Banner --- Deep Thought 1 - GM G. Kasparov, Match Game 1

New York, 1989

Sicilian Defense

1.e4 c5 2.c3 e6 3.d4 d5 4.ed5 ed5 5.Nf3 Bd6 6.Be3 c4 7.b3 cb3 8.ab3 Ne7 9.Na3 Nbc6 10.Nb5 Bb8 11.Bd3 Bf5 12.c4 O-O 13.Ra4 Qd7 14.Nc3 Bc7 15.Bf5 Qf5 16.Nh4 Qd7 17.o-o Rad8 18.Re1 Rfe8 19.c5 Ba5 20.Qd3 a6 21.h3 Bc3 22.Qc3 Nf5 23.Nf5 Qf5 24.Ra2 Re6 25.Rae2 Rde8 26.Qd2 f6 27.Qc3 h5 28.b4 R8e7 29.Kh1 g5 30.Kg1 g4 31.h4 Re4 32.Qb2 Na7 33.Qd2 R4e6 34.Qc1 Nb5 35.Qd2 Na3 36.Qd1 Kf7 37.Qb3 Nc4 38.Kh2 Re4 39.g3 Qf3 40.b5 a5 41.c6 f5 42.cb7 Rb7 43.Kg1 f4 44.gf4 g3 45.Qd1 Rbe7 46.b6 gf2 47.Rf2 Qd1 48.Rd1 Re3 49.Rg2 Nb6 50.Rg5 a4 51.Rh5 a3 52.Rd2 Re2 0-1

In this game, Kasparov tries to take away Deep Thought's computational advantage by engaging in a closed game where sound chess principles rather than deep positional analysis offers the best winning chances. Deep Thought's "character" proves aggressive and it tries to bully Kasparov, an ill-advised stratagem, as you will see.

Kasparov, playing black, opens with one of his favorite openings, the Sicilian defense. Then, with move 3, he transposes into a classic French.

In moves 9 and 10, Deep Thought moves its knight twice, which violates developmental guidelines. This proves problematic for the computer, especially because Kasparov is weaving a diabolical plot.

Kasporov quietly backs up his bishop in move 10, and Deep Thought's Bd3 in move 11 is really the beginning of the end for it. Kasparov responds with Bf5, good, solid middlegame play, offering up his "bad" bishop to the computer.

Deep Thought makes a questionable move with Ra4 in move 13, pushing its rook to a position which effectively takes it out of play. Then the computer helps Kasparov out by getting rid of that "bad" bishop in move 15 Bf5, only to lose its own bishop to Kasparov's queen Qf5. Two odd tactical choices that don't really do Deep Thought any good.

Kasparov's Ba5 in move 19 is the work of a true chess genius; he sees an endgame position and sets events in motion that will leave him with a "good" knight vs. "bad" bishop endgame. Can a computer "see" that far into the game? Deep Thought certainly doesn't. Kasparov has set up a strong Kingside attack than the computer's Queenside set-up.

In moves 25 and 26, Deep Blue tries a show of strength by moving to dominate the E file, but Kasparov calmly counters in kind and further prepares to tighten the noose around his opponent's circuit board. He methodically advances on the computer, picking off its advancing pawns with his rook and knight while he removes white's other weapons. In the end, the computer is hopelessly trapped and has to resign.

In this game, Kasparov has displayed a spatial, positional vision of chess that doesn't require deep analysis of all the possible moves, captures and recaptures. Perhaps the brute force of Deep Blue's computing power may enable it to better grasp this concept -- certainly something to keep a look out for during the upcoming match in Philadelphia!

