An analysis of the second game. (# 2.) 

Deep_Junior (C) (2499) - GM Garry Kasparov (2845) 
[B20]
X3D "Man vs. Machine" Match 
New York City, (NY/USA) (Game # 2),  28.01.2003

[A.J.G.]

Game Number Two of this match.

Kasparov darn-near wins his second game in a row.

Kasparov plays a truly brilliant and very inspired attack, but 
then falters when he almost has the win in hand.

1.e4 c5;  {Diagram?} 

The basic idea of the Sicilian is not so much to equalize completely, as it is 
to introduce a dynamic imbalance into the position. 

 (A counter-attack in a asymmetrical pawn structure.)

2.Nf3 e6;  3.d4 cxd4;  4.Nxd4 a6!?;   {See the diagram just below.}    

This is the older "Paulsen/Kan" Variation. 
This was popular for a great many years. 

This line got played quite a bit in the 1960's at the GM-level, but then was pretty much shelved after Bobby Fischer waxed Tigran Petrosian in this particular variation.

     [  Black could also play: 4...Nc6!?; "~"  {Diagram?}  
         which is the popular line known as: 
         "The Taimanov Variation."  ]  

The position after 4...a6.

5.Bd3, ('!')  {Diagram?} 

According to several books that I have, this is the correct way to play this line ... several theoretical tomes - daunting because of their sheer size - even award this move an exclamation point.

Of course White can also play Nc3 or even c4. 
(And virtually any other move which conforms to the principles of 
 correct play of the opening phase of the game.)

      [  White could try:  5.c4!?,  {Diagram?}  
          with a bind on the center. 

***

          White could also play:  5.Nc3 Nc6;  6.Be2 Qc7;  7.0-0 Nf6;  
          8.Be3, "+/="  {Diagram?}  with over 1000 examples (!) in the 
          ChessBase on-line db.  (White has a slight edge.) 

          GM Sergey Dolmatov (2605)  -  WGM Zsuzsa Polgar (2535);  
          Rubinstein Memorial, 1991. (1-0, 71 moves.)  ]  

5...Bc5!?;  {Diagram?}  

This is a line which is out-dated and superseded ... at least according to 
several (older) chess books that I have.

But this is THE correct way to play against a computer. Even a line which is 
slightly inferior is playable, as long as the way to a clear advantage is not one 
which the program can easily find!

*****

     [  Black could play the line: 5...Nc6;  6.Nxc6 bxc6;  7.Nd2!?,  {Diag?}  
         This is probably the safest way for White to continue. 

           ( Another way to play this line - for the first player - is:   
             7.0-0!? d5!?;  8.c4! Nf6; 9.cxd5! cxd5;  10.exd5 exd5;   
             11.Nc3! Be7;  {Diagram?} Black is trying to defend.    
             12.Qa4+! Qd7!?; ('?!')  This could be dubious.   
             13.Re1!!, "+/="  {Diagram?} & White has virtually refuted    
              Black's whole approach to this line here.   

              GM Robert J. ('Bobby') Fischer - GM Tigran Petrosian;    
              FIDE Candidates Match, (final) (Game Number Seven {7})  
              /Buenos Aires, ARG/1971.  (1-0,  34 moves.) )     

        7...e5!?;  {Diagram?}  This is the modern way of playing this line.

          (Black could also play: 7...d5!?;  8.c4!? Nf6; "~"   {Diagram?}    
          and the second player has made a strong point out of the   
          d5-square.)    

        8.Qh5!?,  "/\"  {Diagram?}  and according to several different 
       opening books, White has a very strong initiative from this position.
       (Several books give Qh5 an exclam here.) 

        [ See MCO-14; page # 309, and column # 3. (All notes.) ] 

***

        The main line for Black is the following continuation; which is the most 
         modern of all the choices - for both parties here: 

        5...Nf6!?;  6.0-0,  {Diagram?}  
        White should simply develop.  

