Computer Chess Club Archives


Search

Terms

Messages

Subject: To Mr. Uri -=- With Love, Deeper Blue

Author: Slater Wold

Date: 03:36:20 09/09/01


I apologize.  This is another one of those senseless posts, that will more than
likely do _NO_ good, and only resurect the timeless arguement over Deep Blue and
todays SW & HW.  I myself _HATE_ to be the one to do this, but Uri has asked me
to prove my point, and by God, I have a point.

Here it goes:

Uri's arguement (as well as Vince, Amir, and a few others): On todays HW, many
commericial (and non-commercial) engines can spot, and play, the same exact
moves made by Deeper Blue.  Therefore, had Kasprarov been playing Deep Fritz 7
on a 8x900mhz machine, he would have lost in the same fashion.

Here's the shocker:  I AGREE!

So why the hell do I think that Deeper Blue is stronger than say, the Deep Fritz
8x900mhz combo that will be playing Kramnik in a few months?

Simple:  Look at the games.  Look for novelties.  Look for outstanding
sacrifices, or brilliant moves.  You will not find anything that even resembles
Super Grandmaster play.

It is no secret that Kasparov did not play like himself in 99% of the 6 played
games.  Kasparov made himself easy prey for Deeper Blue, and I agree, had the
_exact_ same circumstances happened with DF7, the same _exact_ result would have
came.

So what's my point:  Let me give you a clip from Anand, stating his opinon on
the match and Deeper Blue,

"Unfortunately, as a result, we were never able to see the fabulous calculating
abilities of Deep Blue. Not once did we see a spectacular example of brute force
producing a solution that differed significantly from that suggested by
intuition. A lot has been made of Deep Blue's play in the second game, but in
fact only one or two moments can be singled out - 26.f4 and 37.Be4. The rest of
the game is not that difficult, even for a computer."

What he is saying, is that the chess playing champion of the world, did not even
challenge Deeper Blue enough, for it to come up with a result that would boggle
the mind (or DF7 in this case).  A computer program need not be rated 2700 ELO
to beat a 900 ELO human.  It simply had to play smart, straight-forward chess in
order to win, and that's exactly what it did.

Case & Point:  Deeper Blue did not have to play spectacular moves in order to
win.  It could have been searching 2M nodes per second, and the result _may_
have been the same.  Hell, perhaps he could have been playing the same Deep Blue
he defeated just a few years back, and the result could have been the same.

Study:  Perhaps we are going about this _ALL_ wrong.  Instead of comparing moves
played by Deeper Blue, in an effort to defeat Kasparov, perhaps we should be
comparing moves made by Deep Blue I, in it's loss to Kasparov.  Kasparov is
human, and is subject to emotions, trouble, health, ailments, and several other
things that no computer has to deal with.  However, let's all remember that
these "chess playing machines" were developed 7 and 4 years ago, respectivly.
In other words, endgame moves made by Deep Blue I might be way off by what is
made today.  Technology HAS improved in that, Deep Blue I did not have
tablebases.  Also, in the _ONE_ game that Deeper Blue did lose, can you prove to
me that the engines of today might have made a better move?  Can the chess
engines of today, find the mistakes made in this single game?

Proof:  Anand himself said that the games were not spectacular.  And that
perhaps, it did not take all this technology to defeat Kasparov in this
particular match.  He did say that in game 2, there were two moves that did show
that Deeper Blue had great "calculating abilities."  26. f4 and 38. Be4 were the
moves.  Let's look at these, compared to todays engines.  Keep in mind, Anand
states they were correct, and that's all we have to go on.  So if you don't
believe they are correct, or optimal, take it up with Anand, NOT ME!