GM G. Kasparov - Deep Thought 1, Match Game 2

New York, 1989

1.d4 d5 2.c4 dc4 3.e4 Nc6 4.Nf3 Bg4 5.d5 Ne5 6.Nc3 c6 7.Bf4 Ng6 8.Be3 cd5 9.ed5 Ne5 10.Qd4 Nf3 11.gf3 Bf3 12.Bc4 Qd6 13.Nb5 Qf6 14.Qc5 Qb6 15.Qa3 e6 16.Nc7 Qc7 17.Bb5 Qc6 18.Bc6 bc6 19.Bc5 Bc5 20.Qf3 Bb4 21.Ke2 cd5 22.Qg4 Be7 23.Rhc1 Kf8 24.Rc7 Bd6 25.Rb7 Nf6 26.Qa4 a5 27.Rc1 h6 28.Rc6 Ne8 29.b4 Bh2 30.ba5 Kg8 31.Qb4 Bd6 32.Rd6 Nd6 33.Rb8 Rb8 34.Qb8 Kh7 35.Qd6 Rc8 36.a4 Rc4 37.Qd7 1-0

Garry Kasparov opens with the Queen's Gambit, which Deep Thought accepts by taking the pawn at c4. While conventional wisdom once said that the Queen's Gambit should be declined, there has been a recent shift in thought and now it is considered a viable defense. Let's see how it works out here...

Garry takes control of the center with move 3, King's pawn to e4, and although he appears to be down one pawn, he is ahead in development, which can dictate the flow of the game.

Deep Thought's choice at move 7 shows an unwillingness to engage Kasparov in a positional exchange, a good choice given Kasparov's control of the center of the board. Did the computer recognize this and avoid it, or was it simply that it doesn't weigh positional advantages as strongly in it's decision-making process?? Either way, Ng6 was the better move.

When Deep Thought responds to Garry's attack in move 8 by taking his pawn cd5, it creates a glaring disadvantage for itself, as this position remains a weakness for the rest of the game.

Move 10 sees Kasparov advance his queen to d4, a move that demonstrates his genius as his attack builds steam. Deep Thought has moved its' knight four times, violating all developmental guidelines.

Kasparov willingly sacrifices his pawn as fuel for the continued development of his attack. Deep Thought can't help but take the pawn with its bishop. This gambit is definitely a weak link in Deep Thought's programming. How do you explain to a computer that losing a pawn is actually good in certain instances? Can a computer factor in such intangible advantages as initiative and development? In this case, Deep Thought certainly shows no understanding of this, and is left with a very weak position.

Now Garry's plan starts to emerge. He mobilizes his bishop and his knight, which gets Deep Blue's queen moving to alleviate some of the pressure. Kasparov counters with his queen in moves 14 and 15.

The computer responds with a tricky tactical play at move 15 with pawn to e6, offering a trade of material, but Kasparov's attack is already too advanced, and he plays Nc7 as a response, to distract Deep Thought from his queen. The computer is forced to take the knight, and with move 17, Garry closes the noose with Bb5, putting Deep Thought in check.

Deep Thought has to block the check with Qc6, which immediately sacrifices the queen to Kasparov's bishop. Had it moved the king out of harms way, DeepThought would have been stuck in checkmate right away with a move of Qf8. Losing the queen is the only defense, and it is a crushing one.

Garry has effectively won the game right here, and he plays perfect endgame technique from here to the inevitable end. He brings his rooks to the 7th rank and wreaks havoc on the now helpless computer. This is an excellent example of Kasparov taking advantage of Deep Thought's positional weaknesses. He capitalizes on his initiative in the opening, which he established with the sacrifice of only one pawn, and mercilessly assaults the computer into submission. Well done!

