White: Garry Kasparov Black: The World Team
Opening:
Sicilian Defence,
Canal–Sokolsky Attack (
ECO B52)
1. e4 c5 2. Nf3 d6 3. Bb5+ Kasparov played his first move
1.e4 on June 21, and the World Team voted by a 41% plurality to meet him on his home turf with the
Sicilian Defence. Kasparov's third move was unusual for him. He typically plays the more ambitious 3.d4 in this position, which promptly leads to an . Kasparov commented that Bb5+ had something of a weak reputation until the 1990s, but he had been impressed with both
Vasyl Ivanchuk and
Alexei Shirov playing it against him for a win. Additionally,
Vladimir Kramnik had played it twice against
Boris Gelfand at the
1994 Candidates, winning once and drawing the other game. Kasparov was apologetic about the move, but excused himself in light of his expected upcoming championship match against
Vishy Anand, where he expected the
Najdorf Sicilian—a natural extension of 3.d4—to be a potential battlefield.
3... Bd7 4. Bxd7+ Qxd7 5. c4 Some 30% of viewers wanted Black to play 4...Nxd7, but were outvoted by those who thought Black's
queen was safe on the light squares with White's light-squared
bishop off and preferred to the
knight to c6. With c4, Kasparov solidified his hold on the d5-square by advancing the c-
pawn before developing his , which joined the attack on d5 on the next move. The resulting pawn formation for White is sometimes called the
Maróczy Bind, a way of Black's position. The World Team responded by contesting control of the d4-square.
5... Nc6 6. Nc3 Nf6 7. 0-0 g6 Now Black must try to
castle on the and can set it up by developing the to either the e- or g-. Rather than move the e-pawn to e6 or e5 and have the dark-squared bishop blocked by the d-pawn, the World Team opted to
fianchetto the remaining bishop, further contesting the dark central squares. Kasparov immediately broke up the with his , before the black bishop could come to bear.
8. d4 cxd4 9. Nxd4 Bg7 10. Nde2 The center was too hot for the white knight on d4, because the World Team was threatening a
discovered attack by moving the black knight away from f6, unmasking the g7-bishop. Exchanging knights on c6 would have been silly for Kasparov, as it would have brought a black pawn to c6, giving the World Team greater control of d5; instead, a retreat was in order. All of Kasparov's moves up to this point were considered good according to opening theory of the time. If White had wanted to achieve the strategic aims of this opening line, i.e. cramping Black's position without allowing , then either a different move (10.Nc2) or a different move order was necessary. Slightly more popular today is 7.d4 and after 7...cxd4 8.Nxd4 Qe6 would now lose the queen. 7.d4 does allow two other queen sorties with 7... Qg4 or after 7...cxd4 8.Nxd4 Qg4.
10... Qe6 Black finally secured the opportunity to castle but instead opted to complicate the position. This move was a by the World Team, i.e. a move which had never before been played in a recorded game. Krush discovered and analyzed the move, and enlisted Paehtz to recommend it as well, to give it a better chance of winning the vote. Their combined advocacy, plus much discussion on the bulletin board, was enough to gain it 53% of the vote. After this move, MSN requested that the four official analysts not coordinate with each other, perhaps to ensure a greater variety of recommendations. The analysts worked in isolation from each other thereafter. 10...Qe6 has since been played frequently, though 10...0-0 also remains popular, according to
ChessBase. The move was played against Kasparov again in a 2009 match against Gabriel Gaehwiler. The tenth move was a turning point for the World Team, not only because it increased Krush's stature and energized the World Team, but because it blew the position on the board wide open. The black queen
forked Kasparov's central pawns; he could not save them both. Counterattacking with 11.Qb3 would have been met by 11...0-0 12.Qxb7 Rfc8, and the World Team would have won back a with a favorable game. Kasparov was forced to enter the maelstrom with the next several moves.
