Chapter 1 of 22

How the Caro-Kann works

White 48%Draws 4%Black 48%

1,635,111 games

Lichess, 1600 to 1999, September 2026, all speeds

The Caro-Kann starts

1.e4

c6.

The c-pawn move prepares 2...d5, and if White ever takes on d5, the c-pawn is ready to take back.
The point is to fight for the center without shutting in the light-squared bishop, which is the problem 1...e6 creates in the French Defense.
The price is time: if Black later wants ...c5, the c-pawn has moved twice, and the queen's knight cannot use c6 while the pawn stands there.

After

2.d4

d5, White's third move decides the kind of game you get.

3.e5 closes the center and takes space.
3.exd5 cxd5 trades White's central e-pawn for Black's c-pawn and leaves a symmetrical center, unless White follows with 4.c4.
3.Nc3 or 3.Nd2 lets Black take on e4, and after 4.Nxe4 Black chooses between developing the bishop with 4...Bf5, the queen's knight with 4...Nd7, or the king's knight with 4...Nf6.
3.f3, the Fantasy Variation, offers to rebuild the center with a pawn.

In September 2026, Lichess players both rated 1600 to 1999 reached

1.e4

c6 in 1,635,111 rated games, and White and Black each won 48% of them.

White answered 2.d4 in 713,668 of those games, about 44%.
2.Nf3 came next with 449,211, then 2.Nc3 with 150,559 and 2.Bc4 with 83,821.
So at club level more than half of White's second moves are not 2.d4, and the sidelines deserve as much of your preparation as the main line.

Lichess, 1600 to 1999, September 2026, all speeds

To play the Caro-Kann as Black, you need one answer to each of four replies: the Advance, the Exchange and Panov, the 4.Nxe4 positions, and the moves where White avoids

2.d4.

In the main lines Black makes very few compromises in pawn structure and later plays ...c5 to contest d4, which is where the opening's solid reputation comes from.

White’s plan

Take the center with

2.d4, or delay it with

2.Nf3,

2.Nc3 or

2.c4.

Black’s plan

Answer almost everything with 2...d5.

Chapter 2 of 22

Caro-Kann starting moves: 1.e4 c6 2.d4 d5

White 49%Draws 4%Black 47%

669,490 games

Lichess, 1600 to 1999, September 2026, all speeds

White builds a two-pawn center and Black strikes at it at once with ...d5, backed by the c6 pawn.

The e4 pawn is attacked, so White has to decide now: push it, trade it, or defend it with a piece or a pawn.

White’s plan

Pick the structure you want and learn Black's main answers to it: space with

3.e5, a symmetrical or isolated-pawn center after

3.exd5, or quick piece play after

3.Nc3 and 4.Nxe4.

Black’s plan

Develop the light-squared bishop before ...e6 when you can, and use ...c5 to hit d4 at the right moment.

In serious games

2...d5 is chosen more than 95 times in 100. It hits e4 with a pawn the c6 pawn supports, so

3.exd5

cxd5 only trades White's central e-pawn for Black's c-pawn.

White has four ways to meet the attack on e4: close the center with

3.e5, capture with

3.exd5, defend with a knight on c3 or d2, or defend with a pawn on f3.

At 1600 to 1999 on Lichess in September 2026, the position after

2...d5 came up in 669,490 games.

White chose

3.e5 in 230,020 and

3.exd5 in 228,859, almost exactly as often, then

3.Nc3 in 137,208, with

3.Nd2 and

3.f3 at about 28,900 each.

Lichess, 1600 to 1999, September 2026, all speeds

Chapter 3 of 22

Advance Variation

White 48%Draws 4%Black 48%

230,020 games

Lichess, 1600 to 1999, September 2026, all speeds

White pushes past d5 and takes space with a d4 and e5 pawn chain.

Black's usual counter is

...c5, hitting the base of that chain at d4, either at once or after first developing the bishop to f5.

White’s plan

Keep the e5 pawn and the space it gives you, develop the kingside, and be ready to answer

...c5 and ...cxd4 by taking back with the knight.

Against a bishop on f5 you can also chase it with g4 and h4.

Black’s plan

Develop the bishop outside the chain with

...Bf5, then play

...e6 and

...c5 against d4.

The king's knight usually goes to e7, because the e5 pawn covers f6.

3.e5 closes the center and gains space, and it is now the main line at all levels.

Black has two answers:

3...Bf5, which develops the bishop before

...e6 can lock it in, and

3...c5, which attacks d4 straight away.

Once Black's e-pawn reaches e6, a bishop still on c8 would be stuck behind its own pawns, as in the French.

With the bishop already on f5, Black can play ...e6 and ...c5 and gets an improved version of the French Advance.
Reaching c5 costs the c-pawn a second move, but White has also moved the e-pawn twice, so Black does not fall behind.

At 1600 to 1999 on Lichess in September 2026,

3.e5 came up in 230,020 games; Black answered

3...Bf5 in 136,505 and

3...c5 in 73,909.

3...c5 did well at that level: Black won 50% of those games and White 46%.

