Dortmund R5: Leko, Gelfand win, Leko leads

The Dortmunder Sparkassen Chess Meeting is taking place in the State Theatre (Schauspielhaus) in Dortmund, Germany, from July 29th to August 6, 2006. It is a single round robin tournament with eight players, averaging 2720 Elo and making for a category 19 event.

Round five report

Vladimir Kramnik vs Michael Adams was a Petroff which at some stage looked like it would be very long and very tense, with Kramnik grinding out a win. However the game ended after 41 moves with a draw.

Arkadij Naiditsch vs Peter Svidler, a closed Ruy Lopez, was hard fought, lasting almost six hours and ending in a queen and three pawns vs queen and two which White could not win. Naiditsch is playing what most spectators consider the most enterprising chess in Dortmund.

Peter Leko vs Levon Aronian was an Exchange Ruy with an eighth move novelty by Leko. The result was a blocked position in which both sides shuffled their pieces during a protracted middlegame phase. Aronian was intent on holding on to a draw, Leko was casting around for more. On move 55 he exchanged knights and simplified to a rook and pawns ending which Aronian could not hold. The game ended on move 63 with the second victory in this tournament for Hungary's top GM, giving him the sole leadership in the cross tables.

Boris Gelfand vs Baadur Jobava was a marathon encounter lasting six hours and forty-five minutes, coming immediately after his 7½ hour loss (to Michael Adams) which Gelfand had to endure yesterday. But he proved that a 38-year-old world class grandmaster can have world class stamina.

Gelfand,B (2729) - Jobava,Ba (2651) [B39]
Sparkassen Dortmund GER (5), 04.08.2006
1.Nf3 c5 2.c4 Nc6 3.d4 cxd4 4.Nxd4 g6 5.e4 Bg7 6.Be3 Nf6 7.Nc3 Ng4 8.Qxg4 Nxd4 9.Qd1 Ne6 10.Rc1 Qa5 11.Be2 b6 12.Qd5 Rb8 13.Qxa5 bxa5 14.b3 Bd4 15.Bd2 d6 16.0-0 Bc5 17.Na4 Ba3 18.Rb1 Bb7 19.f3 Bc6 20.Bxa5 Bxa4 21.bxa4 Kd7 22.Rfd1 Nc5 23.Bc3 Rhc8 24.a5 Na4 25.Be1 Nb2 26.Rdc1 Bb4 27.Bf2 Bxa5 28.c5 dxc5 29.Rxc5 Rxc5 30.Bxc5 Bc3 31.a4 Ke6 32.Bb5 Kf6 33.Rc1 a6 34.Rxc3 axb5 35.a5 Ke6 36.Rb3 Na4 37.Be3 Kd7 38.a6 Ra8 39.Rxb5 Rxa6 40.Bd4 Rd6 41.Rb7+ Kd8 42.Rb4 Ra6 43.Kf2 Kc7 44.Ke3 Nb6 45.Bxb6+ Rxb6 46.Rxb6 Kxb6

Gelfand has liquidated to a pure pawn ending he believes he can win. 47.Kd4 Kc6 48.Ke5 Kd7 49.f4 Ke8 50.h4 h5 51.f5 f6+ 52.Ke6 gxf5.

This is where computers, equipped with the ruthless perfection of their endgame databases, jump into action. After a second or two of thought and five hundred consultations of the tablebases Fritz announces that 53.e5! is mate in 69 moves. Terrifying. 53.e5! Boris found it in a couple of minutes and without access to the preprocessed analysis of hundreds of millions of positions. 53...fxe5 54.Kxe5 Kd7 55.Kxf5 Kd6 56.Kg5 Ke5 57.Kxh5 Kf4 58.Kg6 e5 59.h5 e4 60.h6 e3 61.h7 e2 62.h8Q e1Q.

Perfectly played, says Fritz, wondering how on earth human beings can do it. The above position is mate in 59, and belongs to the most difficult five-piece endings. Against the perfect defence of a computer probably no human being (except of course John Nunn) would be able to win the full point, but of course Gelfand is facing another human being, whose defence is as weak or strong as Gelfand's attack. We put in some notes by the silicon oracle not to disparage the performance of the players, but to show how complex and incomprehensible these endgames are. Exclamation points indicate that the move played is the only one that wins in the given position.

63.Qb8+! Kg4 64.Qc8+! Kg3 65.Qh3+! Kf4 66.Qf5+ Kg3 67.Qh3+ Kf4 68.Qf3+! Ke5 69.g4! Kd4 70.g5! Qe8+ 71.Kg7 Qe7+ 72.Kh6 Kc4 73.Qf4+ Kb5? Bad defence, says Fritz. After 73...Kb3 White needs 47 moves to mate, now it is just 32 moves. 74.g6 Qe6 75.Kg5 Qe7+ 76.Kg4 Qg7 77.Qd6? More criticism from the database: 77.Qf5+ mates in 30, Gelfand's move increases it to 47.

77...Ka4 78.Kf5 Qc3 79.Qe5. The last moves have brought no improvement, and 79.Qe5 moves the win to 58 moves. 79...Qh3+ 80.Kg5 Qg2+ 81.Kf5 Qh3+ 82.Kf6 Qf3+ 83.Qf5 Qc3+.

