ðHgeocities.com/Baja/Dunes/4153/g4.htmlgeocities.com/Baja/Dunes/4153/g4.htmlelayedx©GÔJÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÈ@‡À]YOKtext/html° h]Yÿÿÿÿb‰.HSun, 09 Aug 1998 23:03:52 GMT Mozilla/4.5 (compatible; HTTrack 3.0x; Windows 98)en, *¨GÔJ]Y Game Four  
Game Four
Extensive analysis provided by GM Neil McDonald, IM Aaron Summerscale, IM Malcolm Pein, IM Richard Tozer and FM Chris Duncan

White: Deep Blue

Black: Garry Kasparov

1.e4  
Deep Blue sticks to the opening move which proved so successful in game 2. 
c6  
Is Kasparov going to play his first love the Caro-Kann Defence? 
2.d4  
A logical space-gaining move. 
d6  
What is this? The commentators were amazed see 2...d6. 
3.Nf3  
IM Summerscale, one of Britain's leading exponents of this defence says- 'This is not the strongest move'. This defence which is known as the Czech or Pribyl system gives White the chance to build a broad pawn centre with f4, so preferable was 3.Nc3 keeping the option of f4 open. 

John Henderson commenting on the choice of opening said: "The Prybl system is not-such a bad idea against a computer. As white, Kasparov can set the agenda and choose what he wants to play. We have to remember that Deeper Blue has a vast opening book, and Kasparov's job is to find a way around it." 

"His 3rd game was a classical example of this. By playing 1 d3 and 4 a3 it "confused" the computer, who misplaced the bishop on e7 allowing Garry to transpose into a favourable English Opening." 

"The interesting point is if Deeper Blue has such a sophisticated opening book then why didn't it notice the possibilty of a transposition into an English?" 

"However, as Black, Garry's job is much, much more difficult. The "tin box" will automatically play 1 e4. Garry doesn't want to get involved in a tactical battle with a Sicilian. Remember, in the last match that was his downfall in the first game; he tried to out-calculate a super-computer in a tactical position, which is akin to challenging a Ferrari in a roadrace with a Sinclair C5!" 

Nf6  
A developing move attacking White's e4-pawn. 
4.Nc3  
Developing and defending at the same time. 
Bg4  

 

A clever move planning to erect a pawn barrier on the white squares e6, d5 and c6. But first bringing the white-squared bishop outside the pawn chain. 

5.h3  
The normal continuation, putting the question to the bishop. Do you exchange the bishop for the f3-knight or retreat the bishop to h5? 
Bh5  
Keeping the tension in the position. 
6.Bd3  
An alternative is 6.Qe2, which is the most aggressive move in this position - preparing to castle queenside and attack on the kingside. GM Hebden from England popularised this system. 
e6  
This is the next stage in Kasparov's plan, putting his pawns on light squares now that his bishop is outside the pawn chain. 
7.Qe2  
White keeps the option of castling queenside, while playing a move that strengthens his e4-square. 
d5  
Kasparov stakes his claim in the centre, as outlined in previous moves. 
8.Bg5  
This appears to be a new move. The only move we can find in our database is 8.exd5. White counters the growing pressure on his centre by pinning the black knight on f6. (8.exd5 8...Bxf3 9.Qxf3 cxd5 10.0-0 (10.Ne2 Nc6 11.c3 a6 12.0-0 Rc8 13.Ng3 Be7 14.Bg5 0-0 15.Rae1 Ne8 16.Bc1 g6 17.Bh6 Ng7 18.Qe3 Re8 19.Re2 Bh4 20.Qf4 Na5 21.Qg4 Papaioannou,I-Oratovsky,M/EUchJ Vejen 1993/½-½ (64)) 10...Nc6 11.Ne2 Be7 12.c3 0-0 13.Bf4 a6 14.Ng3 b5 15.a3 Na5 16.Qe2 Bd6 17.Bxd6 Qxd6 18.f4 Nc4 19.Rf3 Rfe8 20.Raf1 Rac8 21.R1f2 1-0 Chernyshov,K-Maiwald,J/Dresden op 1995 (39)) 
Be7  
Breaking the pin. 

