An analysis of the first game. 

 GM Garry Kasparov (2835) - Deep Junior (C) (2500) 
[D45]
X3D "Man vs. Machine" Match
New York City, NY;  [U.S.A.]  (Rd. # 1),  26.01.2003

[A.J.G.]

This is a game for me - that fails completely to live up to all the pre-game hype and hoopla. ....... it wasn't even that good.

This game should also be a dire warning to the twits who thought that this event was going to be any- thing remotely like Kramnik versus Deep Fritz 7.0 (Garry is obviously out for blood, and is willing to play the very sharpest lines, something Kramnik was NOT willing to do.)

1.d4 d5;  {Diagram?} 
The computer responds in a strictly classical fashion. I would have thought, based on several other  'human-vs. computer'  matches as of late, that this would have been a good idea.

But Kasparov has something very specific in mind.

     [ Just as good was: 1...Nf6!? ]  

2.c4 c6;  {Diagram?} 
The Slav or the Semi-Slav should have been a good choice for the Deep Junior Team.
(The Fritz team did fairly well with this type of pawn structure and classical systems vs.
 GM V. Kramnik.)

One of the basic ideas of the Slav is to bolster the center but without trapping his QB as in the main lines of the other (more) classical systems. Anand is probably the leading exponent of this line.

But ...
# 1.)  Kasparov allowed this for a reason, probably to exploit some type of
          weakness he has detected in the armor of the program.
# 2.)  Kasparov has probably played dozens of test lines against the program,
          (Junior 7.0); and feels that the computer does not handle this position well.

     [ The MAIN LINE is probably:  2...e6;  3.Nc3 Nf6;  4.Bg5,   {Diagram?} 
        which is, ... "The Pillsbury Attack."  
        (The main line of the well-trodden and well known Q.G.D.)  ] 

White's next move conforms to all the principles of correct play in the opening,
but is just a little unusual for this line.
3.Nc3,  {Diagram?} 
Obviously a good move, but not the normal move order for this line.
(Kasparov may be considering playing one of the obscure lines with Nge2
 at a later point.)

     [  More usual would be: 3.Nf3 Nf6;  4.Nc3 e6;  5.e3, "+/=" {Diagram?}
        with a slight edge for White. ( The move,  5.Bg5!?,  {Diagram?}    
        could lead to the main lines of the Classical Variation ... or the very   
        wild lines of the Botvinnik Gambit. (Depending on how Black responds.) ) ] 

3...Nf6;  4.e3,  {Diagram?} 
Simple and safe. White says:
"I never have to worry about ...d5xc4; and maybe Black trying to hang onto the
 newly-captured c-pawn." 

And with the Black Knight already on f6, ....e5; is almost out of the question.
A very sound and well thought-out anti-computer strategy. 

***

     [  Another good line is:  4.Nf3 e6;  5.e3,  {Diagram?} 
         Probably the most solid and sensible move.

        ( White could also  (maybe!)  try:  5.Bg5!? dxc4; 6.e4 b5; 7.e5 h6; 8.Bh4 g5;   
          9.Nxg5! hxg5; 10.Bxg5 Nbd7!; 11.g3 Bb7; 12.Bg2 Qb6; 13.exf6, "~"  {Diag?}   
          which is,  "The Botvinnik Variation,"  one of the sharpest lines in all of chess.   
          I personally think this would be a very good line to play against a computer,  
          as they play positions of a material imbalance VERY poorly.    

          However to off-set this, memory plays a HUGE part in this line ...   
          and the computer would not be likely to forget a key move at a critical   
          point in the game. )  

        5...Nbd7;  6.Bd3,  "+/="  {Diagram?}  
        which I believe is the line known as,  "The Meran Variation."  
        White normally eeks out a small but consistent advantage.  ] 

***

4...e6;  {Diagram?}  
This is probably the best move for Black, and the one most consistent
 with opening principles. 

     [ One Internet player told me that Black should play ....e5;  and open up
        the game. While this might work against the average human, it would fail
        miserably against a GM. 

        I.e.,  </=  4...e5?;  5.dxe5 Ne4;  {Diagram?} 
        This is pretty much the best.

