM players. :)
>>>
>>>The only computers that "could" try and claim that they are Super-GM Players are
>>>Deep Blue, and Deep Junior. We will have to wait and see what Fritz 7 can do.
>>>
>>>Many programs are GM strength(> 2500 elo), but Super-GM...I don't think so
>>>either.
>>
>>"many"???  Shredder has beaten them all in the last couple of major events.
>>Do you think _it_ is a >2500 player after looking at the above game?
>
>You cannot decide about the rating of a player based on one game.
>The fact that Shredder beated other programs does not prove that it is better
>than them.
>
>This kind of game convince me to guess that Shredder beated other programs in
>tournaments mainly because of better opening preperation(better opening
>preperation is not about getting better position out of book but about not
>getting positions when shredder does mistakes like Bg4-c8).
>>
>>I'm not going to argue the point as it is a pretty hopeless thing to do.  But
>>so long as computers play like the above, I don't think any reasonable human
>>will ever say "that is a >2500 player."
>
>They do not play like this every game and it is possible to get more than
>2500 with some weak games.
>
>The facts are that the results of top programs against humans suggest that they
>can get more than 2500.
>
>I also prefer not to count games when a buyer of the program operates the
>program because I suspect that Stefan could use different opening book in a real
>tournament against humans.
>
>Uri

Uri,

I agree with you on this.

At least for me, I look at results since both Digital GM's and Human GM's
make mistakes and have a few bad results from time to time (take a look at any
issue of TWIC and you will see plenty of +200 or -200 performances and lots of
silly mistakes by 2500+ players).  The Digital GM's
continue to have fewer and fewer bad games under "arbitrated events" as the
HW gets faster and the SW gets better (Chess Tiger on the palm is a very
good example of continued SW improvements).  For me, exceptions do not make
a rule, but do provide valuable information for the programer to make
improvements and are great for me as a Human to keep my hopes of winning alive.

I love to see a Human (any rating) get a draw or a win!  However, most of the
draws and wins these days seem to be an "ambush" to me.  I mean that I do not
see many consistant results for a < GM beating a Digital GM over a long period
of time.  I wonder how adaptable the human is if the learning (book and
positional) has been through a few games and the basic "ambush" is avoided by
the Digital GM?  If Digital GM's are very weak and easy to figure out (I can not
do this on a consistant basis, but maybe someone can), I wonder why I have
not seen results where the Human begins to win most of the time and the Digital
GM almost never wins/draws.  I have not seen any reports of this in a very long
time (going back to the 486 days) or any "arbitrated results" like this.  The
current results I have seen show the opposite, the Digital GM's are winning by a
margin of 2.5 to 1 after about 235 games against FIDE rated players.  When will
I see the results go the other way?  I am of course talking about more than a
random game here.  I need to seen several (say 5) different <2300 players
beat several programs (say the top 5) on > 1GHZ machines in a long match (10
games).  If a critic can set up an arbitrated 5 on 5 round robin match at 40/2
and the humans win, I would be convinced.  I can get a couple of wins or a few
draws from time to time if I really practice (my own improvement) and learn to
exploit the Digital GM's mistakes before it can adapt.

Even Kramnik needs three months to prepare his trap to ensure a victory in a
short match.  I wonder...

Best Regards,
Chris Carson



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.