[D]r1rq1bk1/3b1ppn/p2p3p/1p1Pp3/PPp1P3/R1P1B1NP/2B1QPP1/R5K1 w - - 0 1

Deep Fritz T29c - W,S
r1rq1bk1/3b1ppn/p2p3p/1p1Pp3/PPp1P3/R1P1B1NP/2B1QPP1/R5K1 w - - 0 1

Analysis by Deep Fritz T29c:

26.axb5!
  ±  (1.09)   Depth: 1/3   00:00:00
26.axb5 axb5 27.Qh5 Rxa3
  ±  (1.06)   Depth: 2/10   00:00:00
26.axb5 axb5 27.Nf5 Bxf5 28.exf5 Rxa3
  ±  (1.06)   Depth: 3/10   00:00:00  1kN
26.axb5 axb5 27.Ra5 Be7 28.Nf5 Bxf5
  ±  (0.97)   Depth: 4/13   00:00:00  2kN
26.axb5 axb5 27.Ra5 Nf6 28.Bb6 Qe8
  ±  (0.88)   Depth: 5/15   00:00:00  15kN
26.axb5 axb5 27.Ra5 Nf6 28.Bb6 Qe8 29.Qf3
  ±  (0.91)   Depth: 6/16   00:00:00  39kN
26.axb5 axb5 27.Ra5 Be7 28.Bb6 Qf8 29.Nf5 Bxf5 30.exf5
  ±  (0.94)   Depth: 7/19   00:00:00  90kN
26.axb5 axb5 27.Ra5 Be7 28.Qf3 Rxa5 29.bxa5 Ng5 30.Qh5 Ra8
  ±  (0.88)   Depth: 8/24   00:00:00  146kN
26.Qf3!
  ±  (0.91)   Depth: 8/24   00:00:00  316kN
26.Qf3 Ng5 27.Bxg5 hxg5 28.axb5 axb5 29.Ra5 Rxa5 30.bxa5 Ra8 31.a6 Ra7
  ±  (0.88)   Depth: 9/25   00:00:00  510kN
26.Qf3 a5 27.axb5 axb4 28.cxb4 Rxa3 29.Rxa3 Bxb5 30.Ra7 Nf6 31.Nf5 c3
  ±  (0.75)   Depth: 10/27   00:00:01  1460kN
26.axb5!
  ±  (0.78)   Depth: 10/27   00:00:01  1515kN
26.axb5! axb5 27.Ra5 Be7 28.Qf3 Rxa5 29.bxa5 Ra8 30.Bb6 Ng5 31.Qh5 Qe8
  ±  (0.81)   Depth: 10/27   00:00:01  1844kN
26.Qd2!
  ±  (0.84)   Depth: 10/27   00:00:01  2448kN
26.Qd2 Be7 27.Nf5 Bxf5 28.exf5 Bg5 29.Bxg5 Nxg5 30.f6 g6 31.h4 Nh7
  ²  (0.59)   Depth: 11/30   00:00:03  5540kN
26.axb5!
  ²  (0.63)   Depth: 11/30   00:00:03  5731kN
26.axb5! axb5 27.Ra5 Be7 28.Qf3 Rxa5 29.bxa5 Ra8 30.Bb6 Ng5 31.Qh5 Qc8
  ±  (0.91)   Depth: 11/30   00:00:04  6275kN
26.axb5 axb5 27.Ra5 Be7 28.Qf3 Rxa5 29.bxa5 Ra8 30.Bb6 Qb8 31.Nf5 Bxf5
  ±  (0.78)   Depth: 12/28   00:00:05  8561kN
26.axb5 axb5 27.Ra5 Rxa5 28.bxa5 Ra8 29.a6 Be7 30.Nf5 Bxf5 31.exf5 Nf6
  ±  (0.75)   Depth: 13/31   00:00:15  22130kN
26.Bd1!
  ±  (0.78)   Depth: 13/35   00:00:28  42586kN
26.Bd1! a5 27.bxa5 Qxa5 28.axb5 Qxa3 29.Rxa3 Rxa3 30.Qd2 Rb8 31.b6 Ra1
  ±  (0.84)   Depth: 13/35   00:00:31  47105kN
26.Bd1 Be7 27.Qa2 Bg5 28.Bxg5 Nxg5 29.axb5 Bxb5 30.Ba4 Rab8 31.Bxb5 axb5
  ±  (0.84)   Depth: 14/36   00:00:49  73092kN
26.Bd1 Be7 27.Qa2 Bg5 28.Bxg5 Nxg5 29.axb5 Bxb5 30.Qe2 Rab8 31.Nf5 Qf6
  ±  (0.97)   Depth: 15/35   00:01:21  119226kN
26.Bd1 Be7 27.Qa2 Rab8 28.axb5 axb5 29.Ra6 Rb7 30.Qe2 Rbb8 31.Qf3 Ng5
  ±  (0.94)   Depth: 16/36   00:02:09  189645kN
26.Bd1 Be7 27.Qa2 Rab8 28.axb5 axb5 29.Ra6 Rb7 30.Qe2 Rbb8 31.Qf3 Ng5
  ±  (0.88)   Depth: 17/40   00:04:06  371740kN
26.Bd1 Be7 27.Qa2 Rab8 28.axb5 axb5 29.Ra6 Rb7 30.Qe2 Rbb8 31.Qf3 Ng5
  ±  (0.88)   Depth: 18/42   00:09:59  917474kN
26.Bd1 Be7 27.Qa2 Rab8 28.axb5 axb5 29.Ra6 Rb7 30.Qe2 Rbb8 31.Qf3 Ng5
  ±  (0.91)   Depth: 19/43   00:20:46  1934640kN

