Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: LCT II Fin4, Deep Thought, and Deep Blue (was Re: LCT II results...)

Author: Robert Hyatt

Date: 11:23:28 01/07/98

Go up one level in this thread


On January 07, 1998 at 12:06:25, Amir Ban wrote:

>On January 07, 1998 at 08:41:20, Robert Hyatt wrote:
>
>>On January 07, 1998 at 06:24:02, Amir Ban wrote:
>>
>>>On January 06, 1998 at 20:10:07, Robert Hyatt wrote:
>>>
>>>
>>>>This I don't follow.  What micro has beaten a GM in 40/2?  In a match
>>>>of 40/2?  What micro has beaten as many GM's as DB in anything
>>>>(excepting
>>>>blitz, where most micros do ok at times)...
>>>>
>>>
>>>Are you joking ? This is happening all the time nowadays. I didn't
>>>count, but I think the most numbers of computer wins over GM's at full
>>>time come from Aegon. There's also Rebel-Yusupov, and I think I remember
>>>at least one time with Fritz  (Glek ?). My own contribution is Gofshtein
>>>(1994) and Kossashvili (1996). There are only several GM-computer games
>>>played each year, but the computers have decent percentages, which is
>>>not surprising since they are playing at the strong-IM/GM level
>>>nowadays.
>>
>>
>>1.  Last time I looked, Aegon wasn't full tournament time control.  It
>>was faster.
>>
>
>Yes it is. You are splitting hairs. Few tournaments are 40/120 any
>longer anyway, including the WCH. The most common format I saw lately
>was 30/90 + rest/60.
>
>
>>2.  computers are not yet at strong IGM level in tournament chess.  In
>>blitz, yes.  40/2?  Nope.
>>
>
>As you wish, but you are becoming lonely in this opinion, and have to
>ignore or explain away an ever increasing amount of conflicting
>evidence.


I'm not as lonely as you suspect.  I can name a list of IM challengers
that are a pain for computers, because they aren't concerned with the
tactical skills of the computer as much as they are confident in their
own playing skills.  Regardless of what you might want to say, a program
that searches over 10K nodes per second on a PC has *nowhere* near the
chess knowledge that a good IM possesses.  And I mean it is not even
*close*.

if you look at some of Junior's games, I'm sure you come to the same
conclusion that I come to when I study Crafty.  yes it plays nice moves
at times, but yes it plays moves a beginner likely wouldn't try in other
positions.  Moves that *no* IM would consider, much less play.  The
tactical
skills of the programs are beyond reproach.  But their knowledge-base is
still in a very infantile stage of development.  And this doesn't begin
to
consider players like Kasparov, Karpov and so forth, who know so much
more
that is not funny...



>
>>>
>>>Matches ? There weren't any. OK, there was Rebel-Yusupov, and
>>>Junior-J.Polgar (which I lost). These things don't happen because
>>>there's no money. The truth is that there was only one company that
>>>would put up a million dollars to drag the world champion to its back
>>>yard, and even it lost interest now. I know from experience that even a
>>>few thousand dollars is next to impossible to get. We are trying to
>>>organize a Polgar rematch without success for two years, and there are
>>>plans to organize a Junior-Alterman match which proceed smoothly until
>>>money is discussed.
>>
>>
>>there have been other micro matches.  With disastrous results.  There
>>have been DB matches with predicted results.  The two aren't close.
>>
>
>Other GM-computer matches ? Where, when and what result ? (BTW, I didn't
>forget about Hiarcs-Hergott, it's just that Hergott is a mere IM).
>

Bruce, I, CM5000, Genius and I'm not sure who else (perhaps Rebel)
played
5 GM's on chess.net a few months ago, round robin man vs machine,
game/30
time control.  All 5 computers finished with higher scores than any of
the
GM players.  But *all* the games were pretty well marred with tactical
mistakes by the humans (IE against GM Roman Crafty hit on a mate in 26
or
something similar in a 12 piece ending (exchanged into a 4 piece ending
database)...