Lichess, 1600 to 1999, September 2026, all speeds

Botvinnik-Carls Defense

White 46%Draws 4%Black 50%

73,909 games

Lichess, 1600 to 1999, September 2026, all speeds

Black hits d4 at once instead of developing the bishop first.

This sharp line avoids much of the theory after 3...Bf5.
White can take with 4.dxc5, which loosens the center, and Black usually wins the pawn back.

White’s plan

Nc6

7.Bd3 keeps Black's light-squared bishop from a good square.

Black’s plan

Against

4.dxc5, hit e5 with ...Nc6, play ...e6 and take back on c5 with the bishop.

Chapter 4 of 22

Advance Variation: 3...Bf5

White 49%Draws 4%Black 47%

136,505 games

Lichess, 1600 to 1999, September 2026, all speeds

The bishop is out, but it stands alone on f5 with no pawn to support it.

White chooses between quiet development and chasing the bishop with pawns.

White’s plan

Either develop calmly with

Nf3,

Be2 and castling, or treat the bishop as a target with h4 or g4 before Black gives it a retreat square.

Black’s plan

Play ...e6 and ...c5 against d4, and when White pushes h4, answer with ...h5 or ...h6 so the bishop always has a square.

The quiet main line is the Short Variation,

4.Nf3

e6

5.Be2.

White castles and prepares to answer ...c5 and ...cxd4 with Nxd4.
Black can break at once with ...c5, or develop first and break later on the kingside with ...f6.

4.Nc3

e6

5.g4 is the Van der Wiel Attack: the knight takes the e4 square away from the bishop, and White chases it with g4 and then h4 and h5.

After

5...Bg6

6.Nge2

c5

7.h4, White threatens h5 to trap the bishop, and the usual

7...h5

8.Nf4

Bh7

9.Nxh5 gives up a pawn for active play.

4.h4, the Tal Variation, attacks the bishop more directly, and

4.g4, the Bayonet Attack, more directly still.

Both gain time on the bishop, but they loosen White's kingside before White has castled or developed a piece.

At 1600 to 1999 on Lichess in September 2026, White's most common fourth moves here were

4.Nf3 (31,639 games),

4.h4 (17,071),

4.c3 (12,612),

4.g4 (9,590) and

4.Nc3 (6,549).

4.h4 was White's best scorer of those, winning 54% of its games against Black's 42%.

Lichess, 1600 to 1999, September 2026, all speeds

Bayonet Attack

White 52%Draws 3%Black 45%

9,590 games

Lichess, 1600 to 1999, September 2026, all speeds

The Bayonet Attack chases the bishop at once.

It gains time but weakens White's kingside before any piece is developed.

Short Variation

White 48%Draws 4%Black 48%

31,639 games

Lichess, 1600 to 1999, September 2026, all speeds

The Short Variation: White develops quietly and castles.

After

4...e6

5.Be2, White is ready to meet

...c5 and ...cxd4 with Nxd4.

Black’s plan

Break with

...c5, or develop first and break later with

...f6.

Tal Variation

White 54%Draws 3%Black 42%

17,071 games

Lichess, 1600 to 1999, September 2026, all speeds

The Tal Variation goes straight at the bishop.

4...e6 loses the bishop to

5.g4

Be4

6.f3

Bg6

7.h5;

4...h5 is the usual reply.

Van der Wiel Attack

White 46%Draws 3%Black 50%

6,549 games

Lichess, 1600 to 1999, September 2026, all speeds

The Van der Wiel Attack: Nc3 takes e4 from the bishop, and White follows with g4 and h4 to chase it.

After

Bg6

c5

h5

Bh7

9.Nxh5, White wins a pawn, but Black gets good compensation.

Chapter 5 of 22

Exchange Variation

White 48%Draws 5%Black 47%

228,859 games

Lichess, 1600 to 1999, September 2026, all speeds

White trades on d5 and Black takes back with the c-pawn.

The center becomes symmetrical, with White's d4 pawn facing Black's d5 pawn, and the c6 square is free again for Black's knight.

White’s plan

Develop with Bd3, c3 and Bf4, keep Black's bishop off f5, and use Qb3 to hit b7 and d5.

Black’s plan

Put the knights on c6 and f6, develop the light-squared bishop to g4 before

...e6, then play

...e6 and castle.

After

3.exd5

cxd5, White has traded the central e-pawn for Black's c-pawn.

The main line is 4.Bd3, which develops and stops ...Bf5 at the same time, and it usually goes 4...Nc6 5.c3 Nf6 6.Bf4 Bg4 7.Qb3.
The pawns form a Carlsbad structure with the colors reversed, so the plans resemble the Queen's Gambit Declined Exchange Variation with White and Black swapped.
This line is considered to give White a slightly better game, and Bobby Fischer used it.

The sharper 4.c4 is the Panov Attack, which changes the structure.

At 1600 to 1999 on Lichess in September 2026, White's most common fourth move after

3...cxd5 was

4.Nf3 (66,649 games), ahead of 4.Bd3 (50,158) and 4.c4 (49,253).

After 4.Nf3 Nc6, Black won 49% of 26,478 games and White 46%.

