- The Washington Times - Friday, November 21, 2003

The recent four-game match between former world champion Garry Kasparov and a 3-D version of the German software program Fritz provided an illuminating, if unintended, glimpse into the strengths and weaknesses of computer play.

The match, which concluded Tuesday, produced yet another split decision, with each side winning one game and the first and fourth games ending in draws.

Even as computers approach the playing strength of the very top human players, however, it’s still an open question whether they play “better” chess. Computers clearly have their strengths: the amazing ability to calculate; emotional unflappability; hyperaccurate defense; and an invulnerability to fatigue, distraction and simple tactical oversight — even in the sixth hour of a grueling game.

In other ways, though, particularly in trying to make long-range “soft” judgments about a given position, even the best programs can be outclassed by a moderately accomplished Class A player.

Game 3 from the match, played at the New York Athletic Club and televised on ESPN2, provided an ideal display of the computer’s shortcomings. The QGD Slav variation played in the game was first employed in a game from the 1948 world championship tournament in The Hague between American star Sammy Reshevsky and Estonian great Paul Keres.

Keres, perhaps the finest player of the past century never to win the world title, immediately sized up the demand center to keep from getting suffocated. Faced with virtually the same strategic dilemma, Fritz dawdles endlessly and goes down virtually without a fight.

In Reshevsky-Keres, Black sees already on 11. Nd2 Be7 12. Be2 h5! that he must engage in countermeasures fast. His a-pawn in the long run can’t be saved, and his only hope is to drum up pressure on the opposite wing as White goes after it.

With Black building up a head of steam on the king-side, Reshevsky decides to sharpen things by stealing a different pawn: 15. 0-0-0 Ne6 16. Ndxe4!? Nxe4 17. Nxe4, when 17…dxe4?? loses to 18. d5! Bxc5 (Ng5 19. dxc6 is crushing) 19. dxc6 Bxa3+ 20. Kc2! Qf6 21. cxb7+ Kf8 22. Qxa3+ Kg8 23. bxa8=Q. But Reshevsky’s maneuver opens the king file for Black forces and puts the White forces permanently on the defensive with the weak e-pawn.

By 35. Rxg3 Bxd3 36. Qxd3 Re4, a computer would love White’s extra pawn, but it’s clear that Black is better. White faces a long defensive slog, and Black can invade on the open king-side.

Still, Reshevsky was one of the game’s greatest defenders, and the struggle only sharpens on 41. Qd3 Qh5 44. Bd2 g6 43. Rg5 Qxh3 44. R1g3 Qh2 45. Rxg6+ Rxg6 46. Rxg6+ Kf7. Now White could have had the draw with 47. Rxc6! bxc6 48. b7 Re8 49. Qxf5+, with a perpetual check.

But White instead falls for a sly trap on 47. Rg5? Be7! 48. Rxf5+ Bf6 49. Kc3 Qh3, and the rook must give itself up for the bishop. Material equality is restored, but now the Black rook and queen invade through the lines opened up by Black’s aggressive plan dating back 30 moves.

The noose tightens on 54. Qb3 Ra8 55. Bc1 (the tricky 55. Qb2 Qxa3 56. Qxa3 Rxa3 57. e4!? dxe4 58. d5 cxd5 60. c6 just fails to 60…d4+! 61. Kb3 Ra6! 62. cxb7 Rxb6+, collecting the passed pawn) Rh8 56. e4?! (holding tight with 56. Bd2 was tougher) Rh1 57. e5+ Ke7 58. Qe3 Qa2+ 59. Kc3 Rh2 60. Qd3 Qa1+ 61. Kb3 Qxc1 62. f5 Qb2+ 63. Ka4 Rh8!.

With mate on the horizon — 64. f6+ Ke6 65. Qb3 Ra8+ 66. Kb4 Qxd4+ 67. Qc4 Qxc4 mate — Reshevsky resigned.

Presented with virtually the same positional assignment, Fritz failed miserably — and in ways that few human players of any strength would have done.