9.e5  
A commital move, but one which side-steps any problems with Black playing Nxe4. Not 9.0-0? which is a good example of the type of tricks waiting for Deep Blue: 9...Nxe4!! 10.Bxe7 Nxc3 11.Bxd8 Nxe2+ 12.Bxe2 Bxf3 13.Bxf3 Kxd8 Leaving Black with a winning position. 
Nfd7  
Black is preparing to break back in the centre with c5. 
10.Bxe7  
Removing what is commonly known as Black's good bishop which is not obstructed by any pawns. However exchanges generally help Black in this position since he has less space for his pieces than White, therefore it is possible that the plan of 8.Bg5 is inaccurate. 
Qxe7 
11.g4  
An aggressive move from the computer, planning to castle queenside. 
Bg6  

12.Bxg6  
A double-edged move. 'This is dubious, better would have been 0-0-0, followed by h4 and h5 forcing ...Bxd3 when the recapture cxd3 strengthens White's centre.' - GM McDonald. 
hxg6  
Black now has a far more compact pawn formation on the kingside since White's pawns on g4 and h3 look a bit loose. 
13.h4  
White starts a far-sighted plan to open up the kingside. 
Na6  
Preparing to castle queenside. 
14.0-0-0  
Not only castling the king into safety, but connecting the rooks. 
0-0-0 
15.Rdg1  
'I prefer Black in this position' - IM Summerscale. 'I think it is roughly level' - IM Tozer. 
Nc7  
A intelligent re-routing of the knight. However Kasparov spent approx. 35 minutes on this move! 
16.Kb1  
A standard defensive move, removing the White king from the c1-h6 diagonal. 
f6  
Beginning the attack on White's centre. Time left Kasparov 52mins, DeepBlue 1 hour 34 mins. 
17.exf6 
Qxf6  
'This is very surprising, I expected 17...gxf6.' - GM McDonald. Although if here 17...gxf6 then 18.g5 Rdf8 and Black is doing well. 

Joe Hoane from the DB Team commenting said: "The computer will never play perfect chess. That's not the goal of this project, because it's not possible to play perfect chess. In the last six months, the computer has gone from playing [human-type] chess every once in a while to playing chess almost all the time."

18.Rg3  
White anticipates pressure along the f-file and so bolsters his knight on f3. 
Rde8  
Preparing to blast open the centre with 19...e5. 
19.Re1  
Deep Blue has 1:24 to Garry's 0:49. 

 

'Black still has a clear plus' - IM Summerscale. Laying siege to the e5-square and so preventing the e5-break. Deep Blue wants to gain solid contol of the position and then try to exploit some structural weaknesses in Black's kingside. However White also has potential weaknesses in his position e.g. h4, f4 and the f-file in general. 

Rhf8  
Another option was 19...Re7 with the idea of doubling on the e-file and trying to play e5. 
20.Nd1  
Hiarcs thinks White stands better. 
e5!?  
'Bizarre, Kasparov has gone beserk' - GM McDonald. 'An interesting positional pawn sacrifice' - IM Summerscale. 'Why didn't Kasparov play 20...Qf4 first' - IM Tozer. 
21.dxe5  
The computer is only too willing to accept material. 
Qf4  
Black has sacrificed a pawn, hoping to recover it in the next couple of moves, when his superior co-ordination of pieces will give him the advantage. But it is not so clear how he can in fact regain the material since the e5-pawn is heavily fortified. Kasparov is clearly speculating. The whole match may be decided over the next couple of moves. 
22.a3  
Kasparov is shaking his head. The World Champion cnnot believe this move. Has the computer got no respect? Instead 22.h5 looks much better, when if 22...g5? 23.Qe3 attacks a7 and the g5 pawn is doomed. 
Ne6  
This is the perfect blockading square for the knight, however it cuts off the black rook's attack on the e5-pawn. 
23.Nc3  
Re-centralising the knight. 
Ndc5  
Black's pieces are finding good squares, but does he have enough compensation? 
24.b4  
'A Terrible positional move, ask the c4-square what it thinks about the move b4.' - IM Summerscale. 'A human would never play that.' - IM Tozer. 
Nd7  
Kasparov retreats his knight and is delighted to hav provoked the rash b4 move. Perhaps Deep Blue should play 25 Qe3 offering the exchange of queens which it would be difficult for Kasparov to refuse in view of the attack on a7. Then the weakness of his king's position would cease to be of such significance. 
25.Qd3  
Attacking g6. 
Qf7  
Defending f6. 
26.b5  
Trying to justify b4, but it cedes the c5-square to a black 
Ndc5  
Kasparov jumps in at once, taking advantage of White's last move. 
27.Qe3  
Also possible was 27.Qd2. 
Qf4  

 

Garry offers to trade queens since he is possibly afraid to cross swords with the computer in a sharp middlegame especially with time-trouble looming. Once the queens are exchanged White will have strong pressure down the f-file to compensate for the pawn deficit. 

