- The Washington Times - Saturday, May 18, 2002

Like a pair of heavyweight prizefighters fated to an endless series of rematches, man and machine continue to test their prowess at the chessboard.

The casual player may have thought that Deep Blue's celebrated defeat of then-world champ Garry Kasparov in Philadelphia in 1997 marked the definitive triumph of chess-playing computer programs over humanity, but recent matches suggest the question of ultimate superiority remains in doubt.

Computers remain peerless in tactical calculating ability, but top players are developing more sophisticated strategies to steer play into positions particularly closed games featuring long-term maneuvering where the silicon monsters are still weak.

Two recent matches featuring top programs and strong GMs offer a rough idea of the balance of power these days.

America's Boris Gulko, who starred in the celebrated Harvard Cup man-vs.-machine competitions of the early 1990s, came up short in an 8-game match against four top programs in March, struggling to win just one game in a 5-3 defeat.

But Israeli GM Ilya Smirin restored a bit of humanity's amour-propre with his own 5-3 victory over a collection of silicon monsters a month later. We have games from both events.

The emotionless computers paradoxically enjoy a huge psychological edge: Even the greatest players are afraid to enter complicated, unclear tactical lines against them. Kasparov radically altered his style with disastrous results in a bid to stay away from Deep Blue's strengths.

Gulko after his match compared playing strong computer programs to paddling down an uncharted stream, where the player seeks to avoid the rapids at all costs, even rapids he feels in his bones he can handle. Smirin noted that such a strategy is even more difficult to pull off when playing Black, as the human player often accepts a markedly inferior set-up just to avoid an open game.

The program Deep Shredder shredded Gulko's attempts to head for calm waters, dealing the grandmaster his most painful defeat of the match. In positions that play to their programming powers, computers have the ability to make their human counterparts look foolish.

Gulko's French variation with 5. Bd3 c5 6. 0-0 c4!? is a clear anti-computer tack, but does not succeed in keeping the center closed for long. Deep Shredder plays aggressively here, offering two knights for a rook and pawn in order to open lines to the Black king.

Black's poky development and strange shuttling of his queenside pieces are the kinds of moves Gulko ordinarily would not be caught dead doing, and the computer makes him pay.

Having opened the f-file for its attack, Deep Shredder brilliantly exploits his opponent's inaccuracies with a flurry of sacrifices: 19. Qe2 Re8?! (the wrong square, but Black's set-up was already suspect) 20. Bh5+! g6 (weakening, but 20…Nxh5 21. Qxh5+ g6 22. Bg5+ Kg7 23. Qxh6+ Kg8 24. Rf7! Kxf7 25. Qh7+ Kf8 26. Bh6 is mate) 21. Bg4 Nxg4? (Black hastens to remove an attacker, but now the White assault becomes irresistible; better, though not good, was 21…h5 22. Bh3 Nd8 23. Be5 Rf8 24. Rae1) 22. Qxg4 Kg7 23. Bh6+!!.

A beautiful sacrifice, but child's play for a silicon chip. Black's moves are forced: 23…Kxh6 24. Qh3+ Kg7 (Kg5 25. g3 and mate in a couple) 25. Rf7+!!. If now 25…Kxf7 (Kg8 26. Qh7 mate), White mates with 26. Qh7+ Kf6 (Kf8 27. Rf1+ Bf6 28. Rxf6+ Qf7 29. Qxf7 mate) 27. Rf1+ Kg5 28. h4+ Kg4 29. Qxg6+ Kxh4 30. Rf4 mate. Gulko resigned.

Smirin's victory over Gambit Tiger 2 may rate as one of the most impressive games of the year, as the Israeli GM rescues a weak position with a farsighted tactical combination that the computer grasps only too late.

Smirin as White again tries to keep things quiet, but drifts into an inferior position in this King's Indian, with Gambit Tiger enjoying a clear initiative on the queenside. The computer correctly judges that its doubled isolated b-pawns are amply compensated by open lines and piece activity on the queen's flank.