        (The line 5...Nfd7?!; 6.cxd5, "+/"  {Diag?}  is even worse for Black.   
           Nearly  "+/-")   

       6.cxd5 Nxc3;  7.bxc3 Qa5;  8.Nf3 cxd5;  9.Bd2,  "+/"  {Diagram?} 
        White is clearly much better. ] 

5.Nf3 Nbd7;  6.Qc2!?,  {Diagram?} 
This is a different line ... that introduces a severe imbalance in the position.  (You
 can bet this idea was prepared quite a bit in advance by Kasparov and his team.)

Apparently Garry K. has also discovered a chink on the pre-programmed opening
'book'  of the computer's program.

I personally have discovered that lines of severe positional or material imbalance
are perfect to play against the computer, and exactly the type of lines that the
machine/program is most likely to make a severe error against.

***

     [  Last chance to play the  Meran Variation  with:  6.Bd3 dxc4;  7.Bxc4, 
         7...b5;  8.Bd3 a6;  9.e4 b4;  {Diagram?}  One of the more modern lines.

         ( Or  9...c5!?;  10.d5!?, "+/="  {Diagram?}  White has a slight edge here.
            [The Reynolds Variation.]     ( Or - alternatively - MCO gives the sharp line:   
             10.e5!? cxd4!?; 11.Nxb5!, "/\"  {Diagram?}  with continuing complications.   
              [ See MCO-14, page # 457. ] )   )  

        10.Na4 c5;  11.e5 Nd5;  12.0-0,  "+/="  {Diagram?}  
         White has a small edge here.  

         cf. the encounter: 
         GM I. Sokolov - GM A. Shirov;  Stockholm, 1990. ] 

***

6...Bd6!?;  {Diagram?}  
A good developing move ... that is also the main line here. 

     [ Playable is: 6...Be7!? ]  

7.g4!?,  (Maybe - '!')  {See the diagram just below.}  

A move of incredible difficulty that has the added value of looking like a simple 
blunder. This move "ups the stakes" in the war of sharpness, and had Kasparov 
lost this game, I am sure this move would have been criticized very harshly. (!) 

  The basic idea is that if Black takes the g-pawn, White plays Rg1 with some   
  advantage. (At least, that's the way it is supposed to work ... on paper, anyway!)    

I have at least a couple of books on The Slav, a few more on just the Semi-Slav, 
and at least a dozen pamphlets on the various lines in both of these openings. 

7.g4  is a lively and complicated move and is originally the idea of ... 
GM A. Shabalov,  (The current  U.S. Champ.);  according to one author. 
 [ See the book, "The Complete Semi-Slav,"  by  IM Peter Wells.  
    Date of printing, (©) 1994. ]  
   { Wells analyzes the complicated game,  
      GM M. Krasenkow - GM E. Sveshnikov; 
      Tal Memorial Tournament, Moscow/RUS/1992. (1-0, 23 moves.) } 

*****

     [  More often played is Bd3 or even Be2. 

        For example: 7.Be2 0-0;  8.0-0 dxc4;  9.Bxc4 b5;  10.Be2 Bb7;  
        11.Rd1 Qc7;  12.Bd2 a6;  13.b4,  "+/="  {Diagram?}  
        with a very small edge for the first player here. 

        GM Lubosh Ftacnik (2590) - GM Jeroen Piket (2495) 
        Groningen, 1988. 

***

        The main line is probably:  7.Bd3 0-0;  8.0-0 dxc4;  9.Bxc4 b5;  10.Bd3, 
        10...Bb7;  11.e4 e5;  12.dxe5 Nxe5;  "~"   {Diagram?}  
        The end of the column.  

        (MCO goes on to analyze several different tries. Dozens of Master games 
         have been played from this position, and the consensus is that this line 
         pretty much offers equal chances to both parties.)

         (See the encounter:)  
         GM L. Portisch - GM Tukmakov;  Reggio Emilia, 1987/88. 
          [ See  MCO - 14;  page 468, column # 33, and note # (l.). ]  ]   

*****

7...dxc4!?;  {Diagram?}  
Black opens the game, this strikes me as a little inaccurate. 

In this position, I would say that Black needs to complete his development as quickly 
as possible. To open the game now - is to really accentuate White's Two Bishops. 

***********

     [  Another popular line here is:  7...Bb4!?;  8.Bd2 Qe7;  9.a3 Bxc3;  
        10.Bxc3 b6; 11.Bd3, "+/="  {Diagram?}  White holds a small edge. 