Lichess, 1600 to 1999, September 2026, all speeds

Rubinstein Variation

White 53%Draws 5%Black 42%

4,985 games

Lichess, 1600 to 1999, September 2026, all speeds

The Rubinstein line of the Exchange: Bd3 stops

...Bf5, c3 holds d4, and Bf4 develops the last minor piece.

It is considered slightly better for White.

Chapter 6 of 22

Panov-Botvinnik Attack

White 51%Draws 4%Black 45%

49,253 games

Lichess, 1600 to 1999, September 2026, all speeds

White attacks d5 again with the c-pawn.

Once the c-pawns trade, White is usually left with an isolated d-pawn: a long-term weakness, paid for with quick development, control of e5 and kingside attacking chances.

White’s plan

Develop fast with Nc3, Nf3 and the bishops, use the grip on e5, and look for a kingside attack before the endgame arrives.

Black’s plan

Develop with

...Nf6,

...e6 and ...Bb4 or ...Be7, castle, and steer toward an endgame where the isolated d4 pawn is a target.

4.c4 hits d5 a second time and changes the pawn structure.

The d4 pawn is not isolated yet on move four; it becomes isolated once White's c-pawn is traded on d5.
Black almost always answers 4...Nf6, and the main line runs 5.Nc3 e6 6.Nf3.
Then 6...Bb4 often reaches positions from the Nimzo-Indian Defense, while 6...Be7 was once the most common move.

White's compensation for the isolated pawn is rapid development, a grip on e5 and attacking chances on the kingside.

The pawn is a weakness in an endgame, so White has to make the most of the middlegame, and the longer the game goes, the more that weakness helps Black.

At 1600 to 1999 on Lichess in September 2026,

4.c4 came up in 49,253 games, and White won 51% against Black's 45%.

After

4...Nf6

5.Nc3, Black's choices split between

5...Nc6 (4,506 games), 5...e6 (4,040) and 5...g6 (1,077).

Lichess, 1600 to 1999, September 2026, all speeds

Main Line

White 50%Draws 5%Black 45%

744 games

Lichess, 1600 to 1999, September 2026, all speeds

The Panov main line:

6...Bb4 pins the c3 knight and often reaches Nimzo-Indian positions.

White plays for activity and e5, Black for the isolated pawn later.

Chapter 7 of 22

After 4.Nxe4: Black's main choices

3.Nc3 or 3.Nd2 dxe4 4.Nxe4

White 51%Draws 4%Black 45%

125,145 games

Lichess, 1600 to 1999, September 2026, all speeds

Black gives up the center pawn with ...dxe4, and White's knight takes back on e4.

Now Black picks one of three plans: develop the bishop with

4...Bf5, prepare a knight trade with

4...Nd7, or offer one at once with

4...Nf6.

White’s plan

Use the extra space in the center, develop quickly, and decide how to meet the piece Black has chosen to bring out.

Black’s plan

Develop without weakening the pawns, castle, and strike at d4 with

...c5 at the right moment.

3.Nc3 and

3.Nd2 usually reach the same position after

3...dxe4

4.Nxe4, so you can learn them together.

The two move orders differ only when Black avoids taking on e4, mainly with

3...g6.

4...Bf5, the Classical, develops the bishop with an attack on the e4 knight.

4...Nd7, the Karpov, prepares ...Ngf6 so that Black can take back with a knight if White trades on f6.
4...Nf6 offers the trade at once and accepts doubled pawns after 5.Nxf6+.
4...h6, the Finnish Variation, is a rare fourth choice.

Black has no pawn in the center now, while White keeps the d4 pawn and with it extra space.

Black's answer is a sound structure and a well-timed

...c5 against d4.

Chapter 8 of 22

Classical Variation

White 52%Draws 4%Black 44%

8,351 games

Lichess, 1600 to 1999, September 2026, all speeds

Black develops the bishop with an attack on the e4 knight.

White usually chases it with

5.Ng3 and then h4, and the bishop drops back to g6 and later h7, where it keeps working on the long diagonal.

White’s plan

Gain space with

Ng3 and h4 to h5, trade the light-squared bishops with

Bd3, and in the main lines castle queenside.

Black’s plan

Answer h4 with ...h6, drop the bishop to h7 when h5 comes, finish development with ...Nd7, ...e6 and ...Ngf6, and strike at d4 with ...c5.

The Classical is the most common way to handle the Caro-Kann, and much of the opening's solid reputation comes from it.

The main line is 5.Ng3 Bg6 6.h4 h6 7.Nf3 Nd7 8.h5 Bh7 9.Bd3 Bxd3 10.Qxd3.
White's h-pawn gains space and pushes the bishop back, and ...h6 gives the bishop its retreat square.
The pawn on h5 looks aggressive, but in an endgame it can turn into a weakness.

Black can castle queenside, the steady choice that gave the Caro-Kann its quiet reputation, or castle kingside or leave the king in the center for sharper play.

White's sidelines at move six are the Flohr, 6.Nh3, and the Maróczy Attack, 6.f4.

Through the

3.Nd2 move order,

4...Bf5 came up in 8,351 games at 1600 to 1999 on Lichess in September 2026. White answered

5.Ng3 in 5,169 of them and

5.Ng5 in 2,625;

5.Ng5 is the move that aims for the Martian Gambit.