Computers typically punish desperate attempts at counterplay, but this time the human prevails: 21. Qd2 Rxa5 22. h4! (it would take a very farsighted program or an annotator who knows how the game turns out to predict here that this pawn thrust will prove decisive) Kf8?!, one of those hard-to-explain moves to which even the best programs are occasionally prone.

White's strategy is risky, dangling heavy material as bait on the queenside to carry out a kingside attack. But the computer's apparent inability to foresee the long-term danger posed by the h-pawn does it in.

Thus: 26. Qc7 Bc3? (the first step into the quicksand) 27. Qxb6! (eyeing the Black f-pawn; on 27. Qxd7? Bxd7 28. Ra2 Rxa4 Rxa4 30. Rb1 Ra5, White has no compensation for what will soon be a two-pawn deficit) Qe7 (Bxa1 28. Rc7 Qd8 29. Qc5+ Kg8 30. Ng5! Bb3 [fxg5? 31. Rg7+ Kh8 32. Qxe5! Qb8 33. Rxh7+ Kxh7 34. Qg7 mate] 31. Rg7+ [Bh3 is answered by 31[…]Qf8] Kh8 32. Rxh7+ Kg8 33. Rg7+ draws for White; Tiger wants more) 28. Kh2 g5 29. Nd2.

Black finally accepts the poisoned chalice, with fatal results: 29…Bxa1 30. Rc7 Qd8 (see diagram) 31. Bh3!! g4 (Bxh3 32. Qc5+ follows the game line) 32. Bxg4! Bxg4 33. Qc5+ Ke8 (Kg8 32. Qc4+ Kh8 35. Qf7 Qxc7 36. Qxc7 Rg8 37. Qf7 Rg6 38. Qf8+ Rg8 39. Qxf6+ and mate next) 34. Rxh7.

Despite his heavy material advantage, Black cannot stop the White h-pawn. His rooks and the bishop on a1 are terribly misplaced. In the final position, White wins easily on 39…Be6 40. Qh7+ Ke8 41. Qxb7 Rxa4 42. Qc6+ Kf7 43. b6 Rb4 44. b7 Bc3 45. b8=Q Rxb8 46. Qc7+ Kg6 47. Qxb8. Gambit Tiger logged off.

Gulko vs. Computers Match, March 2002

Deep ShredderGulko

1. d4e614. Nxf7Kxf7

2. e4d515. Bf4Bb7

3. Nd2b616. c3Nc6

4. Ngf3Nf617. f3exf3

5. Bd3c518. Bxf3Qd7

6. 0-0c419. Qe2Re8

7. Be2dxe420. Bh5+g6

8. Ng5Bb721. Bg4Nxg4

9. Nxc4Be722. Qxg4Kg7

10. Ne50-023. Bxh6+Kxh6

11. Bc4Bd524. Qh3+Kg7

12. Be2h625. Rf7+Black

13. Ngxf7Rxf7resigns

Internet Challenge II, April 2002

SmirinGambit Tiger 2

1. Nf3d521. Qd2Rxa5

2. g3Nf622. h4Kf8

3. Bg2c523. h5Rca8

4. 0-0Nc624. h6g6

5. d3 e525. Qc2Bb4

6. Nbd2Be726. Qc7Bc3

7. e40-027. Qxb6Qe7

8. c3Bg428. Kh2g5

9. h3Bc829. Nd2Bxa1

10. Re1d430. Rc7Qd8

11. Nc4Qc731. Bh3g4

12. a4Be632. Bxg4Bxg4

13. Bd2Nd733. Qc5+Ke8

14. Qc2f634. Rxh7Rc8

15. Rec1Nb635. Rh8+Kf7

16. Nxb6axb636. Rxd8Rxc5

17. cxd4cxd437. h7Rc8

18. b4Qd738. Rxc8Bxc8

19. b5Na539. h8=QBlack

20. Bxa5Rfc8resigns

David R. Sands can be reached at 202/636-3178 or by e-mail at dsands@washingtontimes.com.

Copyright © 2018 The Washington Times, LLC. Click here for reprint permission.

The Washington Times Comment Policy

The Washington Times welcomes your comments on Spot.im, our third-party provider. Please read our Comment Policy before commenting.


Click to Read More and View Comments

Click to Hide