Garry Kasparov's Forgotten Weapon Against The Queen's Gambit (part six)

Predrag Nikolić (2490) - Kasparov (2690)

Nikšić (Yugoslavia) Round 9 1983

1.d4 d5 2.c4 e6 3.Nc3 c5 4.cxd5 exd5 5.Nf3 Nc6 6.g3 Nf6 7.Bg2 Be7 8.0-0 0-0 9.dxc5!?

This is not as popular as 9.Bg5 but is preferred by the analysis engines Stockfish12 and Komodo11.01.

9...Bxc5

An active alternative is 9...d4!? 10.Na4 Bf5 [10...Bg4?! 11.a3 a5 12.Ne1!? Re8 13.Nd3 was much better for White in David Clear (176 ECF) - Spanton (163 ECF), London League 1995 (1-0, 50 moves)] 11.Bf4, when Black does not have enough compensation for a pawn, according to the engines, although there is a lot of play in the position.

10.Bg5 d4

Pushing the d pawn is the main move.

11.Ne4!?

11.Bxf6, described by Shamkovich & Schiller in Play The Tarrasch as "the only serious attempt to preserve the advantage," will be covered in the next part of this series. The text is slightly preferred by Stockfish12, but not by Komodo11.01.

11...Be7 12.Bxf6 Bxf6

White has given up the bishop-pair but has a well-placed knight on e4 - a theme often seen in ...dxe4 lines of the French.

13.Rc1 Re8 14.Ne1!?

The knight heads for the blockading square d3.

14...Be7!?

Kasparov decides it is worthwhile spending a tempo to preserve the bishop-pair. Emanuel Lasker - Siegbert Tarrasch, Berlin 1918, saw 14...Bf5 15.Nc5 Bc8? (15...Qb6 is a later improvement) 16.Nxb7 Bxb7 17.Rxc6 (1-0, 56 moves).

15.Nd3 Bf8!? 16.Qd2 a5 17.Rfd1

Later the same month, the game Nikolić (2490) - Svetozar Gligorić (2510), Boris Kostić Memorial (Vršac, Yugoslavia) 1983, went 17.a3 a4!? 18.Nec5 with advantage to White, according to the engines (1-0, 58 moves). The engines reckon 17...Bg4!? 18.Rfe1 Rc8 keeps White's edge to a minimum.

17...Bg4 18.Ndc5 Bxc5 19.Rxc5 Qe7

Which reasonable (looking) move should White avoid?

*****

*****

*****

*****

*****

20.h3?

Nikolić must have missed Black's reply. The engines reckon 20.Rcc1 is equal.

20...Bxe2! 21.Re1

21.Qxe2? Qxc5.

21...d3 22.Qc3 Rad8 23.Nd2 Nd4!

The engines agree this is better than 23...Rc8 24.Bxc6 Rxc6 25.Rxc6 bxc6 26.Qxa5, although that too seems good for Black.

24.Qxa5 h6!?

This is a little slow but ensures no nasty accidents on the back rank, and Kasparov believes Black can afford it as the black pieces are much-better coordinated than their white counterparts.

25.Rc3?!

The engines do not like this, preferring 25.Rc7 Rd7 and now 26.Rc3, when ...b6 is not possible, as will be explained in the next note.

25...b6! 26.Qa6

Not 26.Qxb6 Qg5!, eg 27.Ne4 Nf3+ (27...d2 is also strong) 28.Bxf3 Bxf3 29.Qc6 (29.Nxg5? Rxe1+ 30.Kh2 Rh1#)  d2 etc.

If the game had gone 25.Rc7 Rd7 26.Rc3, then 26...b6? 27.Qxb6 Qg5?! runs into 28.Rxd3! Re6 29.Rxd4 Rxb6 30.Rxd7, which the engines reckon is at least equal or may even favour White.

26...Qg5 27.Rxd3

This is best but hopeless, according to the engines. As before, Ne4 fails to ...Nf3+ or ...d2 (or ...Rxe4).

27...Qg6

Other moves also win, including the simple 27...Bxd3.

28.Bf1

White's rook is pinned because of the backward capture ...Bxa6.

28...Qxd3 29.Qxd3 Bxd3 30.Rxe8+ Rxe8 31.Bxd3 Re1+ 32.Bf1 Ra1 33.Nc4 b5 34.Nd6 Rxa2 35.Nxb5 Nxb5 36.Bxb5 Rxb2 37.Bc4

I suspect Nikolić felt like resigning here, but it is instructive to see how Kasparov brings home the full point. A noteworthy aspect is the way Kasparov holds back his pawns while he manoeuvres to get his pieces, especially the king, to their best-possible squares.

37...Rc2 38.Bd5 Kf8 39.h4

It is no improvement to put the pawns on the same-coloured squares as the bishop in order to give the bishop a secure post as Black's pieces would then invade on the dark squares.

