Four Knights Game

Strategic Summary & Main Idea

The Four Knights Game develops all four minor knights before taking any decision, which is why it spent a century with a reputation as a drawing opening. The position after 3...Nf6 measures +0.14, and the main line, the Spanish Four Knights, measures +0.19 — small, but real. What this entry adds is a measurement of the aggressive versions that circulate as shortcuts: the Halloween Gambit, which sacrifices a knight on move four, measures −1.06, and the Belgrade Gambit measures −0.49. Both are entertaining and both are losing, which is more useful information than either one's reputation.

The pawn structure this opening produces

The Four Knights structure is the most symmetrical in open chess: pawns on e4 and e5, knights developed on both sides, and no tension until somebody creates it. That makes the opening deceptive, because symmetry is not a safe state — it is a tempo count. Whoever copies move for move loses at the moment the opponent breaks the symmetry with tempo, and the classical line with ...Bg4 against Bg5 is exactly that: the two pins look identical until White plays Nd5, and then one of them is worth a tempo and the other is not. In the Rubinstein Variation, with 4...Nd4, Black breaks the symmetry on purpose and accepts a structure with unpaired central pawns in exchange for activity — the line that unbalances most without conceding. In the Scotch Four Knights, the exchange on d4 and the capture on c6 restore the doubled-pawn structure familiar from the Scotch proper, with the same verdict: doubled pawns control central squares and are no weakness while the centre stays closed. And in Glek's fianchetto version White abandons the symmetry from the outside, with g3 and Bg2, aiming at d5 instead of contesting the centre head-on.

Main Variations of the Four Knights Game

Evaluations measured by us with Stockfish 19, depth 22, on 14/09/2026.

Frequently asked questions

Is the Four Knights a drawing opening?

It is a low-advantage opening, which is not the same thing. The position after 3...Nf6 measures +0.14 and the Spanish Four Knights +0.19. Symmetry is not safe in itself: it breaks at the first move that gains a tempo, and the line with ...Bg4 copying Bg5 shows the cost — +0.88 after Nd5.

Is the Halloween Gambit worth playing?

Against correct defence, no. The knight sacrifice on e5 measures −1.06, and Black can give material back whenever he likes to keep a solid position. It works in blitz, against someone who has never seen it, and the difference between those two things is what this entry measures.

Can Black capture on e4 against 4.Bc4?

Yes, and it is the best reply: the position after the exchanges measures −0.08. What does not work is White's most popular reaction to that capture — the sacrifice 5.Bxf7+, which leaves the position at −3.78 once Black's king takes and the knight returns home.

Guides on this opening

• An open games repertoire: meeting 1.e4 with 1...e5 without memorising Ruy Lopez theory
• The Four Knights Sicilian: the main line, the check that fails and the test the engine prefers
• The Berlin Defence: the wall, the endgame it forces and the anti-Berlins that give equality back

Common Mistakes & Where Most Amateurs Go Wrong

• 01 Copying the pin with ...Bg4 after Bg5. The move 9.Nd5 hits the b4 bishop and the f6 knight at once, and the symmetry breaks in favour of whoever moved first: the position rises from +0.14 to +0.88. The line: 1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.Bb5 Bb4 5.O-O O-O 6.d3 d6 7.Bg5 Bg4 8.Nd5 +0.88
• 02 Counting on the 6...Qd4 trap after Nxe5. It does not work: White retreats the knight, keeps the pawn and Black recovers nothing — the position measures +2.04 for White. The line: 1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.Bb5 a6 5.Bxc6 dxc6 6.Nxe5 Qd4 +2.04
• 03 Sacrificing on f7 against 4...Nxe4. Black's king takes, the knight comes back and White is a bishop down for two pawns and a handful of checks that run out: the position collapses to −3.78. The line: 1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.Bc4 Nxe4 5.Bxf7+ Kxf7 6.Nxe4 d5 7.Neg5+ Kg8 8.d3 −3.78