11. Nd5 Qxe4 12. Nc7+ Kd7 13. Nxa8 Qxc4 14. Nb6+ axb6 {{Chess diagram After on both sides, Kasparov made a
desperado move with his knight to cause the World Team to have
doubled pawns. the game was still even, with a knight and two pawns balancing a
rook. , the World Team had the disadvantages of doubled pawns and a centralized king, but the advantage of a lead in development and a central pawn mass. With no central pawns, Kasparov had no clear-cut way to expose the black king. In the judgment of many commentators, the World Team was at least equal, and it was perhaps Kasparov who was fighting uphill. Kasparov rose to the challenge with an excellent move picked from alternatives which would have let the World Team take a strong initiative. For instance, it was tempting to harass the black queen and possibly
fianchetto the with 15.b3, but this would have invited the World Team to switch wings and initiate a attack with 15...Qh4. Or, to blindly follow the rule "never move a piece twice in the opening when you can develop another piece" with 15.Be3 would have allowed the World Team to play 15...Nd5, bringing the black knight to the square it most fervently wished to occupy. Kasparov's actual move contested d5, somewhat blunted the effect of the black bishop on g7, and retained a compact, flexible position.
15. Nc3 On the fifteenth move the World Team hotly debated a number of promising alternatives, including 15...e6 (still contesting d5), 15...d5 (occupying d5 outright!), 15...Ne4 (trading off Kasparov's best-placed piece), 15...Rd8 (intending to
artificially castle and mobilize the central pawns), 15...Ra8 (pressuring the and threatening a via a5), and 15...b5 (threatening to dislodge the white knight and pressure the queenside). The plethora of strong options available to the World Team was reflected in the analysts' recommending four different moves. By this point in the game, several chess clubs had begun posting daily analysis to complement what was available on the official bulletin board and Web site of the game. The weightiest of these was the "Grandmaster Chess School" or GM School, a consortium of Russian grandmasters in
St. Petersburg. For the World Team's 15th move, they recommended 15...b5, along with Paehtz. Some people expected the unofficial recommendation of the GM School to be influential, particularly when the official analysts could not agree, but 15...b5 came in a distant second with 15% of the vote. In first place was
Jon Speelman's idea of 15...Ra8 with 48% of the vote. if the world team responded with 38...Ke4. The power of the centralized black king vis-a-vis the white king off in the corner would come into play, showing that even brilliant chess moves have minor disadvantages. Instead of trying the complexities of the 38.Rd1 line, Kasparov said at his press conference that he had no idea how the game would turn out, and began to force the World Team into an ending in which each side got a new queen, and the outcome was still unclear.
38. h6 Ne7 39. Rd1 e5 40. Be3 Kc4 41. Bxd4 exd4 The World Team's moves were all essentially forced. Kasparov's main chance to deviate would have been with 40.Bc1, but then 40...Ke6 appeared to hold. Although the black bishop was at that time temporarily cut off from holding back the pawns, the black king was temporarily not cut off, and an extra tempo in such positions makes all the difference. Kasparov, when making his 38th move, had in all probability already elected to force the World Team's responses all the way to move 50. The World Team's pawns, albeit ungainly, had become a sufficiently potent threat that Kasparov's rook could not serve the double purpose of guarding the home rank and forcing through the passed pawns against Black's knight. Therefore, the white king had to come out of its corner at last.
42. Kg2 b2 43. Kf3 Kc3 44. h7 Kasparov could have made the pawn advance on either of the two previous moves as well, but it would have merely transposed, with the World Team responding as it did in the actual game:
44... Ng6 45. Ke4 Kc2 46. Rh1 The World Team needed to advance the d-pawn as well as the b-pawn, in order to get a second queen after Kasparov sacrificed his rook for the first one. Against the recommendations of the analysts, nine percent of the World Team voted for the tempting prospect of immediately queening the b-pawn, which is a blunder that loses after 46...b1=Q 47.Rxb1 Kxb1 48.Kxd4. The white king arrived to assist the white pawns just as quickly with diagonal moves as with straight ones, and the diagonal moves gave it incidental threats in the center.
46... d3 47. Kf5 For the World Team's 47th move, the analysts were again unanimous, this time recommending immediate queening. 15% of the voters were tempted to try to hang onto the knight a few moves longer with 47...Nh8. This would have led to a lost endgame after 48.g6 d2 49.g7 d1=Q 50.Rxd1 Kxd1 51.gxh8=Q b1=Q+, when Black cannot engineer a
perpetual check.