          ( A common trap is: 6.e5? Qa5+;  7.c3 Qxe5+;  8.Ne2,  {Diagram?} 
            Probably better is Qe2, but now down a pawn - White does not wish 
            to  trade Q's.  

              (Possibly 8.Qe2!?)    

            8...Qc7!;  {Diagram?}  Getting the Q to safety.  
            9.0-0 d5;  "/+"  {Diagram?}  and Black is a clear pawn ahead. )  

        6...Qc7;  "~"  {Diagram?}  with rather unclear play. 

        M. Adamski - M. Marszalek;  
        National Championship. (Semi's)  
        Lubniewice/POL/2002.  (38)  

        MCO quotes the game, GM V. Anand - GM L. Ljubojevic; 
        Linares, 1993.

         [ See MCO-14; page # 309, column # 2, and also note # (i.). ]  

        (Continuing our look at this line.)   
        7.Qe2 d6;  8.c4, "+/="  {Diagram?}  
        White has a very solid space advantage here. 

        There were literally hundreds of examples on the ChessBase web 
        site. I always look for a recent, high-quality contest, (no short draws); 
        between two good (titled) players. 

        The best example I could find was the encounter:  
        GM Z. Almasi - GM J. Dorfman; 
        National Champ. Tourney, #0203. Austria, 2002. 
        (Drawn {1/2-1/2} in 51 moves.)  ]  

*****

6.Nb3 Ba7;  7.c4!?,  {Diagram?}  

White attempts a complete lock on the center. 
(This is a book line, and guarantees White a small but 
 steady edge in the opening.) 

Several opening books recommend Nc3 here as being a little better.

     [  One very strong program - with its opening book turned off - thinks 
         for a long time and recommends the move:  7.Qg4!?,  {Diagram?} 
         which might be playable.  

         Instead  MCO  offers the line: >/=  7.Nc3! Nc6;  8.Qe2! d6;  9.Be3, 
         9...Bxe3;  10.Qxe3 Nf6;  11.f4! Qc7;  12.g4!?, "/\"  {Diagram?}  
         (Maybe - '!')  with a strong initiative to White.  - GM Nick de Firmian. 

         (MCO considers this to be just slightly better for White, or gives the 
          symbol of:  "+/=" ... {"plus-over-an-equal-sign"} meaning White is 
          just slightly better.)  

         GM W. Browne - Ramirez;  Arizona, (USA) 1990.

         [ See MCO-14, page # 309, column # 4, and also note # (o.). ]  ]   

Now both sides continue to try and develop in pretty much a normal manner here. 
7...Nc6;  8.Nc3 d6;  9.0-0 Nge7!;   {See the diagram just below.}     

The line with ...Nf6 is probably also very playable, but Kasparov has something 
very specific in mind for this game. 

The game position after Black's ninth move.

In this game, it will be a key idea for Garry to play ...e5; and then ...Nd4. 
If this happens, then the other Knight can bolster its compatriot on d4 by playing
...Knight from e7 to c6. 

     [  Another line is: 9...Nf6;  10.Qe2!? 0-0;  11.Be3 Bxe3;  
        12.Qxe3 Ne5; "~"  {Diagram?}  with unclear play. 

        For example, see the encounter: GM B. Ivkov - GM R. Huebner;  
        Palma de Mallorca/ESP/1970.

        See also: GM A. Ivanov - A. Stripunsky;  Bradley Open, 
        Windsor Locks, 1998. (1-0, 29.)  ]  

Both sides continue to develop. 
10.Re1 0-0;  11.Be3 e5;  {Diagram?} 
Black nails the center ... before White can play f2-f4. 
(Now an immediate advance of f2-f4 by White could result in
an isolated e-pawn for White, after Black responds ...e5xf4.)

     [ 11...Ne5!? ]  

12.Nd5 a5!;  {Diagram?}  

An excellent idea ... and a common one in certain lines of the Sicilian. 
(Black wishes to gain a tempo off of the White Knight at b3. And a4 
 hands the b4-square firmly over to the second player.) 