Lichess, 1600 to 1999, September 2026, all speeds

Main Line

White 53%Draws 5%Black 42%

1,441 games

Lichess, 1600 to 1999, September 2026, all speeds

The Classical main line:

6.h4 asks the bishop where it will go, and

6...h6 gives it h7.

The h-pawn gains space but can become a weakness in an endgame.

Martian Gambit

White 63%Draws <1%Black 36%

204 games

Lichess, 1600 to 1999, September 2026, all speeds

The Martian Gambit: take the knight with

7...fxe6.

Stockfish 16 at depth 30 rates that clearly better for Black (-1.21).

Chapter 9 of 22

Karpov Variation

White 50%Draws 5%Black 45%

9,861 games

Lichess, 1600 to 1999, September 2026, all speeds

Black prepares ...Ngf6.

If White then trades knights on f6, Black takes back with the other knight and keeps the pawns undoubled.

White’s plan

Use the extra space and the moment before Black castles:

5.Ng5 or

5.Bc4 aim at f7 and e6.

Black’s plan

Play ...Ngf6 and, if White trades on f6, take back with the d7 knight; then develop the bishop, play ...e6 and castle.

The Karpov is also called the Steinitz, Smyslov or Modern Variation.

Its short-term goal is an early knight trade that leaves Black's pawn structure intact.
Compared with the Classical, Black keeps the bishop at home for now and does not have to put it on g6; the cost is that White gets more room in the center.
The knight on d7 also blocks the c8 bishop until it moves again.

White's sharpest try is

5.Ng5, with the knight eyeing f7 and e6 while Black's kingside is still undeveloped.

5.Bc4 leads to the Smyslov line, and 5.Nf3 Ngf6 6.Ng3 is the Kasparov Attack.
The last game of the Deep Blue versus Kasparov rematch, which Kasparov lost after a known blunder, was played in this variation.

At 1600 to 1999 on Lichess in September 2026,

4...Nd7 came up in 1,949 games through

3.Nd2 and in 7,912 through

3.Nc3.

Against

5.Ng5, White won 58% of the 309 games.

Lichess, 1600 to 1999, September 2026, all speeds

Chapter 10 of 22

4...Nf6: Tartakower and Bronstein-Larsen variations

White 50%Draws 4%Black 46%

52,905 games

Lichess, 1600 to 1999, September 2026, all speeds

Black offers a knight trade at once.

After

5.Nxf6+ Black has to take back with a pawn and gets doubled f-pawns, choosing between ...exf6 and ...gxf6, which open different lines.

White’s plan

Trade on f6 to double Black's pawns, develop with c3 and

Bd3, and keep the better structure for the long run.

Black’s plan

After ...exf6, develop fast with ...Bd6 and castle, and use your open lines before the structure starts to matter.

4...Nf6 is played at all levels but a little more often by club players.

5...exf6, the Tartakower or Korchnoi Variation, gives Black quick development, at the price of White's better pawn structure and long-term chances.
In an endgame, White's d-pawn can become a passed pawn.

5...gxf6, the Bronstein-Larsen Variation, is more ambitious: Black gets the open g-file for a rook, but takes on a worse pawn structure and usually has to castle queenside.

It is generally considered somewhat unsound.
Stockfish 16 at depth 30 agrees on the order: it rates 5...exf6 at +0.39 and 5...gxf6 at +0.91 for White.

At 1600 to 1999 on Lichess in September 2026,

4...Nf6 came up in 52,905 games through

3.Nc3, and after

5.Nxf6+ Black took back with the e-pawn in 25,586 and the g-pawn in 1,692.

Lichess, 1600 to 1999, September 2026, all speeds

Alien Gambit

White 54%Draws 3%Black 44%

4,365 games

Lichess, 1600 to 1999, September 2026, all speeds

The Alien Gambit: take the knight with

6...Kxf7.

Stockfish 16 at depth 30 rates that clearly better for Black (-1.84).

Tartakower Variation

White 49%Draws 5%Black 46%

25,586 games

Lichess, 1600 to 1999, September 2026, all speeds

The Tartakower: Black takes back with the e-pawn, accepts doubled f-pawns and gets fast development.

White keeps the better structure for the long run.

Bronstein-Larsen Variation

White 52%Draws 5%Black 43%

1,692 games

Lichess, 1600 to 1999, September 2026, all speeds

The Bronstein-Larsen: the g-file opens for Black's rook, but the structure is worse and Black usually castles queenside.

It is generally considered somewhat unsound.

Chapter 11 of 22

Other moves after 3.Nc3

White 51%Draws 4%Black 45%

137,208 games

Lichess, 1600 to 1999, September 2026, all speeds

Black does not have to take on e4.

3...g6, the Gurgenidze, is the only common alternative;

3...Nf6 and

3...b5 are rare surprises.

White’s plan

Against

3...g6 or

3...Nf6, the push e5 gains space and closes the center on your terms.

Black’s plan

Choose

3...g6 if you want a fianchetto against d4, and expect e5 or h3 in reply; against the gambits, the engine's first choices are 4...exf3 against 4.f3 and 4...b5 against 4.Bc4.