39...g6 40.Kg2 Ke7 41.Kf3 Rc7 42.Ke4 Kd6 43.Ba2 Re7+ 44.Kd4 Ra7 45.Bb3 Ke7 46.Ke4

If 46.Ke5 to try to prevent the black king advancing, White's pawns are too vulnerable, eg one line given by the engines runs 46...Ra3 47.Bd5 (covering a2 and f3) Ra4 (threatening 48...f6#) 48.Bc6 f6+ 49.Kd5 Ra2 50.f3 Rg2 51.g4 Rh2 etc.

46...Kf6 47.Bd5 Re7+ 48.Kf4 Re2 49.Kf3 Rd2 50.Bc4 Rd4 51.Ba2 Rd7 52.Bc4

52.Ke4? Rd2.

52...Ke5 53.Ba2

53.Ke3 cannot keep the black king back for ever, eg 53...Rc7 54.Ba2 Ra7 55.Bc4 Ra3+ 56.Bd3 Rb3 57.f3 Ra3 etc.

53...Kd4 54.Bb1 Rc7 55.Ba2 Ra7 56.Bb1 Ra1 57.Bc2 Kc3 58.Be4 Kd2 59.Kg2

59.Bd5Ra5 60.Bxf7?? Rf5+.

59...Ke1 60.Bd5 Ra7 61.f4 Ke2 62.h5!?

This is best, according to the engines, but breaking up the black pawns does not give drawing chances because the rook is so much more powerful than the bishop in this ending.

62...gxh5 63.Bf3+ Ke3 64.Bxh5 Ra2+ 65.Kh3 f6 66.Be8 Kf2 67.Kh4 Ra8 68.Bc6 Rg8 69.g4 Ke3 70.Kg3

70.f5 Kf4.

70...f5 71.Bf3 h5 0-1

If 72.g5 then 72...h4+ wins.

Garry Kasparov's Forgotten Weapon Against The Queen's Gambit (part five)

IN the same tournament in which Kasparov beat Yasser Seirawan he faced Bent Larsen.

Larsen (2555) - Kasparov (2690)

Nikšić (Yugoslavia) Round 7 1983

1.c4 e6

Kasparov angles for another opportunity to play the Tarrasch Defence.

2.Nc3

Larsen indicates he is happy to oblige. He could have avoided the Queen's Gambit with, for example, 2.Nf3, when 2...d5 3.g3 c5 could be classified as a Réti or as a Symmetrical English.

2...d5 3.d4 c5 4.cxd5 exd5 5.Nf3 Nc6 6.g3 Nf6 7.Bg2 Be7 8.0-0 0-0 9.b3!?

This move goes back to at least 1920 but is dubbed the Uhlmann Variation in Play The Tarrasch by Shamkovich & Schilller (Pergamon (1984). As is pointed out in the book, the position can arise from a Nimzowitsch-Larsen Opening, ie 1.b3.

9...Ne4!?

This is the main move in ChessBase's 2021 Mega database, and is called "very sharp" by Shamkowich & Schiller.

10.Bb2 Bf6 11.Na4

A reasonably popular alternative is 11.e3, when 11...Bg4 12.h3 (12.Ne2 is the main move, but the engines slightly prefer the text) cxd4?! (Black should probably capture on f3 or withdraw the dark-square bishop) 13.Nxd5! was better for White in Spanton (1968) - John G Cooper (Hull - 2193), Hull 2017 (but ½–½, 53 moves).

11...Re8!?

More popular, but not by much, is 11...b6 (note that after 11.Na4, White is not threatening 12.Nxc5?? as that loses to 12...Nxc5 13.bxc5 Bxb2, but White is threatening 11.dxc5).

12.Rc1

12.dxc5 does not even come close to winning a pawn as Black replies 12...Bxb2 13.Nxb2 Nc3 etc.

12...b6!?

This had been played before but largely slipped under the radar because of Spassky's 12...cxd4 13.Nxd4 Bxd4 14.Nxd4 Bf5, when White has the bishop-pair and a slight edge, according to my main analysis engines Stockfish12 and Komodo11.01, but Black's position looks solid enough.

Today the text is the main move.

13.dxc5 Bxb2 14.Nxb2 bxc5 15.Na4

Black's hanging pawns come under immediate pressure.

15...Ba6

Active defence. Black's position looks problematic after 15...Nb4?! 16.a3 Na6, while the pawn sacrifice 15...c4?! 16.bxc4 d4 did not work out in Attila Jakab (2367) - Nathaniel Graham (2346), Budapest IM-A 2002, after 17.Nd2 Bf5 18.Nxe4 Bxe4 19.Bxe4 Rxe4 (1-0, 60 moves).

16.Re1

Not 16.Nxc5? Nxc5 17.Rxc5 Bxe2 etc.

16...c4 17.Nh4?!

