The Four Knights Sicilian: the main line, the check that fails and the test the engine prefers

Short answer

The Four Knights Sicilian comes from 1.e4 c5 2.Nf3 e6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 Nc6, and it is the line where Black swaps memorised theory for one plan: ...Bb4, the exchange on c3 and the ...d5 break. This guide measures White's five tries in Stockfish 18 at depth 24, shows why 6.Ndb5 Bb4 7.Nd6+ is a check that gives the advantage away, and records what the engine actually prefers — 6.Nxc6, which leads to a simplified middlegame and is the test the variation's reputation tends to leave out.

The essentials in a few lines

• The starting position, after 1.e4 c5 2.Nf3 e6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 Nc6, measures +0.50 in Stockfish 18 at depth 24 — the normal range for White's edge in an Open Sicilian.
• The engine's preferred test is not the famous 6.Ndb5 but 6.Nxc6 (+0.56 against +0.33). It removes the target on c6 and leads to positions with no automatic counterplay, which is the opposite of the variation's reputation.
• After 6.Nxc6 bxc6 7.e5, the reflex break 7...d5? drops a piece to 8.exf6 (+2.94). The right move is 7...Nd5 (+0.58), and the alternative 7...Ng8 costs nearly four tenths.
• In the main line 6.Ndb5 Bb4 7.a3 Bxc3+ 8.Nxc3 d5 9.exd5 exd5, only one tenth move keeps White's advantage: 10.Bd3 (+0.34). The natural 10.Be2 (−0.13), 10.Qe2+ (−0.17), 10.Bg5 (−0.20) and 10.Bd2 (−0.37) hand the initiative to Black.
• The check 7.Nd6+ takes the evaluation from +0.37 to −0.11: the best-placed knight on the board ends up traded for the c8 bishop, which was Black's worst piece.

Four knights and one plan

The logic belongs to players who want a Sicilian without keeping two or three theoretical variations current. In the main lines of the Najdorf and the Dragon, home preparation decides games before move 20, and keeping that repertoire warm costs hours every month. Here the entry price is low: five natural moves, none of them hard to remember, and a plan that fits in one sentence.

What Black gets in exchange is the chance to equalise early by trading the dark-squared bishop for a knight. After ...Bb4 the threat is ...Nxe4, and White almost always answers a3 — which gives Black exactly what they wanted: exchange on c3, break with ...d5, and reach an isolated-pawn position on d5 with active pieces and development complete.

What the variation does not deliver is what its reputation promises. It gets described as a counter-attacking weapon, and that is half true: it dodges heavy theory, but the test the engine prefers leads to a dry middlegame with no targets and no tactics. Anyone adopting the line expecting complications will only find them if the opponent cooperates — and the next section is about the case where they do not.

6.Nxc6: the test the engine prefers

The choice of recapture decides the kind of game. With 6...bxc6 (+0.59) the queens stay on, Black gets the b-file and the c6-d7 pawn mass can advance later; with 6...dxc6 (+0.61) White trades queens at once with 7.Qxd8+ Kxd8, and Black loses the right to castle in a position with no counterplay — playable, but unpleasant to defend for thirty moves. The difference in evaluation between the two is irrelevant; the practical difference is large.

The mistake worth knowing comes right after 6...bxc6 7.e5. The ...d5 break is the system's reflex — it is why Black played ...e6 and ...Nc6 in the first place — and here it drops a piece: 7...d5? 8.exf6 (+2.94), because the f6 knight is attacked and the d5 pawn does not defend it. The right move is the retreat to the centre, 7...Nd5 (+0.58); 7...Ng8 also saves the piece but costs nearly four tenths and undoes the development.

What this section implies for a repertoire is simple and worth saying plainly: against an opponent who plays 6.Nxc6, the Four Knights produces no counter-attack. It produces a slightly worse and perfectly playable position in which Black has to handle doubled pawns and a structure without targets. If your reason for adopting the variation is to avoid theory, it delivers; if it is to create complications, it does not.

The main line, and the tenth move that decides it

It is worth understanding why the exchange on c3 is not a concession. Black gives up the dark-squared bishop, but the c3 knight was the piece controlling d5, and without it the break comes for free. The isolated pawn that results is the price, and it is a known one: it gives Black space, the e4 square and open files for the rooks, at the cost of a worse endgame if every piece leaves the board.

