4. Model Game - Balint vs Horvath
By developing both knights, White gives Black the chance to commit one of three mistakes. In the main line, it is not possible to go wrong so early. The first hope is that Black pushes the central pawns.
White is better developed and can easily deploy the light-squared bishop to c4 (or b5). Later, he can consider nudging the black centre with c3 or getting in f4 after some preparatory moves.
This line might occur against opponents who think “aah, don’t care about move orders, I take on e4 and play Bf5 as always”. That careless approach leads to an unexpectedly quick disaster: 5.Ng3 Bg6 6.h4 h6 7.Ne5 Bh7 8.Qh5! g6 9.Qf3 Nf6 10.Qb3 game over.
This was once a respected variation, but nowadays it is rightly avoided even by top players. The reason: the surprising 9. fxg3! in the main line.
Thanks to the half-open f-file, White builds up a strong attacking initiative against the black king.
This leaves Black with only two serious options, but even here, he must know what he’s doing to reach playable positions. The two viable choices are 3...dxe4 4.Nxe4 Nf6 and 3...Bg4.
The first scenario recalls the popularity of 4…Nf6 in the Classical Caro-Kann. If White captures on f6 in the Two Knights variation, it poses no real problems for Black (compare to the 4… Nf6 line of the Classical Variation). But the move 5.Qe2 has proven to be promising. After 5...Nxe4 6.Qxe4, we reach the first starting position.
White is clearly better developed and can threaten mate with Bc4 and Ne5. Black, therefore, tries to exchange queens, leading to a cat-and-mouse game between them. A tip for White: allow the exchange on e3 and recapture with fxe3. Combined with the option of b3 followed by Bb2, this yields a dream position for White, who can exert pressure without any risk.
White’s wishes came true in Mamedov–Paichadze 2014. Here, 15.g4 would have been logical, maintaining a slight advantage, but 15.d4 was also good enough.
The queens’ cat-and-mouse game tends to favour White. Therefore, it is advisable first to develop or exchange minor pieces and only later relieve the position by trading queens.
This is where the queen trade makes sense: the bishop on c4 is biting on granite, the dark-squared bishop cannot reach b2, and without queens, Black will have no trouble surviving the development deficit.
Anyone with a taste for offbeat moves might follow a trend successfully started by GM Eljanov, namely, playing 7...Be6. Comment GM Arjun: “This is a strange move”.
On closer inspection, this move proves surprisingly clever. It prevents White’s Bc4 and prepares the knight manoeuvre Nb8–d7–f6. Occasionally, the bishop on e6 can move to d5, threatening the knight on f3.
The main line against the Two Knights runs 1.e4 c6 2.Nf3 d5 3.Nc3 Bg4 4.h3 Bxf3 5.Qxf3 e6.
White currently holds a slight lead in development, but how should the game proceed? The bishops have no good squares, the knight on c3 is restricted by the d5 pawn, while Black can continue developing smoothly.
Black should not be content with simply barricading on the light squares. An important idea is to play d4 at the right moment. Ideally, this leads to poor coordination among White’s pieces, making it less significant that Black is left with only the “bad” bishop. The closed position often turns out to be favourable for Black’s horses.
White had the idea of playing d4 himself with 8.Rd1. In this position, 8...d4 from Black is particularly strong, as 9.Nb1 can be met by 9...Ne5 10.Qg3 d3!
Another interesting idea is to build pressure on the e4 pawn using Nf6 and Bb4.
White has several options, but none promise an advantage. If he closes the centre with e5, Black obtains a solid French structure without the bad bishop. If he allows captures on c3 and e4, the bishops compensate for the weak pawn structure, resulting in an unclear position.
2.1) Caro-Kann Structure
This pawn structure is the alpha and omega for all Caro-Kann players. Black may have slightly less space, but enjoys a super-solid formation that offers no targets. For White, it is not entirely clear how to make progress. Overly aggressive play on the queenside or kingside may fire back and weaken their own structure. In any case, Black should aim to build pressure against the d4 pawn while keeping options like ...c5 or ...e5 in mind. Usually, Black can also handle White retaining the bishop pair, but it becomes essential to prevent the d4–d5 breakthrough.
This typical Caro–Kann position comes from Gavrilescu–Moranda, Olympiad 2022. White appears to have deployed his forces ideally, but Black knew precisely what to do: to exert noticeable pressure against d4. Now came 18... Rfe8 (opposing the rook to the enemy queen), and after 19.Nxe5 (a bit premature) Black struck with 19... c5!, achieving an effortless equality.