There actually are very few variations to work through in this game after 11. Nd2 Be7 12. b6 (see diagram; the position closely tracks the previous game) Qd8 13. h3 0-0 14. Nb3 Bd6!? (clever but ineffectual; if 15. cxd6?? Nxb6, the White queen is trapped, but Kasparov is under no obligation to capture) 15. Rb1 Be7??.

Black’s last move by itself wasn’t bad, but it demonstrates that Fritz is clueless in the closed position here. A glance at the White array — a pawn chain running like a dagger deep in the Black queen-side, a heavy buildup on the isolated a-pawn — leaves no doubt that Black’s only hope is to push his king-side pawns and counterattack. But the idea proves too nebulous and beyond Fritz’s calculating horizon.

White temporarily gives back his extra pawn to prepare an invasion on the a-file, but Fritz cannot hold it, and the passed White b-pawn must win in the end.

After 42. Nxb7 Rxb7 43. Nxa6 Qd7 44. Qc2 Kh8 45. Rb3, Fritz’s handlers threw in the towel. The computer may not yet appreciate it, but it’s over on lines such as 45….Ne8 46. Rba3 Nc8 47. Nb4 Rab8 48. Ra8 Bg5 49. Rxb8 Rxb8 50. Ra6 Bd8 51 Qa4 Ne7 52. Ra8 Rxa8 53. Qxa8, and the b-pawn runs to glory.

World Championship Tournament, The Hague, April 1948

ReshevskyKeres

1. d4d533. g4hxg3

2. c4c634. Nxg3Nxg3

3. Nf3Nf635. Rxg3Bxd3

4. Nc3e636. Qxd3Re4

5. e3a637. Reg1Rae8

6. c5Nbd738. Rf1Qh4

7. b4a539. Rfg1R8e6

8. b5e540. Qd2f5

9. Qa4Qc741. Qd3Qh5

10. Ba3e442. Bd2g6

11. Nd2Be743. Rg5Qxh3

12. Be2h544. R1g3Qh2

13. b6Qd845. Rxg6+Rxg6

14. h3Nf846. Rxg6+Kf7

15. 0-0-0 Ne647. Rg5Be7

16. Ndxe4 Nxe448. Rxf5+Bf6

17. Nxe4h449. Kc3Qh3

18. Nd20-050. Rxf6+Kxf6

19. Rhg1Re851. Qc2Qf1

20. Bd3Bf852. Qxa4Qa1+

21. Bb2Ng553. Kc2Re8

22. Qc2a454. Qb3Ra8

23. a3Qe755. Bc1Rh8

24. Rde1Ne456. e4Rh1

25. Nf1Qg557. e5+Ke7

26. f3Nf658. Qe3Qa2+

27. Kb1Nh559. Kc3Rh2

28. Bc3Bd760. Qd3Qa1+

29. f4Qh661. Kb3Qxc1

30. Qf2Qf662. f5Qb2+

31. Kb2Bf563. Ka4Rh8

32. Qc2Be4White resigns

X3D Man vs. Machine Match, Game 3, New York, November 2003

KasparovX3D Fritz

1. Nf3Nf624. Kc1Rd8

2. c4e625. Rc2Nbd7

3. Nc3d526. Kb2Nf8

4. d4c627. a4Ng6

5. e3a628. a5Ne7

6. c5Nbd729. a6bxa6

7. b4a530. Na5Rdb8

8. b5e531. g3Bg5

9. Qa4Qc732. Bg2Qg6

10. Ba3e433. Ka1Kh8

11. Nd2Be734. Na2Bd7

12. b6Qd835. Bc3Ne8

13. h30-036. Nb4Kg8

14. Nb3Bd637. Rb1Bc8

15. Rb1Be738. Ra2Bh6

16. Nxa5Nb839. Bf1Qe6

17. Bb4Qd740. Qd1Nf6

18. Rb2Qe641. Qa4Bb7

19. Qd1Nfd742. Nxb7Rxb7

20. a3Qh643. Nxa6Qd7

21. Nb3Bh444. Qc2Kh8

22. Qd2Nf645. Rb3Black

23. Kd1Be6resigns

David R. Sands can be reached at 202/636-3178 or by e-mail at [email protected]


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