This move has not find favour with later players. Today the main line runs 17.Nd2 Qf6 18.Nxe4 dxe4 19.bxc4 Rad8 20.Qb3 e3!, when Black has full compensation for a pawn, according to the engines, eg 21.fxe3 Qh6 24.Nc5 Rxe3 25.Qb2 Ne5, which led to a quick draw in a 'freestyle', ie computer-assisted, rapid game in 2006.

17...Qa5 18.Nf5 g6 19.Nd4 Rac8 20.h4?!

The engines dislike this, reckoning White keeps the balance with 20.Bh3 f5 (20...Rc7? 21.Nxc6 Rxc6 22.Bd7) 21.Nxc6 Rxc6 22.Bg2.

20...Ne5 21.Bh3 Rc7 22.Nc2?

This allows a combination. The engines prefer 22.Kg3, albeit reckoning Black is on top.

22...cxb3

Even stronger is the engines' 22...Nd3! 23.exd3 cxd3 24.Re3 (24.Ne3 Rxc1 25.Qxc1 d2) dxc2, when 25.Rxc2? only makes matters worse after 25...d4 26.Re1 Rxc2 etc.

23.axb3 Bc8 24.Bg2 Ng4 25.Rf1 Bd7?!

And here almost certainly stronger is 25...Ba6, eg 26.Nd4 Rec8, eg 27.Rxc7 Qxc7 28.Bxe4 dxe4, when the threat of ...e3 is extraordinarily hard to meet.

26.Ra1 Bxa4 27.Rxa4 Qc3 28.Bxe4

Forced, but good.

28...dxe4 29.e3?!

White should have protected the b pawn, for example by 29.Na3, according to the engines.

29...Qxb3 30.Rxe4 Rxe4 31.Qd8+ Kg7 32.Qxc7 Rxc4 33.Nd4 Rxc7 34.Nxb3 Rc2

Larsen correctly calculated that 29.e3?! did not lose material, but the ending favours Black thanks to his farside passed pawn and more-active pieces.

35.Nd4 Ra2 36.e4 Rd2?!

Black should have got on with it on the queenside with 36...a5, according to the engines, who reckon the text allows White to equalise.

37.Nc6 a6 38.e5

Stockfish12 is happy with this, but Komodo11.01 does not like it. Perhaps a surer draw is to be had with 28.f3!?, eg 38...Ne3 39.Re1.

38...Re2 39.Ra1 Rxf2 40.Rxa6 Rc2 41.h5 Kh6

Black is a pawn up but with almost zero winning chances after 41...gxh5?!

42.hxg6 hxg6 43.Ra4 Kg5 44.Nd4 Rc3 45.e6 Rxg3+ 46.Kh1!?

Kasparov gave this a question mark, but it seems that the losing move comes later.

46...f5 47.e7 Re3 48.Nc6 f4 49.Ra5+!?

Kasparov also gave this a question mark, but again that may not be right.

49...Kh4 50.Ra8 Nf6

White to play and draw

*****

*****

*****

*****

*****

51.Kg2?

This is the losing move, according to the engines.

I will explain later why they reckon 51.Kh2 is correct, but for now here is how the game ended:

51...f3+ 52.Kf1

52.Kf2 Ng4.

52...Kg3 53.Nd4 Ng4 54.Nxf3 Rxf3+ 55.Kg1 Nh2 56.Rf8 Rc3 0-1

Going back to the last diagram, White seems to draw with 51.Kh2, after which ...f3 does not come with check. The engines reckon 51...f3 is still best, but then comes 52.e8=Q! Nxe8 53.Ra4+ Kh5 54.Kg3, after which there seems no way for Black to save the f pawn, and then White should be able to give up his knight for the other pawn and reach a reasonably easy-to-draw ending, at least for a player of Larsen's calibre, of rook versus rook and knight. Best play seems to go 54...g5 55.Nd4 Re6 (55.Rc3 Ne5) 56.Nd4 Rf6 57.Nxf3! g4 58.Ra5+ Kh6 59.Ra3 gxf3 60.Rxf3 - Draw.

Garry Kasparov's Forgotten Weapon Against The Queen's Gambit (part four)

KASPAROV won his 1983 candidates quarter-final against Alexander Beliavsky by the score of +4=4-1.

His next event, five months later, was a tournament in Yugoslavia, where in round one he was drawn against the Damascus-born American grandmaster Yasser Seirawan.

Seirawan (2600) - Kasparov (2690)

Nikšić 1983

1.d4 d5 2.c4 e6 3.Nc3 c5 4.cxd5

Not everyone always wants to challenge Black in the main lines - even such greats as Karpov, Kasparov and Carlsen are among those who have played 4.e3. The most popular continuation then is 4...Nf6 5.Nf3 Nc6 6.a3!? a6!? 7.dxc5 Bxc5 8.b4 Ba7 9.Bb2 0-0, reaching a position that the analysis engines Stockfish12 and Komodo11.01 reckon slightly favours White.