More complex pawn structures can arise in the line 3... Bg4 4.h3 Bxf3 5.Qxf3.
2.2) The Advance …d4
Against certain move orders, Black’s ...d4!? can be an interesting idea.
In Gavrilescu–Niemann 2024, play continued 7.Nb1 h5, with Black aiming to claim as much space as possible. The closed nature of the position prevents White’s bishop pair from realising its full potential.
Here’s another similar example:
In the model game Balint, P – Horvath, D, 2025, Black expanded on both flanks.
2.3) French Structure
One thing White should be careful about is not closing the centre too early with e5. Otherwise, Black gets a French-style structure without the usual problem piece, the light-squared bishop.
In this game (Manish – Horvath, D. 2023), the White player didn’t follow my advice. On his way to the GM title, the young Austrian Dominik Horvath played a fine game with Black. White’s centre collapses, and the knight on e2 blocks his own bishop.
2.4) Petroff/Berlin Structure
In one of the currently critical lines, the pawn structure resembles that of the Petroff Defence (or a Berlin Defence with Re1, where White parts with the light-squared bishop.)
In this symmetrical pawn structure, the bishop pair plays only a minor role anyway, especially since the bishop on d1 still needs to be developed. A concrete sequence of moves is:14.c3 Ng6, and after 15.Bc2, 15...Qf6! forces White to exchange queens; otherwise, Black can play Qd6, threatening to harass the white king.
1) 3…Nf6 4.e5 Ne4 5.Ne2 Qb6 6.d4 e6 7.Ng3 c5 8.Bd3! Nxg3 9.fxg3! cxd4 10.0-0 Nc6 11.a3 Bd7 12.b4 a6 13.Qe2
White is down a pawn, but calmly continues to build up his attack.
Anyone who wants to play this line with Black should improve on the game Bacrot - Iturrizaga 2019. A comprehensive coverage from White’s perspective is provided by GM Arjun: Expert Repertoire against the Caro-Kann.
2) 3…dxe4 4.Nxe4 Nf6 5.Qe2 Nxe4 6.Qxe4
As mentioned above, I’m not a fan of taking the queen out for a walk too early, whether to d5 or a5. Instead, I would bring a minor piece into play. Let’s start with the bishop:
2.1.) 6…Be6 Eljanov’s pet move.
White now has a wide range of moves. By far the most popular is 7.b3, but 7.c4 and 7.Be2 are also interesting options for White.
2.1.1) 7.b3 with two viable options for Black:
2.1.1.a) 7…Bd5 8.Qe3 Bxf3 9.Qxf3 Qd5 10.Qe3 Nd7 11.Bb2 e6 12.0-0-0 Nf6 13.Bc4 Qa5
Black stands solid. He will castle queenside and use Ba3 to halve White’s bishop pair.
2.1.1.b) 7…Nd7 8.Bb2 Nf6 9.Qe3 (9.Qb4 is the engine move. 9…Qc7 10.Bd3 Nd5 11.Qe4 Qf4 and Black reaches the desire endgame) 9…Nd5!? (9…Bf5 allows White to play 10.Nd4, followed by f4) and after almost any reasonable queen move, 10…Bf5 equalizes.
The queen retreated to e2. After Bf5, Black stands comfortably.
White can restrict the bishop on e6 with 7.c4 and chase it down with Ng5:
2.1.2) 7.c4 Nd7 (8.d4 Nf6 9.Qh4 Bf5 10.Be2 e6 11.0-0 Be7 Black has reached a nice position) 8.Ng5 Nf6 9.Nxe6 Nxe4 10.Nxd8 Rxd8 11.d3 Nc5 12.Be3 e5 13.0-0-0 Ne6 with a pleasant ending for Black.
The bishop pair is practically useless here; the weakness on d4 is far more significant.
GM Arjun recommends a move that looks modest but conceals plenty of bite: 7.Be2. At first glance, White is developing quietly, but the idea contains hidden venom.
2.1.3) 7.Be2 Nd7 8.d4 Nf6 and now the key move is 9.Qe5!
Black can now play 9…Qd6 to trade off the queens. White simply castles. Because of the slightly awkward bishop on e6, the endgame after a queen exchange promises White a slight but lasting edge.
9…g6 fails to 10.g4! Bd7 11.Bc4 with a strong attack.
So Black’s best bet is 9…Bg4. After 10.h3, the bishop can take on f3 or drop back to h5. In both cases, the position should be assessed as somewhere between equal and slightly better for White.