        GM B. Gelfand - GM V. Kramnik;  Germany, 1996. 
        [ See MCO-14; page # 468, col. # 33, and also note # (j.). ]  

*****

        Can Black take the pawn?  Different games give various results. Black could 
        play the move:  7...Nxg4!?;  "~"  {Diagram?}  with an unclear position.  

        In some games, taking the pawn has been good for Black, in others it has been 
        a total disaster. 

        For example: 8.Rg1 f5!?;  {Diagram?} 
        This looks good, but could be just a little risky.  

***

          ( Maybe best is: 8...Nh6!?; "~" {Diagram?}  with an unbalanced position.   

            But Black should NOT play: </= 8...Nxh2?!; 9.Nxh2 Bxh2;     
            10.Rxg7, "/\"  ("+/=")  {Diagram?}  when White has an initiative.   

            Another (interesting) line is: 8...Qf6!?; 9.Rxg4 Qxf3; 10.Rxg7 Nf8; 11.Rg1,   
            11...Ng6;  12.Be2 Qf6; 13.Bd2 Bd7; 14.0-0-0 Bxh2!?; 15.Rh1 Bc7;    
            16.e4!?, "~" {Diagram?}  with a very complex position.    
            GM M. Krasenkow - GM J. Piket; Bundesliga/GER/2002.  (1/2-1/2, 39) )   

***

        9.h3 Ngf6;  10.Rxg7 Ne4;  11.Bd2 Qf6;  12.Rg2 b6!?;  13.Qa4! Bb7;  
        14.cxd5 b5!?;  15.Nxb5!! cxb5;  16.Bxb5 Rd8;  17.Bc6 Nxd2;  18.Kxd2!,
        18...Bxc6;  19.dxc6 Nb6;  20.Qxa7, "+/"   {Diagram?}  
        White has a very large advantage, and went on to win in just a few more 
        moves from this position.  

        GM Mikhail Krasenkow  (2555)  -  IM Alexander Filipenko  (2375);  
        Tal Memorial Tournament, Moscow/RUS/1992. (1-0, 28.)  ]   

***********

8.Bxc4 b6!?;  {Diagram?}  
This strikes me as weakening a lot of key squares, and lays the groundwork 
 for what happens later. Maybe just ...h6 instead?  

*****

     [ Possibly  8...Nd5!?;  {Diagram?}  with interesting play.  

***

        Or  8...h6;  9.h3!? 0-0!?; "~"  {Diag?}  with an unclear position. 

***

        The computer  'book'  for  Fritz 6.0  {program}  gives the following line: 
        8...b5!?;  9.Be2 Bb7!?;  10.e4 Nxg4!?;  11.h3!?,  {Diagram?}  
        I don't know about this. 

         (Maybe better is: >/=  11.Rg1!, "+/=")   

        11...Nh6;  12.Rg1 e5!?;  ('?!')  {Diagram?}  
        It is senseless for Black to open the game here. 

        13.Be3!? Qe7?!;  14.0-0-0 b4;  15.Na4 exd4!?;  16.Bg5 f6?;  
        17.Bxh6 gxh6;  18.Nxd4 Bf4+!?;  19.Kb1, "+/"  {Diagram?}  
        White is clearly much better here, and is very close to having a totally 
        winning position. (An atrocious line for the second player here, to be sure!)

        This is - I am quite sure - based on the following contest: 
        GM Vladimir Akopian (2650) - Darius Ruzele (2520); 
        14th Summer Tournament,  Berlin/GER/1996. (1-0, 22 moves.) 

          Another example of a hole in the computer's opening book!   ]   

*****

Now White grabs the center. 
9.e4!,  {Diagram?}  
Not only dominating the middle of the board, but this threatens to fork, 
(and win) a piece with e5. 

This is by far the most energetic move, and much better than the wimpy h3, 
to protect the g-pawn. (One could almost expect as much from Kasparov, 
who is noted for his energetic and forceful play.)   

     [  9.h3, h6; "=" ]  

9...e5!? ; {Diagram?} 
This strikes me as a little bit risky, Black is opening the game up, when he is 
not really suited to handle the resultant complications. 

Maybe better is the simple retreat ...Be7; or maybe Black could try ...Bb7!?; 
with some very tricky tactics. 

According to several sources, the computer was already out of its opening book, (!) 
and took a very long time to replay here. 

*****

     [  If Black plays: 9...Be7!?; {Diagram?}  he reaches a known position, but a 
         whole tempo down! (But this still might have been better than the game!)  