Another popular choice is 4.Nf3, although this usually transposes to the game continuation via 4...Nc6 5.cxd5 exd5.

4...exd5 5.Nf3

White can grab a pawn with the somewhat underrated 5.dxc5!?, but then comes 5...d4 6.Na4!? (this is much less popular than 6.Ne4 in ChessBase's 2020 Mega database, but scores 15 percentage points better) b5!? (this sharp response is the only move stopping White from getting an advantage, according to the engines) 7.cxb6 axb6 8.b3, and now the engines reckon the little-played 8...b5!? 9.Nb2 Bb4+ gives Black a slight edge, despite being a pawn down. A possible improvement for White is 8.e3!? (marked as a blunder in Play The Tarrasch by Shamkovich & Schiller) Bb4+ 9.Bd2 Bxd2+ 10.Qxd2 (Shamkowich & Schiller may have missed that this recapture is possible) Rxa4 11.Bb5+ Bd7 12.Bxa4 Bxa4 with an unclear position that the engines reckon is dynamically balanced.

Instead of grabbing a pawn, White can sacrifice one with Frank Marshall's 5.e4!?, when the main continuation is 5...dxe4 6.d5 (Marshall first played 6.Bb5+!? but later switched to 6.d5) Nf6 7.Bg5 Be7 8.Bb5+. Black then usually plays 8...Kf8?!, but White has tremendous compensation for a pawn. The engines prefer 8...Bd7 9.d6!? (fear of this move seems to account for the popularity of 8...Kf8?!) Bxb5! 10.Nxb5 (10.dxe7 Qxd1+ 11.Rxd1 Bd3!) Qa5+ 11.Nc3 h6 12.Bh4 Bd8, when Stockfish12 prefers Black but Komodo11.01 reckons White has full compensation for a pawn. Instead of 6...Nf6, Black can try Dawid Janowski's 6...f5!?, which is recommended by Shamkowich & Schiller. The commonest continuation is 7.Bf4 Bd6, when White's best may be 8.Nh3, but 8...Nf6!? (better than Tarrasch's 8...a6, according to the engines) seems to leave White with not enough compensation for a pawn.

Although 5.Nf3 is overwhelmingly the most popular move, some of these other continuations illustrate how sharp the Tarrasch Defence can become at an early stage.

5...Nc6 6.g3 Nf6 7.Bg2 Be7 8.0-0 0-0 9.Bg5 cxd4 10.Nxd4 h6 11.Bxf6!?

This "does not make much sense," according to Shamkowich & Schiller, but has been played by Efim Geller and Oleg Romanishin, as well as by Seirawan (three times in 1983).

11...Bxf6 12.Nb3!?

Two other continuations are more popular, but neither 12.e3 Nxd4 13.exd4 Be6 nor 12.Nxc6 bxc6 13.Rc1 Re8 seems to give White anything.

12...d4

Pushing the d pawn is thematic in IQP positions, but it is not the only move here.

The earlier game Seirawan (2600) - Murray Chandler (2500), OHRA (Netherlands) 1983, saw 12...Be6!? 13.Rc1 (13.Nxd5 Bxb2 14.Rb1 Ba3 is roughly equal, according to the engines) Rc8!? 14.Nc5 Qe7 15.Nd3 Rfd8 16.Nf4 with maybe a slight edge for White, according to the engines (½–½, 68 moves).

At the same tournament, Seirawan (2600) - Eric Lobron (2515) went 12...Bg4!? 13.Bxd5 Bxc3 14.bxc3 Qf6 15.Qc2 Rac8, when the engines reckon Black does not quite have enough compensation for a pawn (but 0-1, 24 moves).

13.Ne4 Be7 14.Rc1 Qb6 15.Nec5 Rfd8 16.Rc4

The engines prefer 16.Qd3 Bxc5!? 17.Nxc5!? Qxb2 18.Rb1 Qxa2 19.Nxb7 Bxb7 20.Rxb7 Rac8, when they reckon White has compensation for a pawn, but no more.

16...Bxc5

The engines reckon 16...a5 gives Black a slight edge.

17.Nxc5

17.Rxc5 Be6 is slightly better for Black.

17...Qxb2 18.Qc2 Qxc2 19.Rxc2

White is a pawn down but his pressure against Black's queenside gives him equal chances, according to Komodo11.01, although Stockfish12 prefers Black.

19...Rb8 20.Rb2

Milorad Knežević (2425) - Milan Draško (2460), Yugoslav Championship (Budva) 1986, went 20.Nxb7 Bxb7 21.Bxc6 Bxc6 22.Rxc6 Rb2 with a slight edge to Black.

20...Rd6 21.Rd1 b6 22.Nb3 Bb7 23.Rbd2 Rbd8 24.Kf1 Ba6 25.Bxc6 Rxc6 26.Nxd4 Rc5 27.Nb3 Rxd2 28.Rxd2 Rc7

How would you assess this ending?