28.bxc6  
It is not possible for Black to exchange queens yet due to the intermezzo 29.cxb7 check! 
bxc6 
29.Rd1  
Preventing any tricks with d4. 
Kc7  
Kasparov now has 19 minutes for 11 moves, while the computer has 51 minutes. Kasparov anticipates an endgame and so improves the position of his king. This quiet move also clears the b-file for the rook and gives Deep Blue the chance to make the positional blunder 30.Qxf4 Rxf4 when the f2-pawn is very weak. Kasparov now has 19 minutes for 11 moves, while the computer has 51. 
30.Ka1  
Removing the White king away from any danger on the open b-file. 
Qxe3  
'I'm very suspicious of the White e-pawns' - IM Summerscale. If 30...Qc4 then 31.Nd2 Qa6 32.Ne2 with a slight advantage for White. 
31.fxe3 
Rf7  
Preparing to double on the f-file. 
32.Rh3  
'I can't see the point to that move, surely there is something more constructive. I think White should have played h5 at some point to break things open on the kingside.' - GM McDonald. 
Ref8  
Intending to invade along the f-file. White's position is beginning to creak. 
33.Nd4  
White's best chance of safety is to challenge the knight on e6. 
Rf2  
The invasion begins. 
34.Rb1  
Can the computer generate enough counterplay on the queenside. 
Rg2  
Threatening to double rooks on the seventh rank which in combination with Nxd4 would win the c-pawn, when White's position would crumble. 
35.Nce2  
Deep Blue prefers to sacrifice the g-pawn, rather than the disastrous position given in the note above. 
Rxg4  
Kasparov regains the pawn and leaves White's pawn structure shattered. 
36.Nxe6 
Nxe6 
37.Nd4  
This is the best fighting chance to gain some counterplay in a rook ending. As we know 'All rook endings are drawn'. 
Nxd4  
We feel that Kasparov has at least a draw in the coming rook endgame with good winning chances. The key to his position is gaining passed connected c- and d-pawns. If he is able to achieve this he has tremendous winning chances. 
38.exd4 
Rxd4  
Picking up the loose d-pawn. 
39.Rg1  
Deep Blue wants to capture both the g-pawns and place a rook on the seventh rank. 
Rc4  
Garry is optimising the activity of his pieces, while trying to gain the key connected c- and d-passed pawns. 
40.Rxg6 
Rxc2 
41.Rxg7+  
The pawns are coming off, but the passed c- and d-pawns are stronger than the three isolated white pawns. 
Kb6 
42.Rb3+ 
Kc5!  
A much better move than 42...Ka6, which would hold the a7-pawn. The key to the position is keeping the key active. 
43.Rxa7 
Rf1+  
A strong move forcing White to place his rook passively on the back rank. 
44.Rb1  
The Deep Blue team offered Garry a draw at this moment, Garry just laughed and said 'let's play chess'. 
Rff2  

 

Garry can now force a draw at any moment with perpetual check from the rooks along the second rank. But Garry undoubtedly wants more. 

45.Rb4 
Rc1+ 
46.Rb1 
Rcc2  
If instead 46...Rxb1+ then 47.Kxb1. 
47.Rb4  
Is Garry content with a draw? 
Rc1+  
Alternatively 47...Ra2+ 48.Kb1 Rxa3 49.Rxa3 Kxb4 is another way to try and play for a win. 
48.Rb1+  
No! Garry will play for the win. 
Rxb1+ 
49.Kxb1 
Re2 
50.Re7 
Rh2 
51.Rh7 
Kc4  
Not as good is 51...d4. It is better to keep the king in front of the pawns. This will enable Black to threaten back rank checkmate as well as trying to promote his pawns. 
52.Rc7 
c5 
53.e6 
Rxh4 
54.e7 
Re4 
55.a4 
Kb3 
56.Kc1  
Draw Agreed 
½ ½
ÅðéóôñïöÞ óôçí áñ÷éêÞ óåëßäá ãéá ôï óêÜêé