Concerning the 6…Be6-line: The case for White is pleaded by GM Arjun. Expert Repertoire against the Caro-Kann. The dark side is defended by GM Grigorov: Play the Caro-Kann, Part 2.
Let’s go back to Black’s move number 6.
2.2) 6…Nd7 is a logical development move.
2.2.1) 7.Bc4 is the big main line. 7…Nf6 8.Ne5 e6 9.Qe2
The critical line runs 9…b5 10.Bd3 Qc7 11.a4! Bd6 12.Nxc6! b4! followed by another series of computer moves, a chaotic position, and the engine’s reliable evaluation of 0.00. Anyone blessed with a good memory who enjoys following the suggestions of the silicon monster is warmly invited to study this line. Detailed instructions can be found here: GM Pavel Eljanov, My Caro-Kann - Part 3.
The Caro-kann gone wild: This is not the end of the complications, only the beginning.
Those who favour a classical Caro-Kann approach may want to follow Dreev’s recommendation of 9…Qd4. After 10.c3 Qe4 Black swaps queens and drives away the cheeky knight on e5 with ...f6.
This is the position GM Dreev aims for. With ...Ba6 and ...Kf7, White’s slight pressure is neutralised. More detailed explanations can be found here: Caro-Kann According to Dreev, Part 2.
White has meanwhile recognised that the quick mating threat with Ne5 and Bc4 offers no theoretical advantage. A sensible alternative is the calmer 7.d4.
2.2.2.) 7.d4 Nf6 8.Dd3!? Practice has shown that this is the best square for the queen. 8…Bg4!? (8…g6 9.Be2 Bg7 10.0-0 0-0 11.h3 Bf5 12.Qb3 is more pleasant for White) 9.Ne5 Bh4 10.c3 e6 11.Qh3! Bd6!. Black reached an equal endgame in Costa–Donchenko 2023.
Let’s turn to Black’s key reply in the Caro-Kann Two Knights variation:
3) 3…Bg4 4.h3 Bxf3 5.Qxf3 e6
We’ll take a look at three main continuations for White.
3.1. 6.d4 Nf6 7.Bd3 (7.e5 Nd7 is promising for Black) dxe4 8.Nxe4 Qxd4! 9.Be3 Bb4+! 10.Ke2 Qd8.
White has just enough compensation for the missing pawn, but no advantage is in sight.
3.2. 6.g3 Nf6 (6…d4!? was the recent over-the-board choice from Niemann and D. Horvath, also employed by Keymer in online blitz) 7.Bg2 Bb4!? 8.0-0 0-0 9.d3 Nbd7 10.Qe2 Bxc3 11.bxc3 dxe4 12.dxe4 Qa5
An exciting battle is about to unfold: White has the bishop pair, but the pawns are fractured.
3.3. 6.Be2 is the main line. 6…Bc5 prevents d4. 7.0-0 (7.Qg3 is answered with 7…Bd4) Nd7 (7…Ne7 8.Sa4! offers a slight edge to White) 8.Qg3 (8.Rd1 allows 8…d4 9.Nb1 Ne5!, followed by 10…d3!) 8…Bd4! 9.exd5 exd5 10.Ld1!? Ne7 11.Ne2 Be5 12.Qf3
Now Black can play 12...d4!? to hinder White’s development and raise the question of whether the pieces on c1, d1, and e2 are ideally placed. 12...O-O 13.d4 Bc7 14.c3 Ng6, along with Qf6, is, of course, also perfectly reasonable.
You can find a complete analysis here: Jobava’s Creative Caro-Kann.
The Two Knights System (1.e4 c6 2.Nc3 d5 3.Nf3) gives Black several ways to go wrong early (mistakes to avoid are pointed out in 1. Basic Concepts). Black has two viable ways to meet the Two-Knights Variation:
5.Qe2 (5.Nxf6+ exf6 transposes to a line of the Tartakower Variation of the Classical System) Nxe4 6.Qxe4, and now either 6…Be6, or 6...Nd7. Black has less space, while his position has no weaknesses, and he has already exchanged a pair of knights. Once Black manages to complete his development without making concessions, he can forward to an equal middlegame.
Black has parted with his light-squared bishop to obtain a solid position with easy development. Typical ideas include playing ...d4 at the right moment to disrupt White's coordination, or building pressure on e4 with ...Nf6 and ...Bb4.
Key Strategic Themes