3...g6 prepares ...Bg7 to put pressure on White's d4 pawn.

White can expand on the kingside with 4.e5 Bg7 5.f4 h5, or play 4.Nf3 Bg7 5.h3 to stop the

...Bg4 pin.

3...Nf6, the Campomanes, invites

4.e5, which drives the knight to d7 and gives a closed center like the Advance Variation.

3...b5, the Gurgenidze Counterattack, threatens ...b4 to kick the c3 knight.
It is unsound but performs well enough at club level, and the September 2026 Lichess games agree: Black won 52% of 276 games at 1600 to 1999, though Stockfish 16 at depth 30 rates White clearly better (+1.03).

After

3...dxe4 White can also gambit the pawn with

4.f3, the Rasa-Studier, or 4.Bc4, the von Hennig.

At depth 30 both favor Black slightly (-0.64 and -0.49), but at 1600 to 1999 White won 56% and 59% of those games.

Chapter 12 of 22

Other moves after 3.Nd2

White 51%Draws 4%Black 45%

28,870 games

Lichess, 1600 to 1999, September 2026, all speeds

3.Nd2 defends e4 like

3.Nc3 but keeps the c-pawn free to go to c3.

Its main point is against

3...g6.

White’s plan

Meet

...g6 and

...Qb6 alike with c3.

Black’s plan

Take on e4 and play the 4.Nxe4 lines, or try

3...g6 knowing White has c3 ready.

White plays

3.Nd2 to take back on e4 with the knight while keeping c3 for a pawn.

Against 3...g6 that matters: after 4.c3 the bishop on g7 has little to do against White's dark-squared pawn chain, which is why 3.Nd2 grew more popular in the 1970s.
Black's rare try is 3...Qb6, the Edinburgh Variation, which attacks the d4 pawn; 4.c3 defends it.

At 1600 to 1999 on Lichess in September 2026,

3.Nd2 came up in 28,870 games, and 23,477 of them continued

3...dxe4.

Lichess, 1600 to 1999, September 2026, all speeds

Chapter 13 of 22

Fantasy Variation

White 54%Draws 3%Black 43%

28,869 games

Lichess, 1600 to 1999, September 2026, all speeds

White defends e4 with the f-pawn and hopes for a full pawn center and a half-open f-file if Black takes on e4.

Black's solid answer is

3...e6, preparing

...c5 against the dark squares.

White’s plan

Build the e4 and d4 center with f3, and if Black takes, recapture with the f-pawn and use the half-open f-file.

Black’s plan

Play

3...e6 and

...c5 to hit the dark squares, or take on e4 only if you know the 4...e5 lines.

3.f3 supports e4 and offers to recapture there with the f-pawn, keeping two pawns in the center.

It also moves a kingside pawn, which changes the cover around White's king.
Black does not have to take: 3...e6 is probably the most solid answer, and 3...g6 and 3...Qb6 are the other tries.

If Black takes,

3...dxe4

4.fxe4

e5 is the main move.

5.Nf3 exd4 6.Bc4 is the Maróczy Gambit, a sacrifice first played in 1899, and Stockfish 16 at depth 30 rates it at +0.67 for White.
In the other main line, 5...Bg4 6.Bc4, Black must play 6...Nd7 to keep e5 protected.

At 1600 to 1999 on Lichess in September 2026,

3.f3 came up in 28,869 games and White won 54% against Black's 43%.

Black took on e4 in 16,166 of them, and after 6.Bc4 in the Maróczy Gambit White won 60% of 2,017 games.

Lichess, 1600 to 1999, September 2026, all speeds

Maróczy Gambit

White 60%Draws 3%Black 38%

2,017 games

Lichess, 1600 to 1999, September 2026, all speeds

The Maróczy Gambit: White gives the d-pawn for quick development and pressure on f7.

Stockfish 16 at depth 30 rates it at +0.67 for White.

Chapter 14 of 22

Other third moves after 1.e4 c6 2.d4 d5

Each of these is a gambit or an unusual way to hold e4.

Learn the full idea of each, because the third move alone does not show it.

Black’s plan

Against 3.Bd3, take on e4 and develop with ...Nf6; against 3.Be3, taking on e4 is the engine's first choice; against the Landau, avoid 3...Nf6.

3.Be3 is the Mieses Gambit: the usual idea is 3...dxe4 4.f3 exf3 5.Nxf3, a pawn for a lead in development.

Stockfish 16 at depth 30 rates the position after 3.Be3 as slightly better for Black (-0.62).

3.Bd3 defends e4 with the bishop, but 3...dxe4 4.Bxe4 Nf6 gives Black the initiative.

If Black plays 3...Nf6 instead, 4.e5 Nfd7 5.e6 is the Landau Gambit, which depth 30 rates at +1.95 for White, with 5...Nf6 6.exf7+ Kxf7 as the engine's line.

3.Nf3 dxe4 4.Ng5 is the Ulysses Gambit; after the engine's 4...Bf5, depth 30 rates it slightly better for Black (-0.93).

All three are rare: at 1600 to 1999 on Lichess in September 2026, 3.Bd3 came up in 7,905 games, 3.Nf3 in 1,943 and 3.Be3 in 480.