        See the game:  R. Kempinski - A. Czerwonski;  51st National Championship, 
        Gdansk/POL/1994. (0-1, 30);  

***

        If  9...Nxg4!?;  then  10.e5, "+/="  {Diagram?}  with a slight edge to White. ]  

*****

10.g5 Nh5!?;  {Diagram?}  
Once again the computer chooses a slightly risky move. It might have been 
much more prudent to simply play ...Ng8; in this position.  

     [  Maybe better was: 10...Ng8!?;  {Diagram?}  Is this forced?/best? 
        (The computer may have avoided this line because Qd1! - indirectly 
         threatening d6 - seems to give white a solid edge, with little counter-
         play for Black.)  ]  

11.Be3 0-0!?;  12.0-0-0!,  {Diagram?}  
Once again White plays the most aggressive continuation. It is almost a paradox
that the King might be safer here than on the K-side, because White played the 
pawn advance, g2-4 earlier. 

     [  If White had played: 12.Qd1!? Nf4;  "<=>"  {Diagram?}  
         Black has good counterplay, the position is almost equal. 

***

         Possibly:  12.h4!?, "="  or even  12.Qb3!? exd4;  "<=>"  {Diag?} 
         But probably not:  12.0-0?! exd4;  13.Nxd4 Bf4; "=/+"  ]  

12...Qc7;  13.d5!,  (TN)  (Maybe - '!!')   {Diagram?} 
Apparently this is a new move here.

I would (almost) be willing to bet that Kasparov had prepared this particular idea
weeks - or months! - in advance of this contest. (Specifically for Deep Junior.)

The move Be2 has also been played here, but has not yielded White much
 advantage. (cf. C. Ward - E. Gausel; Copenhagen, 2002.)

*****

     [  In yet another game,  White played the move: 13.Kb1,  "+/="  {Diag?} 
        which also yielded White a fairly decent advantage. 

        T. Hillarp.Persson - G. Borgo; 
        /Batumi, GEO; 1999.  (1-0, 37);  

***

       The move: 13.d5!,  {See the diagram just below.} 

       greatly affects a whole complex of light-colored-squares here, in this position. ]  

*****

13...b5?!;  {Diagram?} 
A terrible move, (maybe - '?');  yet many of the programs that I tested this 
position on played this move here. 

(I tested this position on over twenty different programs here: Rebel, Chess Genius, 
 Essentia, Flash, ChessMaster, Crafty, GM-Chess, & GM Chess ULTRA, Gnu-Chess, 
 Chess-X, Combat Chess, and many others. Many show a distinct proclivity for the 
 pawn advance of ...b5.) 

Knaak gives this a dubious. (GM D. King also condemned this move as well.)

Virtually any move was better than this lemon of a move, ChessMaster 
recommends Black try ...Bb7!? in this position.  

*****

     [  Simply terrible was: 13...cxd5?;  14.Bxd5 Rb8;  15.Nb5! Qxc2+;  16.Kxc2, 
        16...Bc5;  17.Bxc5 bxc5;  18.Nxa7,  "+/"  {Diagram?}  
        White is clearly much better and has practically a won game here. {A.J.G.}; 

***

        ChessMaster 7000  gives:  >/=  13...Bb7;  14.dxc6 Bxc6;  15.Nb5, "+/="  {D?} 
        and White has a small but clear advantage here. 

***

        While not lovely, (MAYBE) better than the game was: 13...c5!?;  {Diagram?}  
        Black tries a strategy of a complete blockade. 14.Nb5 Qb8;  15.Qc3 a6;  
        16.Nxd6 Qxd6;  17.Be2,  "+/="  {Diagram?}  with at least a small advantage 
        for White here. (Maybe "+/")  ]  

*****

14.dxc6 bxc4;  {Diagram?}  
This is virtually forced. (Now.) 

     [ 14...Qxc6??; 15.Bd5, "+/-" ]  

15.Nb5!,  {See the diagram just below.}  
This is best according to the pundits.  
(An enviable scenario! Kasparov may have several different winning lines.) 

 A completely dominating position for White. 

***

     [  I almost immediately saw: 15.cxd7!? Qxd7?!;  {Diagram?} 
        This is probably bad.  