47... b1=Q 48. Rxb1 Kxb1 49. Kxg6 d2 50. h8=Q d1=Q The race to queen a pawn ended in a tie. The position was not equal, though. Kasparov's far-advanced g-pawn was an imminent threat to queen, while the World Team's pawns were potentially liabilities as much as assets, since they could give the white king a modicum of shelter from threatened checks. The general plan of defense for the World Team was to place the king favorably so that Kasparov could not threaten a deadly queen trade, and then determinedly check Kasparov's king to prevent him from queening the g-pawn. In some lines the World Team could gain counterplay by advancing its own pawns, but this was a secondary strategy. The World Team had relied heavily on computer analysis for much of the game, but at this point the forward-searching chess engines began to produce worthless suggestions. This type of position was not well-handled by 1999 chess programs, until the position became simple enough to use an
endgame tablebase. As of October 1999, however, there were no seven-piece endgame tablebases, and seven pieces remained in the actual position. After the game was over, Peter Karrer of Switzerland constructed a specialized tablebase for the purpose of fully understanding this endgame. With the aid of the tablebase, Krush and IM
Ken Regan were able to prove that the position after the World Team's 50th move was drawn with best play on both sides. Both Kasparov and the bulletin board suspected that the position was drawn, but as the further course of the game proved, no one fully understood the position at the time. Some World Team members tried to gain insight from the position by consulting state-of-the-art five-piece tablebases, with the black pawns missing, and were encouraged to find the position dead drawn. Unfortunately for the World Team, the extra black pawns complicated the position enormously, and not necessarily to the World Team's advantage. For example, analysis positions arose which were theoretically drawn with both black pawns in place, but which would be won for White if exactly one black pawn were not there, and then again drawn with both black pawns gone. Since 2012, seven-piece Lomonosov tablebases have become available which prove that the position is drawn. Moreover, using a special EGTB generator "Hoffman" developed by Brent Baccala, it can be shown that Black can draw without trying to promote its pawns and without moving its king out of the a–d × 1–4 square.
51. Qh7 This move threatened to gain Kasparov a
tempo by advancing the king with
discovered check. The bulletin board and Krush came up with the response 51...Ka1, which Kasparov later said he had considered as a dead draw, and the subsequent
endgame tablebases confirmed to be so. But Felecan recommended 51...d5 and Paehtz favored 51...b5, while Bacrot sat out the turn. For the first time in 40 moves, Krush's recommendation was not selected, receiving only 34% for 51...Ka1. 39% of votes were for Elisabeth Paehtz's 51...b5. Someone using the alias "Jose Unodos" joked that he rigged the vote for the Paehtz suggestion 51...b7–b5 instead of Kb1–a1 that Krush and most others thought the best move, although MSN said afterward vote-stuffing only affected move 59.
51... b5 52. Kf6+ The bulletin board and Krush now concluded that 52...Kc1 offered the best chances of holding a draw. Kasparov versus the World 1999 analysis via Peter Karrer's special computer program for the position, which was fashioned two weeks after the game was over: If 52...Kc1! (but 52...Kb2 was played in the game), 53.Qe4! Qf1+ 54.Ke7 b4. Here if 55.Qb4, Black must find 55...Qf5 (the only move) 56.Qc3+ Kb1 57.Qf6 Qe4+ (only move) 58.Kf7 Qc4!+ 59.Kg7 d5 (only move) so Black can draw. Upon 55.g6! Qg1! 56.Qf4+ Kd1! 57.Kf7 b3=. Also on 55.g6 Qg1 56.Qc4+ Kd1 57.Qd3+ Kc1—here White cannot theoretically progress. However, with Bacrot recommending 52...Ka1 and both Felecan and Paehtz favoring 52...Kb2, the latter move eked out a victory with 42%. The idea was to use the king to support the b-pawn toward promotion, and tablebases show that either move will hold a draw.