*****

*****

*****

*****

*****

Black has bishop versus knight on an open board with rival pawn-majorities, a queenside pawn-majority and the better minor piece for cooperating with a rook ... but the position is almost dead-equal, according to the engines.

29.Rd8+ Kh7 30.Ke1 Bc4 31.Kd2 g6!? 32.Nc1 Kg7 33.a3 Kf6 34.e3!?

Kasparov called this dubious but it seems it is OK if followed up correctly. However, White has an easier time of it after the engines' 34.Nd3.

34...Ke7 35.Rd4 Rd7!?

Although rook and bishop usually work together better than rook and knight, and there is no reason to think this position is an exception, Kasparov decides bishop-versus-knight with rival pawn-majorities (and no rooks) gives better winning chances.

36.Kc3 Rxd4 37.Kxd4 b5

The bishop is maintained on c4, from where it dominates the knight.

38.Nd3!?

This is Stockfish12's choice, but Komodo11.01 reckons it is the losing move.

38...Bxd3!?

Although a bishop is usually better than a knight on an open board with rival pawn-majorities, and there is no reason to think this position is an exception, Kasparov decides he has better winning chances in the pawn ending.

What makes this ending remarkable is that Kasparov goes against accepted wisdom on two major occasions - i) swopping rooks, ii) swopping minor pieces - in order to win, although it has to be said he gets a big assist from his opponent.

39.Kxd3 Kd6 40.e4?

Kasparov called this a blunder, and it does seem this is the real losing move. Drawing, according to Stockfish12, is 40.Kd4, which conforms with first world correspondence champion Cecil Purdy's rule-of-thumb that in an ending you should get your pieces to their best possible squares before advancing pawns. Komodo11.01, however, reckons for a long time that 40.Kd4 is also losing. But after the engines' joint-continuation 40...f5 41.e4 fxe4 42.Kxe4 Kc5 43.g4 a5 44.f4 they agree the ending is drawn.

40...g5!?

Kasparov gave this an exclamation mark, but the engines reckon 40...Kc5 also wins.

41.f4 gxf4 42.gxf4 Kc5 43.Kc3 a5 44.Kd3

If 44.e5 then 44...Kd5.

44...h5

Also winning is 44...b4.

45.h4 b4 46.a4

Equally hopeless is 46.axb4+ axb4, eg 46.f5 f6 47.e5 fxe5 48.f6 Kd6 49.Kc4 Ke6 50.Kxb4 Kxf6 51.Kc4 Kf5 52.Kd3 Kg4 53.Ke4 Kxh4 54.Kxe5 Kg3 etc.

46...f6 47.f5 Kc6 48.Kc4 Kc7 49.Kd3

Or 49.Kd4 Kd6.

49...Kd7

Black takes the distant opposition.

50.Ke3

Or 50.Kc4 Kc6, which is the same position as after 48.Kc4 but with White to move rather than Black.

50...Kc6 51.Kd3

If 51.Kd4 then 51...Kd6 52.Kd3 Ke5 53.Ke3 b3 etc.

51...Kc5 52.Ke3

Or 52.e5 fxe5 53.f6 Kd6.

52...b3 53.Kd3 Kb4 54.e5 Ka3 0-1

Garry Kasparov's Forgotten Weapon Against The Queen's Gambit (part three)

KASPAROV'S use of the Tarrasch Defence in his first Candidates quarter-final game with the black pieces against Alexander Beliavsky was both surprising - it was the first time in his career he answered 1.d4 with 1...d5 - and successful.

Beliavsky no doubt came to the board armed with an improvement the next time he had the white pieces, but Kasparov switched to the Nimzo-Indian Defence ... and lost, levelling the match score at one win each with two draws.

Kasparov restored his lead in game five by winning on the white side of the Exchange Variation of the Queen's Gambit Declined.

In the next game it was Beliavsky's turn to play the Queen's Gambit.

Beliavsky (2570) - Kasparov (2690)

Candidates Quarter-Final (Moscow) Game 6 1983

1.d4 d5 2.c4 e6 3.Nc3 c5 4.cxd5 exd5

This is not compulsory. Black scores four percentage points better in ChessBase's 2021 Mega database by playing 4...cxd4!?, the Von Hennig-Schara Gambit. The main line runs 5.Qxd4 Nc6 6.Qd1 exd5 7.Qxd5 Bd7 8.Nf3 Nf6 9.Qd1 Bc5 10.e3 Qe7 11.Be2 0-0-0 12.0-0 g5!? 13.b4!? Bxb4 14.Bb2 g4 15.Nd4. Mega21 has 40 games with this position, the most-recent high-level encounter being Simon Kim Williams (2442) - Spyridon Kapnisis (2464), Dundee 2017, which continued 15...Kb8 16.Ncb5 Nxd4 17.Bxd4 Bxb5 18.Bxb5 Bc3 19.Rb1 Bxd4 20.exd4 Rd6 with a position that Stockfish12 reckons is much better for White, but Komodo11.01 rates as equal (but 0-1, 50 moves). There are lots of alternatives along the way for both sides, starting right at the beginning when White often plays 5.Qa4+. The main line then runs 5...Bd7 6.Qxd4 exd5 7.Qxd5 Nc6, which is a transposition to the main line after 5.Qxd4.