GM A. Karpov - Deep Thought 1

Boston, 1990

1.e4 c6 2.d4 d5 3.Nd2 g6 4.c3 Bg7 5.e5 f6 6.f4 Nh6 7.Ngf3 O-O 8.Be2 fe5 9.fe5 c5 10.Nb3 cd4 11.cd4 Nc6 12.o-o Qb6 13.Kh1 a5 14.a4 Bf5 15.Bg5 Be4 16.Nc5 Qb2 17.Ne4 de4 18.Rb1 Qa3 19.Bc1 Qc3 20.Bd2 Qa3 21.Bc1 Qc3 22.Rb3 Qa1 23.Bc4 Kh8 24.Bh6 Qd1 25.Bg7 Kg7 26.Rd1 ef3 27.gf3 Ra7 28.Bd5 Rd8 29.Rb5 Ra6 30.Be4 Ra7 31.Bd5 Ra6 32.Rc5 Rd7 33.Kg2 Rb6 34.Bc6 bc6 35.Kf2 Rd5 36.Rd5 cd5 37.Rc1 Rb4 38.Ke3 Ra4 39.Rc5 e6 40.Rc7 Kg8 41.Re7 Ra3 42.Kf4 Rd3 43.Re6 Rd4 44.Kg5 Kf7 45.Ra6 a4 46.f4 h6 47.Kg4 Rc4 48.h4 Rd4 49.Rf6 Kg7 50.Ra6 Kf7 51.h5 gh5 52.Kf5 Kg7 53.Ra7 Kf8 54.e6 Re4 55.Rd7 Rc4 56.Rd5 h4 57.Rd3 Ke7 58.Rd7 Kf8 59.Rh7 h5 60.Ke5 h3 61.f5 Kg8 62.Rh5 a3 63.Rh3 a2 64.Ra3 Rc5 65.Kf6 1-0

In a odd twist, Deep Thought plays the Caro-Kann defense against former world champion Anatoly Karpov. The Caro-Kann is Anatoly's favorite defense; everyone associates him with it. Watch how the master of the Caro-Kann plays against the defense he knows so well.

Karpov's king pawn to e5 in move 5 establishes a strong control of the center with a long pawn chain that barricades Deep Thought's fianchettoed bishop in the corner.

All of Karpov's early pawn play is unusual, and it disrupts the computer's plan, helping Karpov to develop real dominance in the center of the board. Deep Thought opens the E file for his rook in move 8 fe5, but cannot take advantage as Karpov is busy developing the rest of the board.

The interesting aspect of computer chess is the game is played with a truly unique style that can't be predicted. They "think" differently than humans do. For example, Karpov's Nc5 at move 16 gambits his b pawn, inviting the computer to snatch it up with its queen, which it immediately does. That plays right into Anatoly's plan, clearing out that area for his queen, and his rooks - they need some room to maneuver. Does Deep Thought understand the concept of a "gambit"? How can that really be programmed?

Now Karpov begins an attack on Deep Thought's queen. In moves 18 through 22, he toys with the computer, using his bishop and rook to keep the black queen on the run, and then he swiftly moves his bishop to c4 in move 23 for a quick check.

Karpov resumes his attack with Bh6 in move 24, willing to leave his queen open for an even-up trade so he can move towards the endgame. The trade happens between moves 24 and 26, and Karpov has the desired result.

As an interesting aside, notice that Karpov is able to keep his hanging knight in play for 10 moves before Deep Thought finally takes it with move 26 ef3. After that, it seems the computer struggles with Karpov's stranglehold-style positional play. It moves its queen side rook back and forth, not seeing a clear way to advance its position.

Karpov moves with slow precision. On move 38, he centralizes his king at e3. Karpov is intent on establishing the position he desires; if the computer could see what Anatoly sees, it might choose a different tack. Instead, Deep Thought seems to be trying to steer this into a drawish rook and pawn endgame.

Karpov will not allow a draw. He slowly baits Deep Thought, and the pawns he swaps with the computer help him to achieve desired position. In move 56, Karpov shows how to use control of the center and connected passed pawns to create a threat that forces Deep Thought to compromise defense of his flank pawns. This is positional play at its best. Karpov confuses the computer as he prepares for the kill.

With move 64 Ra3, Anatoly eliminates the last pawns that threaten to reach the first rank for promotion. He secures his king with his final move, Kf6, leaving Deep Thought with no choice but to resign.