Chapter 15 of 22

Black avoids 2...d5

White 49%Draws 4%Black 47%

713,668 games

Lichess, 1600 to 1999, September 2026, all speeds

Black delays

...d5 and lets White keep both center pawns.

These moves are legal but rare, and the engine prefers White after each of them.

White’s plan

Keep the full center with e4 and d4, develop normally, and push e5 when it gains time on a knight on f6.

Black’s plan

Prefer

2...d5; if you choose a sideline, know its piece route, such as

...Na6 to c7 or

...Nf6 to d5 and c7.

After

2.d4, Black plays

2...d5 in more than 95 games in 100.

2...Na6 is the De Bruycker Defense, with the knight heading for c7 by way of a6.

2...Nf6, the Masi Variation, invites 3.e5, after which the knight goes to d5 and, if 4.c4 kicks it again, to c7.
2...f5 is the Massachusetts Defense.

Stockfish 16 at depth 30 rates White slightly better after

2...Na6 (+0.98) and

2...Nf6 (+0.99), and clearly better after

2...f5 (+2.17).

A good score for White is not a forced loss for Black, but none of these moves gives Black the normal ...d5 fight for the center.
At 1600 to 1999 on Lichess in September 2026 they almost never came up: 866 games for 2...Nf6, 116 for 2...Na6 and 64 for 2...f5, out of 713,668 after 2.d4.

Chapter 16 of 22

White avoids 2.d4

White holds the d-pawn back.

Some of these lines reach the main lines later; others are sidelines with ideas of their own.

Black’s plan

Answer with 2...d5 and look for the main-line structure you know, since many of these transpose.

At club level White plays something other than 2.d4 more often than not.

In September 2026 Lichess games between players rated 1600 to 1999, 2.d4 came in 713,668 of 1,635,111 games after 1.e4 c6, about 44%.
2.Nf3 followed with 449,211, then 2.Nc3 with 150,559, 2.Bc4 with 83,821, 2.d3 with 46,477 and 2.c4 with 40,258.

Lichess, 1600 to 1999, September 2026, all speeds

Several of these meet the main lines again: 2.Nf3 d5 3.exd5 cxd5 4.d4 is an Exchange Variation, and 2.c4 can become a Panov Attack.

The others ask Black a different question before d4 is played.
Against almost all of them, 2...d5 is still the move.

Chapter 17 of 22

Two Knights Attack

White 48%Draws 4%Black 48%

150,559 games

Lichess, 1600 to 1999, September 2026, all speeds

White develops both knights before deciding where the d-pawn goes.

Black's logical answer is ...Bg4, pinning the f3 knight.

White’s plan

Develop fast, keep the d-pawn flexible, and if Black copies the Classical with ...dxe4 and ...Bf5, look for Ne5 and Qh5 against f7.

Black’s plan

Play 3...Bg4 and give up the bishop for the f3 knight after 4.h3, then build with

...e6 and

...Nf6.

2.Nc3

d5

3.Nf3, or

2.Nf3

d5

3.Nc3, is the Two Knights Attack.

White wants quick development and keeps the choice of d3 or d4 for later.
Bobby Fischer played it in his youth.

Black's logical reply is 3...Bg4.

After 4.h3 Black usually takes, 4...Bxf3 5.Qxf3, and then chooses 5...Nf6 or 5...e6.
Retreating with 4...Bh5 is playable, but Black has to be careful.
White can also play 3.Qf3 instead of 3.Nf3, the Goldman Variation, which Graham Burgess calls sound.

At 1600 to 1999 on Lichess in September 2026,

2.Nc3

d5

3.Nf3 came up in 32,869 games.

Black took on e4 in 20,387 of them and played

3...Bg4 in 5,276, and after 3...Bg4 Black won 49% and White 47%.

Lichess, 1600 to 1999, September 2026, all speeds

Mindeno Variation

White 47%Draws 5%Black 49%

5,276 games

Lichess, 1600 to 1999, September 2026, all speeds

3...Bg4 pins the f3 knight, the logical reply to the Two Knights.

After

4.h3, taking on f3 is the usual choice;

4...Bh5 is playable but needs care.

Chapter 18 of 22

2.Nf3: move orders and the Apocalypse Attack

White 47%Draws 4%Black 48%

449,211 games

Lichess, 1600 to 1999, September 2026, all speeds

2.Nf3 is usually a move-order difference: after

2...d5 it often becomes an Exchange Variation or a Two Knights Attack.

Two lines are its own:

3.exd5

cxd5

4.Ne5 and 3.d3.

White’s plan

Use

2.Nf3 to reach the Exchange Variation with d4 later, or try 4.Ne5 after the trade on d5 to set early traps.

Black’s plan

Answer

2...d5, and after

3.exd5

cxd5

4.Ne5 play

4...e6 or 4...g6 rather than an early ...Bf5.

After

2...d5,

3.exd5

cxd5

4.d4

Nc6 is the Exchange Variation, and 3.Nc3 is the Two Knights Attack.

3.d3 heads for the Breyer Variation and offers a queenless middlegame.
4.Ne5, the Apocalypse Attack, sets traps for Black; 4...e6 is the most common answer and a perfectly good one, and 4...g6 is the critical one.