5.Nf3 Nc6 6.g3 Nf6

Black has a major alternative with the Swedish Variation, 6...c4!?, which scores five percentage points better in Mega21.

The starting position of the Swedish Variation

Shamkovich & Schiller in Play The Tarrasch (Pergamon 1984) call this "one of the most-dynamic plans at Black's disposal." They add: "The line has constantly been 'refuted' and resurrected, and both White and Black must be well-versed in theory in order to survive the opening. Although the variation is considered suspect it still has great surprise value and in many cases Black can emerge relatively unscathed, provided that he chooses his lines wisely."

That is far from being a ringing endorsement, but 6...c4!? is quite liked by the engines, especially Stockfish12.

The main line runs 7.Bg2 (the thematic counter 7.e4?! is probably premature, eg 7...dxe4 8.Ng5 Qxd4 9.Be3 Qxd1+ 10.Rxd1 Bxg4 with a sharp but equal position, according to the engines) Bb4 8.0-0 Nge7 9.e4 (White is now much more ready for this break) dxe4 (9...0-0 is also popular) 10.Nxe4 0-0, and here the engines reckon 11.Be3 is just one move that gives White a large edge.

7.Bg2 Be7

The engines prefer trying to save a tempo with 7...cxd4!? 8.Nxd4 Bc5. The main line continues 9.Nb3, when the engines reckon 9...Bb6!? is an improvement on the more-popular 9...Bb4. Hikaru Nakamura (2761) - Daniil Dubov (2690), Fide Grand Prix (Moscow) 2019, went 9...Bb6!? 10.0-0 (10.Nxd5 can be met, according to Stockfish12, by 10...Be6 11.Nxb6 Qxd1+ 12.Kxd1 axb6, when Stockfish12 prefers White but Komodo11.01 reckons the position is equal. Komdo11.01 varies early with 11.e4, continuing 11...Bg4 12.Qd3 Nxd5 13.exd5 Ne5 14.Qb5+ Bd7 15.Qe2 0-0, when the engines agree Black has enough compensation for a pawn) d4 11.Na4 0-0!? 12.Bg5 Re8 13.Nxb6 axb6 14.e3 d3!? with an unclear but probably balanced position (½–½, 28 moves).

8.0-0 0-0 9.Bg5 cxd4 10.Nxd4 h6 11.Be3

The bishop goes here, rather than, say, f4, because on e3 it helps restrain the d pawn, possibly freeing the d4 knight to move at a crucial moment.

11...Re8 12.Qc2!?

So this is what Beliavsky prepared. The move is given an exclamation mark by Shamkovich & Schiller.

12...Bg4 13.Rfd1

The main line today is 13.h3 Bd7 14.Rad1 Qc8 15.Kh2 Bf8 (a common retreat in the Tarrasch, unmasking the king's rook), which first occurred in Beliavsky (2640) - Miguel Illescas Córdoba (2530), Linares (Spain) 1990. The engines reckon White has an edge (but 0-1, 42 moves).

13...Bf8 14.Rac1 Rc8 15.Nxc6!?

Transforming the pawn-structure is often available and has to be taken into account by Black. The engines prefer 15.Qa4.

15...bxc6

Clearly the d pawn is no longer a target for White, who should change his attention to the new, backward c pawn and, to a lesser extent, to the newly isolated a7 pawn. Black's chances, as is the case when he has an IQP, lie mainly on the kingside and down the e file.

16.Bd4

If Beliavsky entertained hopes of immediately capturing the black a pawn, he would have changed his mind on seeing 16.Bxa7?! c5! as 17.Bxd5? Qa5 18.Qa4 Qxa4 19.Nxa4 Nxd5 20.Rxd5 Ra8 21.Bxc5 Ra8, when the black bishop is worth more than White's three pawns. Less clear is the engines' 17.h3 Be6 18.a4! Qa5 19.Nb5, when White is a pawn up but badly discoordinated, and the engines prefer Black.

16...Bb4 17.Rd2

17.Bxa7!? Bxc3 18.Qxc3 Bxe2 19.Re1 is slightly better for White, according to the engines.

17...Qe7 18.a3 Ba5 19.b4 Bb6 20.e3 Qe6 21.Qb2 Bxd4 22.Rxd4 c5!?