         ( >/=  15...Bxd7; 16.Rhg1!, "+/=" {Diagram?}  and White is clearly better. 
             (Or 16.Qd2 Bg4; 17.Qxd6 Qxd6; 18.Rxd6 Bxf3; 19.Rg1, "~"  {Diag?}  
               Knaak/Schoen/Greengard/Schulz)  )   

        16.Bc5, "+/-"  {Diagram?}  with (also) a winning game for White. 
          (16.Qe2!?, "+/")   ]  

***

The next couple of moves are pretty much forced. 
15...Qxc6; 16.Nxd6 Bb7!?;  {Diagram?}  
This strikes me as second best. The computer gets into a terrible bind from 
 this position. 

"Black is hoping for counterplay against the pawn on e4." 
  - Greengard and Schulz. 

     [  I think the computer had to play for counterplay and an active position by 
        playing the continuation: >/=  16...f6!;  17.Qxc4+ Qxc4+;  18.Nxc4 Nb6!?; 
        19.Ncd2! Ba6;  20.Kb1 Rfc8;  21.Rhe1!, "+/="  {Diagram?}  
        White has a clear edge, but it is far from a won position. 

        (I spent several hours looking at this line with an Internet student who was 
         running  ChessMaster8000  on a Pentium IV computer. Our conclusion was 
         White MIGHT win, but this line is many times a better defense than what 
         Deep Junior actually played!)  ]   

17.Qc3!,  {Diagram?}  
Again this is the best move here. 

Many annotators like the ChessBase group, ... ... ... 
[Knaak/Schoen/Greengard/ and Schulz.];  ... 
praised this as the correct move here. (And I agree.) 

(Many on-line players wanted to play Nf5 ... to go for the fork of the King 
 and Queen on e7 ... but this is easily avoided.) 

     [ 17.Nf5?! Qxe4!; "<=>"  {Diagram?} 
         Black has fair play here.  ]  

17...Rae8?; (Maybe - '??')  {Diagram?}  
Why this bone-headed and very lame move - which is equivalent to 
resignation??? 

[The computer decides to give up a rook for a knight in order 
 in order not to lose the pawn on e5.] 

I think this problem is caused by an evaluation problem I have noticed that was 
purposely built into Junior.  (Many times the program will sacrifice material to 
 maintain key pawns.)  While this might make it a more interesting opponent to 
play against, and increase its commercial value; this little  'tic'  needs to be disabled 
while it is playing a chess giant like Garry Kasparov. 

*****

     [  The best defense had to be: 
         >/=  17...f6;  18.gxf6 Nhxf6; 19.Nxe5, "+/" , {Diagram?}  
         and while White might still win, this is the fault of 13...b5.  

        GM Danny King  and  GM R. Knaak  give the following continuation 
        as an improvement over the game:  17...Rab8!?;  18.Nxe5 Nxe5;  
        19.Qxe5 Qa4!?;  {Diagram?}  as the best line for Black. 
        Now White plays: 20.a3 Bc6;  21.Rd4, "+/"  {Diagram?}  
        with virtually a won game for the first player here. ]  

*****

The rest really needs no comment - 
Garry is quite capable of winning a won position.
18.Nxe8 Rxe8;  19.Rhe1! Qb5;  {Diagram?}  
Black has few good choices in the current position.

     [  Much worse was: 19...Qxe4?;  20.Rxd7 Qxf3;  21.Qxe5!, "+/-"  {Diag?}
         and if Black captures, he will be mated on the last rank. ] 

20.Nd2 Rc8;  21.Kb1! Nf8;  22.Ka1 Ng6;  23.Rc1 Ba6;  24.b3! cxb3; 
25.Qxb3 Ra8; ('?!')  26.Qxb5 Bxb5;  27.Rc7,  "+/-"  {Diagram?} 
The programming team resigns the game for Deep Junior.
(It was an embarrassment to play on against Kasparov here.) 

*****

A great game by Garry Kasparov ... one which he aptly and skillfully showed
ALL the weaknesses of the program.
(MUCH of this was probably prepared in advance!)

This game also broke a streak where Deep Junior had not lost a game to a
 human opponent in over two years.

***

Those interested in the analysis and news stories surrounding this match 
 should visit my ... "Computers vs. Humans" web site.
 (https://www.angelfire.com/fl5/ human_fan02/index.html)
This site is completely dedicated to ONLY the chess played between the 
 best human players and the best programs!

***

Copyright (c) A.J. Goldsby I. Copyright (©) A.J.G; 2003. 

  1 - 0 

   (Code initially)  Generated with  ChessBase 8.0