(W,  08.08.2001)

Also, at 45 minutes and 20 ply, the result is the same, and eval window had not
updated.  (There was a slip on the wheel on my mouse, that cut the search
short.)  2400k nps for over an hour, at ply 20, and no f4!  Can anyone out there
get it?


[D]r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - 0 1

Deep Fritz T29c - W,S
r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - 0 1

Analysis by Deep Fritz T29c:

37.Be4 Rd8
  ²  (0.50)   Depth: 2/7   00:00:00
37.Be4 Rxa2 38.Rxa2 Qh5
  ²  (0.34)   Depth: 3/8   00:00:00
37.Be4 Rxa2 38.Rxa2 Qh5
  ²  (0.34)   Depth: 3/8   00:00:00
37.Be4 Rxa2 38.Rxa2 Qh5
  ²  (0.34)   Depth: 3/8   00:00:00
37.Be4 Rxa2 38.Rxa2 Ra8 39.Rxa8
  ²  (0.34)   Depth: 4/9   00:00:00  1kN
37.Be4 Rxa2 38.Rxa2 Rd8 39.Qh4 Qd7
  ²  (0.31)   Depth: 5/11   00:00:00  4kN
37.Qg3!
  ²  (0.34)   Depth: 5/12   00:00:00  7kN
37.Qg3 Rxa2 38.Rxa2 Rd8 39.Be4 Qh5 40.Ra7 Qd1+ 41.Kh2
  ²  (0.31)   Depth: 6/13   00:00:00  15kN
37.Be4!
  ²  (0.34)   Depth: 6/13   00:00:00  17kN
37.Be4 Rxa2 38.Rxa2 Rd8 39.Qh4 Qd7 40.Ra5 Kh7
  ²  (0.34)   Depth: 7/15   00:00:00  36kN
37.Qb6!
  ²  (0.38)   Depth: 7/18   00:00:00  56kN
37.Qb6! Rxa2 38.Rxa2 Qd7 39.Ra5 Rb8 40.Qc6 Qxc6 41.dxc6
  ²  (0.56)   Depth: 7/19   00:00:00  76kN
37.Qb6 Rxa2 38.Rxa2 Qd7 39.Ra5 Rb8 40.Ra7 Qf7 41.Rxf7 Rxb6
  ²  (0.56)   Depth: 8/21   00:00:00  115kN
37.Qb6 Rxa2 38.Rxa2 Bc7 39.Qe6+ Kf8 40.Be4 Ra8 41.Rxa8 Qxa8 42.Kh2 Qd8
  ±  (0.84)   Depth: 9/24   00:00:00  334kN
37.Qb6 Rxa2 38.Rxa2 Bc7 39.Qe6+ Kf8 40.Be4 Ra8 41.Rxa8 Qxa8 42.Kh2 Qd8
  ±  (0.84)   Depth: 10/25   00:00:00  611kN
37.Qb6 Rxa2 38.Rxa2 Bc7 39.Qe6+ Kf8 40.Be4 Rd8 41.Ra6 Qxe6 42.dxe6
  ±  (0.72)   Depth: 11/28   00:00:01  1124kN
37.Qb6 Rxa2 38.Rxa2 Bc7 39.Qe6+ Kf8 40.Be4 Rd8 41.Ra6 Qxe6 42.dxe6 Rd2
  ±  (0.72)   Depth: 12/30   00:00:02  2517kN
37.Qb6 Rd8 38.Be4 Rab8 39.Qe3 Bc7 40.Kh2 Bb6 41.Qf3 Ra8 42.Rxa8 Rxa8
  ²  (0.63)   Depth: 13/31   00:00:10  10335kN