Kasparov dissolves the backward pawn, confident the re-established IQP will not be a decisive liability.

23.bxc5

23.Rd2 Red8 24.Na4!? cxb4 25.axb4 Rxc1+ 26.Qxc1 Rc8 seems at least equal for Black.

23...Rxc5 24.Ne2 Rec8 25.Rxc5 Rxc5 26.Nf4 Qc8 27.h3

White's back-rank weakness means he cannot capture the d pawn, eg 27.Nxd5?? Nxd5 28.Rxd5 Rc1+ 29.Bf1 Bh3 etc.

27...Rc1+ 28.Kh2 Rc2 29.Qb3 Bf5 30.Kg1 Rc1+ 31.Rd1 Be4 ½–½

Kasparov gives a likely continuation as 32.Nxd5 Nxd5 33.Bxd5 Rxd1+ 34.Qxd1 Nxd5 35.Qxd5 Qc1+ 36.Kg2 Qxa3, which he regards as equal.

Garry Kasparov's Forgotten Weapon Against The Queen's Gambit (part two)

GARRY Kasparov is well-known for playing asymmetrical defences against 1.d4, ie starting with 1...Nf6 and often continuing with the King's Indian or Grünfeld.

At the September 1982 Interzonal in Moscow, which he won with a score of 10pts out of 13 (1.5pts clear of the field), putting him on the road to the world championship, Kasparov faced 1.d4 three times, always replying 1...Nf6.

At the Olympiad the following month in Luzern, he again faced 1.d4 three times and again always replied 1...Nf6.

Then came game two of his Candidates' quarter-final match against Alexander Beliavsky.

Beliavsky (2570) - Kasparov (2690)

Candidates Quarter-Final (Moscow) Game 2 1983

1.d4 d5!?

The first time, at least in ChessBase's 2021 Mega database, that Kasparov met 1.d4 with 1...d5. If nothing else, it must have rendered Beliavsky's preparation for the white pieces null and void.

2.c4 e6 3.Nc3

So we reach the starting point of what this series is about - what to do if you are a 1.d4 d5 2.c4 e6 player as Black but do not want to face the QGD Exchange when White is not committed to playing Nf3.

Kasparov's style with the black pieces, as several commentators have noted, is one in which he would rather have an unbalanced position with counterplay than a theoretically more-or-less equal position, but one in which White has the initiative.

So continuing 3...Nf6 4.cxd5 exd5 would be alien to Kasparov's style. Indeed, throughout his career he never met 1.d4 d5 2.c4 e6 3.Nc3 with 3...Nf6.

Instead, in this, his first ever 1.d4 d5 game as Black, Kasparov revived the Tarrasch Defence.

3...c5!?

Statistically, this is an improvement on 3...Nf6. The latter scores a miserable 39% in Mega21, while 3...c5 scores 45%, which is about average for Black in large databases.

4.cxd5

This is played 72% of the time in Mega21.

4...exd5 5.Nf3 Nc6

Shamkovich & Schiller in Play The Tarrasch (Pergamon 1984) state: "When heading for the Tarrasch it is important to remember the dictum that knights should be developed in alphabetical order!" Actually, there seems nothing wrong with 5...Nf6, but the text is a lot more popular.

6.g3

This move was introduced by Carl Schlecter at Prague 1908 and quickly became White's main weapon against the Tarrasch.

6...Nf6 7.Bg2 Be7

Other moves are possible, eg 7...Be6, but according to Shamkovich & Schiller "it is much too early to commit the (queen's) bishop to a particular post."

8.0-0

Black can be said to have lost a tempo after 8.dxc5 Bxc5, which is indeed the continuation favoured by the analysis engines Stockfish12 and Komodo11.01 In practice, games usually transpose to one of the main lines after the further moves 9.0-0 0-0 (more on this later).

8...0-0 9.Bg5

9.dxc5 Bxc5 would transpose to the previous note, although Black has the added option of 9...d4!? The text is the absolute main line in Mega21.

Shamkovich & Schiller comment: "White eschews subtlety in favour of a direct attack on the black centre. He threatens Bxf6 followed by Nxd5."

9...cxd4

The main alternative, 9...c4!?, is preferred by the engines.

A solid alternative, liked by Komodo11.01 but not by Stockfish12, is 9...Be6!?, when 10.Rc1 cxd4 11.Nxd4 Rc8? 12.Nxe6 fxe6 13.Bh3 gave White a large advantage in Burak Akguc (unrated) - Spanton (2065), Hastings Challengers 1997-98 (but ½–½, 53 moves). Black can improve with 11...Nxd4, but White can probably improve before that with 10.dxc5 Bxc5 11.Rc1.

10.Nxd4 h6

We are still very much in theory - there are almost 4,000 games with this move in Mega21.

11.Be3

I will cover the important alternative 11.Bxf6 in part four.

11...Re8

Decision time for White

This is possibly the first point at which either player needed a serious think.