This is quite typical of the endgame. At move 62 it was 59 moves to win, now it is 58. And White must find 84.Qe5 to achieve even this. The only other move that wins is 84.Kf7, which would require 68 more moves. In the game White plays neither: 84.Kg5? Now the position is a theoretical draw. 84...Ka3 85.Qf8+ Ka4? Now it is mate in 59 again. Black should have played 85...Kb3 or 85...Ka2 to keep the draw. 86.Qa8+? The position is drawn again, White needed to find 86.Kg4! (only move to win). 86...Kb4 87.Qb7+ Ka5? Mate in 36. If Black had played 87...Ka3! the position would still be a draw. 88.g7 Qe5+ 89.Kg6 Qe6+ 90.Kh7 Qf5+ 91.Kg8 Ka4.

What would you play in this position? Hint: there are only two moves that win for White. We would like to once again stress that we are not criticizing the play of the two GMs, but merely demonstrating the complexity of this endgame. Some day, maybe, computers will annotate games as follows: 1.e4 e6?? Allows mate in 23 million moves. 1...e5, 1...d5 or 1...a6 was required to hold the draw.

92.Qh1? Draw again. 92.Qa7+ wins in 35, 92.Qe7 wins in 39 moves. 92...Qc8+? Black needed to find 92...Kb3 or 92...Ka3 to hold on. Now it is mate in 32 moves. 93.Kh7 Qf5+ 94.Kh8 Qe5 95.Qh3 Qd4 96.Qe6 Qh4+ 97.Kg8 Qf4 98.Qd5 Ka3 99.Kh7 Qh4+ 100.Kg6 Qg3+ 101.Kf7 Qf4+ 102.Ke8. White has not allowed his opponent to escape into a theoretical draw, but it is still 32 moves to mate. 102...Qb8+ 103.Qd8. Now it is 45 moves. 103...Qb5+ 104.Qd7 Qh5+ (31 moves left) 105.Kf8 Qf3+ 106.Ke7 Qe4+ 107.Qe6 Qb7+ (22 moves) 108.Kf6 Qf3+ 109.Kg5 Qg3+ 110.Qg4 Qe5+ 111.Kh4 Qf6+ 112.Qg5 Qd4+ 113.Kh3 and Black resigned, 14 moves before mate. For those of you who are still awake, here is an "optimum" continuation: 113...114.Kh2 Qe6 115.g8Q Qe2+ 116.Qg2 Qh5+ 117.Qh3+ Qxh3+ 118.Kxh3 Kb4 119.Qd5 Kc3 120.Kg3 Kb2 121.Qc4 Ka1 122.Qb5 Ka2 123.Kf2 Ka1 124.Ke3 Ka2 125.Kd2 Ka1 126.Kc1 Ka2 127.Qa4 mate. [Click to replay]

The above endgame is particularly interesting because there has been a fair amount of human analysis done on the endgame queen and g-pawn vs queen. This resulted from an endgame that appeared in practical play and was deeply analysed by the great endgame expert André Chéron.

Botvinnik,Mikhail - Minev,Nikolay N [D47]
Amsterdam ol (Men) fin-A Amsterdam (5), 17.09.1954

Here the tablebases tell us that White must play 77.Kg4 for a mate in 60 moves (or 77.Kg5/Kh4/Kh6 or 77.Qf3/Qf2/Qd3 for slightly longer mates). But Botvinnik played 77.Qf4+? which allows Black to escape with a draw. Chéron showed that the best defensive strategy for Black is, somewhat incredibly, to move his king to the square a1. So Minev's reply in the game, 77...Ka5?, is a deadly mistake and allows White to win again, now in 35 moves. Botvinnik did it with 78.Qd2+ Ka4 79.Qd4+ Ka5 80.Kg5 Qe7+ 81.Kf5 Qf8+ 82.Ke4 Qh6 83.Qe5+ Ka4 84.g7 Qh1+ 85.Kd4 Qd1+ 86.Kc5 Qc1+ 87.Kd6 Qd2+ 88.Ke6 Qa2+ 89.Qd5 Qe2+ 90.Kd6 Qh2+ 91.Kc5 1-0. [Click to replay]

If you replay the Gelfand-Jobava game you will see that the errors seem to arise when Black fails to move his king in the right direction, towards the square a1 – and when White almost forces him to do so. Moving the black king to a1, or preventing it from getting there, is one of the few strategies that are comprehensible to human beings. Unfortunately the endgame database just provide us with the best moves in any given position, but are unable to give reasons for the choice. Fritz may play the endgame with absolute perfection, but if you ask it why a certain move must be played its only answer is "because that's what it says in the list" (of hundreds of millions of positions).

Additional information

• Workshop: Fritz Endgame Turbo 3 - Part 1
• Workshop: Fritz Endgame Turbo 3 - Part 2
• Buy Fritz Endgame Turbo 3
• Shredder 10 and the Shredderbases

Standings after five rounds

• Download all games so far in PGN
• Games from the Pähtz-Krush match

Full Schedule and Results

Links

• Official web site
• All ChessBase reports on Dortmund 2006
• ChessBase reports on Dortmund 2005
• ChessBase reports on Dortmund 2004
• ChessBase reports on Dortmund 2003