37.Qb6 Rxa2 38.Rxa2 Ra8 39.Ra5 Rxa5 40.bxa5 Qe7 41.Kh1 Bc7 42.Qxb5 Qa3
  ²  (0.63)   Depth: 14/32   00:00:28  26854kN
37.Qb6!
  ±  (0.94)   Depth: 15/38   00:01:12  69863kN
37.Qb6 Rxa2 38.Rxa2 Ra8 39.Ra5 Qb8 40.Qxb8+ Rxb8 41.Ra7 e4 42.Kf2 Re8
  ±  (0.84)   Depth: 16/36   00:01:45  103866kN
37.Qb6 Rxa2 38.Rxa2 Ra8 39.Ra5 Qb8 40.Qxb8+ Rxb8 41.Ra6 Bc7 42.d6 Bb6+
  ±  (1.03)   Depth: 17/42   00:04:07  247234kN
37.Qb6 Rxa2 38.Rxa2 Bc7 39.Qb7 Bd6 40.Be4 Rb8 41.Qc6 Qxc6
  ±  (1.03)   Depth: 18/42   00:08:15  493570kN
37.Qb6 Rxa2 38.Rxa2 Ra8 39.Ra5 Rxa5 40.bxa5 Qb8 41.Qxb8+ Bxb8 42.Kf1 Kf7
  ±  (1.09)   Depth: 19/45   00:24:30  1480448kN

(W,  08.08.2001)



Once again, I do not get the move that Deeper Blue gets, or that Anand said was
correct.  This search is a little slower, and a little less deep, but if anyone
out there can get it, I'd love to know about it.

Conclusion:  Deeper Blue, and the chess programs & HW combo's of today may not
be "worlds" apart.  Simply, they cannot be.  They are both chess playing
programs!  However, I believe that there are certain circumstances that arrive
that make it easy for you to claim they are not better, and for me to claim they
are better.  However, with 6 games, and possibly only 425 positions to compare
with, is there really enough to determine, that in the most complex, and
difficult of positons, Deeper Blue and Deep Fritz are always going to agree?  I
don't believe so.

My arguement:  That at 200M nodes per second (not to mention some pretty
revolutionary software, at that time), I don't care _WHAT_ positon you throw at
Deeper Blue, it will _ALWAYS_ solve it faster than Deep Fritz on todays HW.
_PERIOD_.  (Which in essence, leads me to belive that Deeper Blue, would be
stronger, of course.)


The End


Slate

* This post was created very late in the nite.  I apologize ahead of time, for
any incomplete thoughts.  :) *



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.