The measurement of the tenth move is this guide's most useful figure for White. The four moves intuition suggests — 10.Be2 to develop, 10.Qe2+ to annoy the king, 10.Bg5 to pin, 10.Bd2 to finish — all travel from a white advantage to a small black one, between −0.13 and −0.37. The only one that holds +0.34 is 10.Bd3, which attacks the isolated pawn along the diagonal and prepares 0-0 and Bg5 in the right order.

For Black the reading is the same in reverse: the position after 9...exd5 is comfortable against everything except the most precise move. That makes the line excellent in club games and average against preparation — which is exactly the profile of the player who chooses the Four Knights. Two isolated-pawn ideas recur and are worth knowing: ...Bg4 pressing f3, and ...d4 at the moment a black rook owns the e-file.

The check that looks free

The temptation is understandable. The black king loses the right to castle, the knight occupies an advanced square and the move is forcing, which excuses you from calculating. It is the classic shape of the move nobody verifies — and in practice it turns up in rapid games and in the last rounds of tournaments, when checking the opponent's reply is the first thing to go.

What 7...Ke7 reveals is that the d6 knight has no job. It attacks nothing of value, it cannot be supported by pawns, and it blocks the very square White would like to press on. After 8.Nxc8+ Rxc8, Black has both rooks connected along the c-file, the king one move from f8 and the structure intact — the evaluation is −0.15.

If White insists on keeping the knight with 8.Bf4, the position is still slightly preferable for Black, and the game becomes one in which White is the side with something to prove. That is the exact inversion of what the check promised, and the reason this line enters the guide as a warning rather than a recommendation.

Move orders and transpositions

This is the move-order detail that costs the most points among players who adopt the system without understanding it. ...d6 is the most natural move in the world in a Sicilian, and here it swaps a one-plan opening for a theory-heavy one. Anyone who wants the Four Knights has to play ...Bb4 at the first opportunity, even when the square looks exposed.

There is also the reverse route: opponents who play 6.a3 precisely to stop ...Bb4. The move costs a tempo and measures +0.29; the answer is ...d6 or ...Be7 with normal development, accepting a Scheveningen Sicilian with a classical structure. It is no cause for alarm, but it does prove the variation is not a closed system — it depends on White cooperating with the move order.

For players who already play the Sicilian and only want to simplify the repertoire, this catalogue's guide on the Sicilian Defence for intermediate players covers how to pick one variation and stay in it; the Four Knights is a legitimate item on that menu, with the profile described here: little theory, few complications, and a position that requires understanding the isolated pawn.

Who the variation suits

The ideal profile is the club player with little study time who has already lost games by forgetting move 15 of a memorised line. In the Four Knights there is no move 15 to memorise: there is one plan, three positions to recognise and one trap to avoid, 7...d5 after 6.Nxc6 bxc6 7.e5. That fits in a single study session and survives months without review.

The wrong profile is the player who needs to win with Black. Against a precise opponent the variation delivers lukewarm equality in the best lines and a small, stable disadvantage in the rest — and the isolated-pawn structure, which is where Black's chances live, demands exactly the kind of middlegame understanding that takes time to build. It is not a fast-win opening.

As a study plan, the order that pays best is: the main line with ...Bb4 and ...d5 first, because it is the one you will play; then the answer to 6.Nxc6, because it is what a strong opponent plays; and the transpositions last, since they are settled by one move-order rule. Two hours cover the set, and reviewed games will show within a fortnight whether the isolated-pawn structure suits you.

• Play ...Bb4 at the first opportunity; ...d6 transposes into the Sveshnikov.
• After a3, exchange on c3 and break with ...d5 — that is the whole plan.
• Against 6.Nxc6, recapture with the b-pawn and never play ...d5 after e5.
• With the isolated pawn on d5, press f3 with ...Bg4 and save ...d4 for when the e-file is yours.
• If White plays 6.a3, accept the Scheveningen and develop normally.

Frequently asked questions

Openings mentioned in this guide

• Sicilian Defense: Four Knights Variation (B45)
• Sicilian Defense (B20)
• Sicilian Najdorf (B90)
• Sicilian Defense: Alapin Variation (B22)
• Four Knights Game (C47)
• Sicilian Defense: Sveshnikov Variation (B33)
• Sicilian Defense: Rossolimo Variation (B30)
• Grand Prix Attack (B23)