52... Kb2 Kasparov made a reasonable-looking queen maneuver to shelter his king from checks on the f-file:
53. Qh2+ It was postulated at the time that 52...Kb2 was a game-losing mistake, and that Kasparov could have won the game by force after 53. Qe4, but modern tablebases don't bear that out.
53...Ka1! 54. Qf4 Krush and others on the MSN forum recommended sacrificing the b-pawn with 54...b4 in order to allow the black queen to give check on the f-file. This was partly based on the modified tablebases, which needed to assume exchanges to reach an error-free conclusion, by reducing the number of pieces and possibilities. Bacrot advocated centralizing the black queen with 54...Qd5!, while Felecan and Paehtz suggested 54...Qd3. Later analysis showed that Bacrot's recommendation could hold the draw in a relatively comprehensible fashion, and Felecan and Paehtz's move could hold after some desperate ingenious squirming, but Krush's move would concede Kasparov a forced win, if he could find it.
54... b4? This move was the first mistake made in the seven-piece endgame. The tablebase shows that either 54...Qd3 or 54...Qd5 gives a theoretical draw. The idea was to get more activity with their queen and be able to force White's king into several checks; however, it was shown that with good piece coordination the white king would be able to escape the checks and White would promote first with a winning position, demonstrated by Kasparov in the game.
55. Qxb4 Qf3+ 56. Kg7 d5 57. Qd4+ Kb1 58. g6 Kasparov played forcefully on moves 55–57, and the World Team responded each time with a large majority for the best plausible move. On move 58, both 58...Qe4 and 58...Qf5 looked reasonable, but the bulletin board had analyzed the former to a forced loss. Krush duly recommended the latter. Due to an e-mail mix-up, her recommendation and analysis were not received on time by the MSN site, and voting proceeded for some time with Bacrot and Paehtz both recommending 58...Qe4 against Felecan favoring 58...Qf5. When the weaker move won the vote 49% to 44%, there were dismayed cries from the bulletin board that the game was lost due to the delay of Krush's recommendation, as Krush's recommendations had been closely followed by voters in earlier moves. Kasparov agreed that 58...Qf5 would have put up stiffer resistance, but claimed it was also losing, and published a forced win sequence. Subsequently, tablebases showed an error in Kasparov's analysis, but confirmed that 58...Qf5 could not have saved the draw with
best play on both sides. The main line starts with 58...Qf5 59.Kh6 Qe6 60.Qg1+ Ka2 61.Qf2+ Kb1 62.Qd4 Ka2 63.Kg5 Qe7+ 64.Qf6 Qe3+ 65.Qf4 Qg1+ 66.Kf6 Qb6+ 67.Kf7 Qb7+ 68.Ke6 Qc8+ 69.Kf6 Qd8+ 70.Kf5 Qc8+ 71.Kg5 Qc3 72.Qh2+ Ka1 73.Qe2 Kb1 74.Qf2 Qc1+ 75.Kg4 Qc3 76.Qf1+ Kc2 (or 76...Kb2 77.Kf5 Qc7 78.Qe2+ Kb1) 77.Kf5 Qc7 78.Qe2+ Kb1 79.Qd3+ Ka2 80.Qa6+ Kb3 81.Qe6 Ka2 82.Qf7 Qc2+ 83.Ke6 Qe2+ 84.Kxd5.
58... Qe4 59. Qg1+ At this point several disgruntled members of the bulletin board, knowing the game to be lost, suggested the
blunder 59...Qe1. The move 59...Qe1?? got the majority of the votes the next round of voting, but Microsoft invalidated all votes due to
ballot stuffing. MSN announced after the game that this was the only move on which ballot stuffing had taken place at a level where it was significant. Microsoft added a
resign option to the next vote, which got 28% of the vote after it was first added – a high number, but not the plurality.
59... Kb2 60. Qf2+ Kc1 61. Kf6 d4 62. g7 With his 62nd move, Kasparov announced a in 28 moves found by the computer program
Deep Junior. In light of this, 51% of the World Team voters opted to resign on October 22, four months after the game commenced. ==Aftermath==