*****

*****

*****

*****

*****

12.Qa4

White's important alternatives will be covered later.

12...Bd7!? 13.Rad1

White can win a pawn with 13.Nxd5!? Nxd5 14.Bxd5, but then comes the forcing 14...Nb4 15.Qb3 Nxd5 16.Qxd5 Bh3 17.Qxd8 Raxd8, when Black is considered by theory (and the engines) to have enough compensation for a pawn.

13...Nb4 14.Qb3 a5

"Black must play aggressively on the queenside to offset White's central pressure," according to Shamkovich & Schiller.

15.Rd2!?

This move has been widely criticised, but it is the choice of the engines.

The main line runs 15.a4 Rc8 16.Ndb5 Be6 17.Bd4 Bc5 with what the engines reckon is an even position (White can hardly play 18.Bxf6?! Qxf6 19.Bxd5 Bxd5 20.Nxd5 Nxd5 21.Qxd5 because Black equalises the material and gains an initiative with 21...Rxe2).

Note that the d pawn was shown to be immune in Rafael Vaganian - Borislav Ivkov, USSR-Yugoslavia Match (Odessa) 1975, which saw 15.Nxd5? Nbxd5 16.Bxd5 Nxd5 17.Qxd5 Bh3, when 18.Qxd8 fails to 18...Raxd8 19.Rfe1 Bb4.

15...a4 16.Qd1 a3 17.Qb1

The engines like 17.Qb3!? Qa5 18.bxa3 Qxa3 19.Nxd5 Nfxd5 20.Bxd5 Nxd5 21.Qxd5, when White is a pawn up but Black's bishop-pair and White's kingside light-square weaknesses give obvious compensation.

17...Bf8 18.bxa3

This is Komodo11.1's choice.

Stockfish12 prefers 18.Nb3, which was played in Vitaly Kunin (2494) - Ernst Weinzetti (2327), Oberwart (Austria) 2006, which continued 18...Qc8 19.Rc1 Bf5 20.Qa1 axb2 21.Qxb2 Rxe3!? 22.fxe3 Qe6, when Black had full compensation for the exchange, according to the engines (but 1-0, 34 moves).

18...Rxa3 19.Qb2 Qa8 20.Nb3!?

Unblockading an IQP is not a decision to be taken lightly, although the move is Stockfish12's choice. Kasparov called the move dubious, suggesting 20.Ndb5 (Komodo11.01's choice) Bxb5 21.Nxb5 Rxa2 22.Qb3, which he and the engines call equal. One possible continuation runs 22...Rxd2 23.Bxd2 Qa5 24.Nd4, re-establishing the blockade, when Black is a pawn up but has two weak pawns, and White has a pair of bishops on an open board - still equal, according to the engines.

20...Bc6 21.Bd4

This is Beliavsky's idea - the dark-square bishop becomes the blockader, indirectly adding pressure against d5.

21...Ne4 22.Nxe4

The engines prefer 22.Bxe4!? dxe4 23.Be3, claiming equality.

22...dxe4 23.Ra1

Kasparov in Mega21 gives as better 23.Nc5!? e3!? 24.Bxc6 exf2+ 25.Bxf2 Nxc6 with a slight edge for Black.

23...Bd5 24.Qb1?

The engines reckon White needs to block the e4 pawn with 24.e3, the point being Black cannot reply, as in the game, by occupying the newly created outpost at d3 as 24...Nd3? loses to 25.Rxd3! If Black tries to set up ...Nd3 with 24...b6, similar to what happens in the game, White has 25.Nc1 and is only a little worse.

24...b6

Now Black is threatening to play ...e3.

25.e3 Nd3 26.Rd1

The engines prefer 26.Nc1, which Kasparov said can be met by 26...Ne1, when one line given by the engines runs 27.Qxb6 Nf3+ 28.Bxf3 exf3 - White is a pawn up but Black is better, according to the engines.

As Kasparov points out, 26.Bxb6? loses to 26...Bb4 27.Rc2 (stopping ...Bc3) Bxb3.

26...b5 27.Bf1 b4 28.Bxd3!?

This gets rid of the strong knight but opens the h1-a8 diagonal, which may be a greater evil.

28...exd3 29.Qxd3 Rxa2 30.Rxa2 Qxa2 31.Nc5

Not 31.Rb1? Be4.

After the text, material is level but White is positionally lost - it is just a question of whether Black will win by pushing the b pawn or by exploiting the weak light squares around the white king.

The game finished:

31...Bf3 32.Ra1 Qd5 33.Qb3 Qh5 34.Nd3 Bd6 35.Ne1 Bb7 36.Rc1 Qf5 37.Rd1 Bf8 38.Qb1 0-1

White lost on time as he made his 38th move. Kasparov gives a likely continuation as 38...Be4 39.Qb3 h5, when Black has a large advantage.