Deep Thought 2 - GM J. Polgar - Game 1 Yorktown Heights, 1993

Sicilian Defense

1. e4 c5 2. Nf3 e6 3. d4 cd4 4. Nd4 Nc6 5. Nc3 Qc7 6. Be2 a6 7. o-o Bb4 8. Nc6 bc6 9. Qd4 Bd6 10. Qg7 Bh2 11. Kh1 Be5 12. Bf4 Bg7 13. Bc7 d5 14. Rad1 Ne7 15. Na4 Ra7 16. Bb6 Ra8 17. c3 Ng6 18. Bc7 Ra7 19. Bb8 Rb7 20. Bg3 O-O 21. ed5 cd5 22. Rfe1 f5 23. Bd6 Rd8 24. Ba3 Rc7 25. Bb4 Rc6 26. Ba5 Rf8 27. Bb6 Rf7 28. Kg1 Bf8 29. b3 Bb7 30. Bh5 Re7 31. Kf1 Kf7 32. c4 Rc8 33. Ba5 dc4 34. Nb6 Rb8 35. Nd7 Rc8 36. Bb4 cb3 37. Nf8 Rf8 38. ab3 Rb8 39. Be7 Ke7 40. Rd4 a5 41. Ra4 Bc6 42. Ra5 Rb3 43. Ra7 Kf6 44. Rh7 Nf4 45. g3 Bb5 46. Kg1 Nh3 47. Kg2 Ng5 48. Rh6 Ke7 49. Re5 Bc6 50. Kf1 Rb1 51. Re1 Rb2 52. Be2 Ne4 53. Rd1 Bd5 54. Rc1 Nd2 55. Ke1 Ne4 56. Ra1 Nc3 57. Bd3 Na2 58. Kf1 Nb4 59. Bf5 Nc6 60. Rd1 Bc4 61. Bd3 Bb3 62. Rb1 Rb1 63. Bb1 Bd5 64. Rh7 Kf6 65. Rh4 Ne5 66. Ke2 Bf3 67. Ke3 Bc6 68. f4 Nf7 69. g4 e5 70. g5 Kg7 71. Rh7 Kg8 72. g6 ef4 73. Kf4 1-0

Judit opens with the Sicilian defense, establishing an aggressive tactical game. Deep Thought seems willing to play aggressively as well, initiating an exchange of knights in move 8 with Nc6, and then boldly brings its queen out to attack both Polgar's bishop and her pawn at g7.

Judit counters by removing her bishop from the line of fire to d6, which provides a launching point for an attack on White's king Bh2 at move 10. The real brilliance of her play is that when Deep Thought moves its king out of check to h1, Polgar retreats her bishop to e5, which attacks the computer's queen from a protected position (note the black queen at c7 covering the bishop!) Deep Thought's queen attack has been neutralized, and the game begins anew.

The combatants swap queens in moves 12 and 13, and advance to the middlegame on even ground. The quest now is to gain control of the center, but Deep Thought sees a weakness on Polgar's queenside, and move 15 Na4 diverts its attack to the black pawn at a6. Judith defends the pawn with her rook, and forces the computer's bishop to take refuge at g3 in move 20.

Now, Polgar takes the opportunity to castle to the kingside, and the focus again returns to the center of the board. Judith struggles to get there, but finally gains a secure position with Rf7 at move 27. From here, she endures Deep Thought's tactical pressure as she looks to maintain an adequate and flexible defense, all the while hoping to see an opportunity to counter-attack.

Move 32 c4 sees Deep Thought pressing a pawn, and Judit backs off. The combatants are clearly feeling each other out - waiting for the other to tip their hand ... then Deep Thought strikes with Nd7 in move 35, attacking both Judith's bishop at f8 and her rook at f6, while ignoring Judith's pawn at c4. Pretty advanced, even for a computer!

The sequence from move 34 to move 38 is great play by the computer, overpowering Judit. This really is the beginning of the end for her, even though the game ends at move 73! By Rh7 in move 44, Deep Thought is really in control, but Polgar has one more counter-attack left in her. The computer keeps its cool (computers don't get nervous), and successfully defends its king by move 62. Judit's last effort has failed.

Deep Thought has the game in hand, and makes sure not to let it slip away. Judit resigns in move 73, after a very impressive showing by both contestants.

---

[ The Game | The Program | Commentary | Grandstand | Deep Blue ]

[ IBM home page | (C) | (TM) | IBM Small Planet Pavilion | Internet 1996 World Expo ]