2.Nf3 was White's second most common move at 1600 to 1999 on Lichess in September 2026, with 449,211 games, and 287,325 of them went

2...d5

3.exd5.

After 3.d3 dxe4 4.Ng5, the Dinic Gambit, White won 63% of 6,800 games, although depth 30 rates the position at -0.43, slightly better for Black.

Lichess, 1600 to 1999, September 2026, all speeds

Apocalypse Attack

White 53%Draws 4%Black 43%

2,251 games

Lichess, 1600 to 1999, September 2026, all speeds

The Apocalypse Attack sets traps:

4...Bf5 runs into

5.Bb5+.

4...e6 is the most common answer and

4...g6 the critical one.

Chapter 19 of 22

Accelerated Panov Attack

White 51%Draws 4%Black 45%

40,258 games

Lichess, 1600 to 1999, September 2026, all speeds

White plays c4 before d4, so d5 is hit twice once the pawns meet.

The usual result is a trade of two pawn pairs and an isolated d-pawn for White.

White’s plan

Develop with Nc3 and Nf3 before d4, and use the open c- and e-files while the queens are still on.

Black’s plan

Take back on d5 with a pawn or a piece, never early with the queen, and aim for the endgame, where the isolated pawn is a target.

With pawns on c4 and e4, White fights for d5 at once.

The main line goes 2...d5 3.cxd5 cxd5 4.exd5 Nf6 5.Nc3 Nxd5, and White's d-pawn is left without neighbours.
That pawn is a weakness in an endgame, so White has to use the open c- and e-files in the middlegame.
Because the d-pawn stays on d2 until White castles, the c3 knight cannot be pinned.

The line can also become a normal Panov Attack with 3.exd5 cxd5 4.d4.

2...e5, the Open Variation, is a different game: it looks like

1.e4

e5, but after 3.Nf3 Black cannot defend e5 with ...Nc6 because the c-pawn already stands on c6.

At 1600 to 1999 on Lichess in September 2026,

2.c4 came up in 40,258 games, and White won 51% against Black's 45%.

After

2...d5, White chose

3.cxd5 in 25,290 games and 3.exd5 in 8,145.

Lichess, 1600 to 1999, September 2026, all speeds

Pseudo-Scandinavian

White 100%

2 games

Lichess, 1600 to 1999, September 2026, all speeds

3...Qxd5 is a blunder:

4.cxd5 takes the queen.

Take back with the pawn instead.

Chapter 20 of 22

Breyer Variation

White 49%Draws 4%Black 47%

46,477 games

Lichess, 1600 to 1999, September 2026, all speeds

White supports e4 with the d-pawn and lets Black take on e4 later.

The game becomes either an early queen trade or a King's Indian Attack setup.

White’s plan

Choose early: 3.Nf3 for a quiet queenless game, or 3.Nd2 with g3, Bg2 and castling for a full middlegame.

Black’s plan

Play

2...d5, and against

3.Nd2 keep the tension with

3...e5.

2.d3 prepares to take back on e4 with the d-pawn.

After 2...d5 3.Nf3 dxe4 4.dxe4 Qxd1+ 5.Kxd1 the queens are off by move five.
To keep the queens on, White plays 3.Nd2, so that ...dxe4 dxe4 no longer opens the d-file between the queens, and Black's main reply is to keep the tension with 3...e5.

From there White often sets up g3, Bg2 and castles, the King's Indian Attack.

After 3.Nf3 dxe4, White can also gambit the pawn with 4.Ng5.
At 1600 to 1999 on Lichess in September 2026, 2.d3 came up in 46,477 games, and White won 49% against Black's 47%.

Chapter 21 of 22

Hillbilly Attack

White 46%Draws 4%Black 50%

83,821 games

Lichess, 1600 to 1999, September 2026, all speeds

The bishop goes to c4, where

...d5 attacks it at once, so White loses time.

The real point is the gambit

3.Bb3

dxe4

4.Qh5, aimed at f7.

White’s plan

Play 3.Bb3 and, if Black takes on e4, 4.Qh5 at f7 for sharp play.

Black’s plan

Play

2...d5; after

3.Bb3

dxe4

4.Qh5, cover f7 with

4...g6 or 4...e6, and you keep the extra pawn.

2.Bc4 has a poor reputation: after

2...d5

3.exd5

cxd5 the bishop is attacked again and has to move, losing a tempo.

The d5 pawn is defended by both the c-pawn and the queen, so White cannot hold the bishop on c4.

It became a surprise weapon in online blitz because of 3.Bb3.

If Black takes with 3...dxe4, 4.Qh5 hits f7, which only the king defends, and Black usually blocks with 4...g6 or 4...e6.

Stockfish 16 at depth 30 rates

2.Bc4 at -0.38, slightly better for Black, and the position after 4.Qh5 at -0.82.

At 1600 to 1999 on Lichess in September 2026, Black won 50% of the 83,821 games after 2.Bc4 and White 46%.
But in the 978 games that reached 4.Qh5, White won 58%.