>
>>
>>>
>>>Most GM wins at non-blitz ? I have eight: Gofshtein, Bikhovsky, Har-Zvi
>>>(twice), Liss, Finkel, J.Polgar, Kossashvili, plus a few draws including
>>>Yudasin.
>>>
>>>Amir
>>
>>
>>what time controls?  I beat Walter Browne in a (roughly) 30 second per
>>move match.  5 computers beat 5 GM's a year ago in a game/30 tournament.
>>But those are *not* tournament time controls.
>>
>
>(Big sigh) Non-blitz means 30 min/game or more. I was rather careful to
>look at your question and answer it in the exact terms you phrased it,
>but I had the feeling you will shoot me down anyway.

I don't consider 30 min/game non-blitz, really.  That is still a
favorable
time control for computers, because by the time the middlegame is
underway,
we are into a blitz time/move.  I agree that at game/30, we can beat
some
GM players.  But I also believe that at this same time control we are
going
to lose many games too, particularly as GM players adopt anti-computer
styles, rather than playing "normal" chess.

pop over to ICC and try Adolf, Greg1, Schroer, (in times past
Dreamteamer)
and others.  They are deadly because they play the opponent and not the
board, and once they learn how to block the center, I haven't seen a
program
that understands how to play the resulting position, yet...



>
>Actually on a recount it's 9, since I forgot Gurevich. I'm not sure I
>want to count it since it was played in the Harvard Cup where nothing
>much was at stake. All the other wins were in serious play with money
>and rating points at stake.
>
>
>>I can post perhaps a *hundred* wins against GM's at non-blitz.  But I
>>can also post several hundred losses.  :)
>>
>>But the
>>problems are:  (1) they are still "skittles"-type games with no prizes
>>or anything;  (2) they are *not* 40/2...  many are game/30, some are
>>of the form 30/30 (30 mins + 30 sec increment) and so forth.  But I'm
>>smart
>>enough to know that crafty is *not* a gm-level player.  That's wishful
>>thinking...  The best GM's are still *very* difficult, even at blitz.
>>Much
>>less "standard" time controls...
>
>Your rhetorical skills are admirable. You list some rather unserious
>facts about crafty, dismiss them out of hand, and imply that I have done
>the same.
>
>No, ICC and private games don't count, for good reason. Must make one
>important exception to this of course: Any unsupervised and unpublished
>10-game match played at your home/office at an unknown date in unclear
>conditions against uncertain opponents and the result spread through
>hearsay counts as irrefutable evidence and is given precedence over any
>other fact.
>
>Amir

I don't believe I have done this.  However, I have *enough* GM games so
that statistically the results mean something.  IE My current database
here
has 619 games against GM Larry Christiansen.  *619* games total.  Of
that
total, 18 are longer than game/30.  1 was game/30, the rest are odd
blitz
games like 5 5, 5 8 and so forth.

If you'd like to pick a GM, I can give you my results against him, as I
am fortunate enough to get the opportunity to play *many* games against
them (mostly blitz, but some longer games.)  Yes, I win 90% of the blitz
games, but by tactical smoke and mirrors.  *not* by superior
understanding
of the game.  They know an *awful* lot about the game that programs have
not
even been exposed to yet...

The only reason I respond is that you seem to overestimate what the
programs
are doing.  A synopsis is they are catching GM's in tactical mistakes,
*not*
outplaying them in endgames or anything.  Which is certainly part of the
game
of chess.  But when GM's start playing anti-computer, this turns ugly,
as we
have seen at Aegon.  And more are beginning to adopt this style, which
is
going to turn back the "silicon tide" for a few more years.

But if you *really* think that Junior understands the game of chess, we
should pick some positions and sit down with an IM and discuss them.
The
weaknesses of Junior (or Crafty or Ferret or Genius or Hiarcs or ...)
will
become apparent quite quickly.  tactically brilliant.  Positionally
barely
adequate.



This page took 0.01 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.