     [ 12...Bxe3!? ]  

13.Rc1,  {Diagram?}  

Around this point, Mig Greengard and GM M. Ashley were saying that White 
had dissipated most of his advantage with too many 'normal'  moves. 

     [ 13.h3!? ]  

13...a4;  14.Bxa7 Rxa7;   {See the diagram just below.}    

Times: White - 1:36;  Black - 1:11 

The actual game position after Black's fourteenth move. 

The correct way to play. Black's position looks passive, 
but it is also very resilient.  

15.Nd2 Nd4;  {Diagram?}  

As advertised, Black claims his outpost.

     [ About the same was: = 15...a3;  16.b3, "~" ]  

16.Qh5!?,   {See the diagram just below.}    

An extremely aggressive move, but there is also the danger that the 
White Queen will become may become misplaced. 

Perhaps White could have tried this approach much earlier in the
game.

The game position after White plays Qh5.
(White has the makings of a possibly strong K-side attack.)

     [ 16.g3!? ]

16...Ne6; ('!')  {Diagram?} 

Now Black possibly menaces the f4-square in some lines. 
(Forking The Queen and Bishop.) 

     [ 16...h6!? ]  

17.Rc3, (Almost - '!')  {Diagram?}  

A cute move (rook luft), almost worthy of an exclam in this position.
In some lines - if Black is not real careful - White will play Rh3 ...
with a very dangerous attack on the Black King.

     [  White could try:   17.Bb1!?, "~"  {Diagram?}  
         with approximately level chances.   

***

        To show this position is fraught with danger ... (for BOTH sides!); 
         ... I offer the following line:  
        17.f3!? Nxd5;  18.cxd5??,  {Diagram?}  
        To open the c-file. 
         (But the move is a mistake.)

           (Correct was:  18.exd5, "=")   

        18...Nf4;  "-/+"  {Diagram?}  and Black wins a piece. ]   

17...Nc5;  {Diagram?}  

A near ideal square for the Black Knight here. In some lines, especially 
if White is not careful, Black can trade down to a  'Good Knight versus 
a Bad Bishop' end-game. 

The other purpose for this move is now the Black QB covers h3, and stops 
the Rook from coming over to the h3-square. 
(To use the Q+R combo to threaten the Black King ... mate on h7.) 
  {After the move,  17...Nc5; {D?}  the h3-square is covered, and 
   White is currently unable to pace a Rook there and threaten
   the Black King.}  

     [  After the continuation of: 
        </=  17...Nxd5?!;  18.exd5 g6;  19.Qh6 Nf4;  20.Ne4,  "--->"  {Diag?} 
        (Probably  "+/=")  
        and White has the makings of a nice King-side attack.  ]  

18.Bc2 Nxd5!?;  {Diagram?}  

This looks a little risky, but Garry seems to have everything 
under control.

     [  Or  18...Bd7;  19.Rf3, "+/="  {Diagram?}  
         and White probably has a slight advantage here.  ]  

19.exd5 g6;  {Diagram?}  

This looks weakening, but is actually a maneuver designed to blunt the 
 power of the White Bishop on c2.  

     [  Even more dangerous is:  19...h6!?;  20.Rg3, "--->"  {Diagram?} 
        White will have a solid attack from this position. ]  

20.Qh6, ('!?')  {Diagram?}  

This looks aggressive, but the Queen is actually slightly out of play here. 
 (Its mobility is limited, and if the position requires a fast maneuver on 
  another part of the board, White will find this very difficult to do.)  

     [  Maybe slightly safer than the game would have been: 
        20.Qd1 Qb6, "="  {Diagram?}  with approximately equal play.  ]  

20...f5;  {Diagram?}  

Times: White - 1:02;  Black - 0:39

Now in some lines, Black's Rook comes in quite handy on the second 
rank!

     [  Black could (also) try:  20...Qc7!?;  {Diagram?}  
         with fair play.  

        or  20...Bd7; "~"  {Diagram?}  also with a reasonable 
        position for Black. ]  

21.Ra3!?,   {See the diagram just below.}   

A typical computer move, after (seemingly) extensive King-side preparations, 
the box suddenly switches to the Queen-side. 
(It thinks it can win the a-pawn.) 