Schaeffer Gambit

White 58%Draws 4%Black 38%

978 games

Lichess, 1600 to 1999, September 2026, all speeds

The Schaeffer Gambit:

4.Qh5 and the b3 bishop both aim at f7.

Block with

4...g6 or

4...e6 and keep the pawn.

Chapter 22 of 22

White's wing-pawn moves against the Caro-Kann

Flank pawn moves that leave the center to Black.

2...d5 is a sound answer to all three.

Black’s plan

Take the center with 2...d5 and develop normally.

These moves are generally seen as ineffective or doubtful.

2.b3, the Euwe Attack, prepares to fianchetto the bishop and gives the center to Black in a hypermodern style.
2.b4 is the Labahn Attack, and 2.g4 the Spike Variation.
The sharpest is the Labahn Double Gambit, 2.b4 d5 3.b5, which Stockfish 16 at depth 30 rates clearly better for Black (-1.69) after 3...dxe4.

All three are rare at club level: at 1600 to 1999 on Lichess in September 2026, 2.b3 came up in 4,456 games, 2.b4 in 287 and 2.g4 in 122.

Lichess, 1600 to 1999, September 2026, all speeds

White’s move	Variation	What changes	Open chapter
3.e5	Advance Variation	White closes the center and takes space. Black develops the bishop to f5 before ...e6, then attacks d4 with ...c5.	Chapter 3 ›
3.exd5 cxd5	Exchange Variation	The center becomes symmetrical. Black puts the knights on c6 and f6 and the bishop on g4 before ...e6.	Chapter 5 ›
then 4.c4	Panov-Botvinnik Attack	White hits d5 again and usually ends up with an isolated d-pawn, paid for with quick development and kingside attacking chances.	Chapter 6 ›
3.Nc3 or 3.Nd2	After 3...dxe4 4.Nxe4	Black gives up the center pawn and picks a plan: 4...Bf5, the Classical; 4...Nd7, the Karpov; or 4...Nf6.	Chapter 7 ›
3.f3	Fantasy Variation	White defends e4 with a pawn and hopes for a full pawn center. Black's solid answer is 3...e6, preparing ...c5.	Chapter 13 ›

White’s move	Variation	What changes	Open chapter
3.e5	Advance Variation	White closes the center and takes space. Black develops the bishop to f5 before ...e6, then attacks d4 with ...c5.	Chapter 3 ›
3.exd5 cxd5	Exchange Variation	The center becomes symmetrical. Black puts the knights on c6 and f6 and the bishop on g4 before ...e6.	Chapter 5 ›
then 4.c4	Panov-Botvinnik Attack	White hits d5 again and usually ends up with an isolated d-pawn, paid for with quick development and kingside attacking chances.	Chapter 6 ›
3.Nc3 or 3.Nd2	After 3...dxe4 4.Nxe4	Black gives up the center pawn and picks a plan: 4...Bf5, the Classical; 4...Nd7, the Karpov; or 4...Nf6.	Chapter 7 ›
3.f3	Fantasy Variation	White defends e4 with a pawn and hopes for a full pawn center. Black's solid answer is 3...e6, preparing ...c5.	Chapter 13 ›

Choose the Caro-Kann variation to study

How the Caro-Kann works

White’s plan

Black’s plan

Caro-Kann starting moves: 1.e4 c6 2.d4 d5

White’s plan

Black’s plan

Advance Variation

White’s plan

Black’s plan

White’s plan

Black’s plan

Advance Variation: 3...Bf5

White’s plan

Black’s plan

Black’s plan

Exchange Variation

White’s plan

Black’s plan

Panov-Botvinnik Attack

White’s plan

Black’s plan

After 4.Nxe4: Black's main choices

White’s plan

Black’s plan

Classical Variation

White’s plan

Black’s plan

Karpov Variation

White’s plan

Black’s plan

4...Nf6: Tartakower and Bronstein-Larsen variations

White’s plan

Black’s plan

Other moves after 3.Nc3

White’s plan

Black’s plan

Other moves after 3.Nd2

White’s plan

Black’s plan

Fantasy Variation

White’s plan

Black’s plan

Other third moves after 1.e4 c6 2.d4 d5

Black’s plan

Black avoids 2...d5

White’s plan

Black’s plan

White avoids 2.d4

Black’s plan

Two Knights Attack

White’s plan

Black’s plan

2.Nf3: move orders and the Apocalypse Attack

White’s plan

Black’s plan

Accelerated Panov Attack

White’s plan

Black’s plan

Breyer Variation

White’s plan

Black’s plan

Hillbilly Attack

White’s plan

Black’s plan

White's wing-pawn moves against the Caro-Kann

Black’s plan

White's choices

Questions players ask

How do you play the Caro-Kann?

Where does the name Caro-Kann come from, and what is it known for?

Is the Caro-Kann good for beginners?

How do you beat the Caro-Kann as White?

Should I play the Caro-Kann or the French?

Caro-Kann or Sicilian: which should I learn?

Can you play the Caro-Kann against 1.d4?

What are the main traps in the Caro-Kann?

In the Advance, should Black's king knight go to f6 or e7?

Why play ...Bf5 before ...e6?

What does ...Qb6 do in the Caro-Kann?