(The actual game position after the program plays its 21st move.) 

This type of constantly flip-flopping ('wishy-washy') strategy is what used to 
 kill the computers when they played really good players. 
(Here DJ gets away with it only because Garry does not follow up the 
 attack correctly.)

      [  A very reasonable line would have been the continuation: 
         21.Nf3 Qd7;  22.Ng5, "+/="  {Diagram?}   
         when most computer programs award a small edge to White. 

***

         Even less convincing was:  21.Rh3!? b5!;  22.cxb5 Qa5; 
         23.Rc3 Qxb5; "<=>"  {Diagram?}  and I would say Black 
         has very good play on the Queenside. 
         (He also controls a bit more space than White does. 
          And to be completely honest, most strong programs still 
          award White a very small edge here.)  ]  

21...Qf6!?; (Maybe - '!!')  {Diagram?} 

I commented right after this move was made, that  - - -
"This is either bad  ...  or very, very, very good." 

Typical Kasparov, he prepares to sacrifice material for an attack.  

     [  Several players and computers pointed out that Black could 
        also play the safe:  21...Ra6; "="  {Diagram?}  
        with a very solid position.   

***

        Awkward for Black might have been:  21...Qb6!?;  22.Rb1!,  "+/="  {D?}  
        White is a little better here. 

        (One of my students in Ohio spent hours in this position, and even pitted 
         Fritz 7.0  against  ChessMaster 9000  in this position. I had originally 
         planned to include some of the highlights of this deep investigation, but 
         lost the file when my computer crashed.)  ]   

22.b4!?,  {Diagram?}  

The computer "thinks" it is winning a pawn here, but it is not really 
 prepared for what happens next. 

     [  The same idea is:  22.b3!?,  {Diagram?}  
         and Black could respond exactly the way he does in this game. 

***

         Maybe better was:  22.Nf3!?, "+/="  {Diagram?}   
         when White seems no worse. 

***

         Totally passive would be:  22.Nf1 e4;  23.Qc1!?, "="  {Diag?} 
         when White risks ceding the initiative over to Black.  ]   

22...axb3; ('!')  {Diagram?}  

This is virtually forced now, the passive retreat of the Knight to d7 
 loses a pawn for nothing.  

     [  A completely doubtful idea is:  22...Nd7?!;  23.Rxa4, "±"  {D?}  
         ("+/")  when White has won a free pawn in this position.  ]   

********************

Now White is forced to capture the Rook. (axb3??, Rxa3; "-/+") 

23.Rxa7 bxc2;   24.Rc1,   {See the diagram just below.}      

One commentator said this was virtually forced.  

The game position just after White plays 24.Rc1.

     [  White could also try: 24.Nb3!?, "~"  {Diagram?}  
        which might also slightly favor the first player here, although that is 
        not 100% certain.  (Black would continue with ...f4!?)  ]

24...e4;  {Diagram?}  

This is probably the correct way to play this position, although Black 
 obviously had other ways to pursue his attack here. 

     [  Black could also play:  24...f4!?; "~"  {Diagram?}   
         with a strong attack. {Unclear?} 

***

        But not:  </=  24...Bd7?!;  25.Rxc2, "±"  {Diagram?}  
        and White is clearly better. 

***

        One titled commentator (IM) also suggested that:  
         24...Nd3;  {Diag?}  was possible as well. ]  

25.Rxc2,  {Diagram?}  

This is close to being forced. 

White could easily lose this position, as the next few moves show. 

Most human opponents would be sweating bullets here ... but Deep Junior 
does not have any such problems. 

     [  I was hoping the box might fall for something like: 25.Nb3!?, 
         25...Nxb3!?;  {Diagram?}   Very tricky, but ........  

          ( >/= 25...f4!; "<=>" )  

        26.axb3 Qd4;  27.Qe3 Qd3; 28.Ra2 f4;  29.Qxd3?! exd3;  "/+"  
        (Maybe  "-/+")  {Diagram?}   
        when Black is much better. 

         (White should have played Qe1 on his twenty-ninth move.)  ]  

25...Qa1+!?;  {Diagram?}  

Thinking it does not matter, Kaspy plays a typical human move, and 
 throws in a check. But this may not be the most accurate way of 
 pursuing the attack here. 

Several annotators have already given this move a full question mark, 
('?') ...  but until a clear and irrefutable win is found for Black, I think that 
is a bit premature.  (And unfair.) 

*****

     [  Probably the best line is:  >/=  25...f4!;  26.Nf1,  {Diagram?} 
        This could be forced.  (The alternatives look worse.) 

***

           a).  Much worse would be: </=  26.h3?! Qa1+;  27.Kh2,  {Diag?} 
                  Now ... this is forced. 

                   (Simply ludicrous would be:  </=  27.Nf1? e3;  28.fxe3 fxe3;    
                    29.Qxe3 Qxf1+;  30.Kh2 Bd7; "/+" {Diagram?}     
                    One GM reached this position in his analysis and then he    
                    claimed, "But White should be able to draw this."  ???     
                    Maybe against a  'Class C'  player, but not against Garry Kasparov!!!)     

                 27...Qd1;  28.Rb2[]  Qc1!;  29.Rbxb7 Nxb7;  30.Nxe4 Qxc4;  
                 31.Ng5 Qc7; "/+"  {Diagram?}  
                 and Black is clearly MUCH better. (If not winning.  Maybe "-/+".)  

           b).  =  26.Ra8!? e3;  27.Nf1,  {Diagram?}  
                 This is probably forced. 

                   (</=  27.fxe3? fxe3; 28.Qxe3 Qa1+; 29.Nb1 Nd3!!; "-/+")    

                 27...Qf5!; "=/+"  (Maybe - "/+")  {Diagram?}  
                 and Black is very clearly better in this position.   

***

        (Returning now to the main {analysis} line here.) 
        26...e3;  {Diagram?}  
        Black must try to open lines in this position.

        27.Ra8!?,  {Diagram?}  
        This could be forced.  

           (Not  27.Nxe3!? Bf5!; "=/+")    

        27...Qf5!; "~"  {Diagram?}  (Probably  -  "=/+")  
        Black has tremendous compensation in this position.  

        (This is the line I predicted Black would play - on one chess server. 
         {I was one of the official commentators.} 

        I searched this position for hours, but never found a forced win. 
        But Black is definitely better!)  ]   

*****

The next couple of moves are all forced. 

26.Nf1 f4;  {Diagram?}  

The correct way (now) of pursuing the attack for Black. The pawns on e4 
and f4 form an attacking wedge, and threaten to advance and destroy 
White's position. Meanwhile White's forces are scattered far and wide 
over the whole of the chess board, and have very little co-ordination. 

     [ </=  26...Nd3?!;  27.a4, "±" ("+/")  ]  

27.Ra8!,   {See the diagram just below.}   

The program cleverly finds the ONLY defense in this position. 
 White can give back material, and emerge with a decent position.  

(The position just after White played Ra8!, 
 on his twenty-seventh move.) 

      [  To show how dangerous White's position can become, I offer the 
          following plausible line: </=  27.a4!? e3;  28.Qxf8+,  {Diagram?} 
          This is pretty much forced now for the first player here.  

            (</=  28.fxe3?! fxe3;  29.Qxf8+ Kxf8; 30.Ra8 Qd1; "-/+")    

         28...Kxf8;  29.fxe3,  {Diagram?}  
         This is pretty much forced.  

            (</=  29.Ra8? Qd1; "-/+")   

         29...Bf5;  30.Re2 Bg4;  31.Rf2 Nd3; "-/+"  {Diagram?}  
         Black is winning easily here.  

          (If  32.Rd2?!, then 32...Qe1!; {"-/+"}  so that if  33.Rxd3,  
           then  33...Be2;  is completely decisive.)  ]  

27...e3;  28.fxe3 fxe3;   {See the diagram just below.}    

Suddenly Black has dire threats, but the computer can give back 
the Queen to help eliminate many of Black's attacking ideas here. 

 The position after Black just played 28...fxe3. 

29.Qxf8+, {Box}   {See the diagram just below.}    

No exclam here, if White does not play this - the program will get shredded.  

 The position after White played his twenty-ninth move here. 

     [  An example of what happens if White should choose  NOT  to give up 
         his Queen is:  29.h3? Rxf1+;   30.Kh2 Rh1+;  31.Kg3 Qe1+;   32.Kf4, 
         32...Rf1+;  33.Rf2 Qxf2+;  34.Kg5 Ne4#.   

***

         Several players on the Internet suggested the move: 
         29.Rxc8??,  {Diagram?}  .....  missing the threat!  

         Of course Black does not recapture, but simply plays: 
         29...Qxf1#; Check-Mate!  ]   

29...Kxf8;  30.Rxc8+ Kf7;   {See the diagram just below.}     

Draw agreed.

I don't know who offered a draw here, but the position is finely
balanced. (And the material is close to equal, with White having 
two Rooks for the Queen, but a totally unorganized position.)

White has an easy draw by  "perpetual check"  ...  to which
Black has no defense. 

Nor can White afford to take his time or dilly-dally about. Black 
has many threats, such as ...Qd1; and then ...e2, winning. 
Additionally the second player has such ideas as ...Ne4-d2; and 
even threats to the White a2 and c4-pawns. 
(Should White move his Rook.) 

In the line given, (see just below); the Black King must prevent the 
penetration/activation of the White Rooks. (In a long ending, the 2 
Rooks are generally better than a single Queen.)  

My thanks to one of my (California) Internet chess students who saved 
the entire transcript of  "Chess-FM's"  broadcast of this event and 
mailed it to me. 

The final position of the game. 

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

     [  The continuation of:  30...Kf7;  31.Rc7+,  {Diagram?}  
        This looks forced for White. 

          (It is also a draw after:  31.Re2 Ne4;  32.Rc7+ Kf8;  33.Rxe3 Nd2;     
           34.Rc8+ Kf7;  35.Rc7+ Kf8; "="  {Diagram?}     
           which is a draw by the ...  "3-time repetition of the position" rule,   
           according to FIDE.  

           Very bad is: 31.Ra8?? Ne4!; "-/+" {Diagram?}  
            with the winning threat of ...Nd2.   

            A simple blunder is: 31.Rh8?? Qxh8; "-/+" {Diagram?}   
            and Black should win.)    

        31...Kf8;  {Diagram?}  
        The King continues to guard the e7 and f8 squares. 
        32.Rc8+,  {Diagram?}  
        This is forced.  

          ( Losing for White is:  32.Rxh7? Ne4!;  33.Rh4,  {Diagram?}  
            This is probably now White's only chance in this position.

             (33.Rc7? Nd2; "-/+")   

            33...Nd2;  34.Rf4+ Kg7;  35.g3 g5!;  36.Rf5,  
            36...Qb1!; ("-/+")  {Diag?}  and Black wins material. )  

        32...Kf7;  {Diagram?} 
        This looks forced as well.  

             ( But not:  </=  32...Kg7?;  33.Re8, "+/="  {Diagram?}    
                and Black loses his important King's Pawn.    

                (Without this pawn, Black has almost no threats.    
                 Black risks losing because if White consolidates his position,   
                 the two Rooks are probably better than a Queen.) )   

        33.Rc7+ Kf8;  "="  {Diagram?} 
        is a fairly obvious draw here. 

          (Much too risky is:  33...Kf6?;  34.Re2, "±"  {Diagram?}  
           and all the programs agree that Black is in trouble.)   ]  

  (Code initially)  generated with  ChessBase 8.0  

  ½ - ½