Queen's Gambit Declined: Marshall Defense

The Marshall Defense (1.d4 d5 2.c4 Nf6) is a rare and fairly dubious way of declining the Queen's Gambit. Black develops the knight at once instead of supporting d5 with ...e6 or ...c6, and invites White to take on d5.

Origins

The defence is named after Frank Marshall, who played it in the 1920s. He gave it up after losing with it to Alexander Alekhine at Baden-Baden in 1925, and it is no longer used by experienced players.

The idea

Black hopes to recapture on d5 with the knight and develop quickly. The problem is that after 3.cxd5 Nxd5 White can build a big centre with e4, or first control e5 with Nf3 and then play e4.

Main lines

• 3.cxd5 Nxd5:
  • 4.e4, a common continuation, though White may be playing e4 too early; after 4...Nf6 White can play 5.e5, 5.f3 or 5.Nc3;
  • 4.Nf3!, controlling e5 and preparing e4 without allowing ...e5. After 4...Bf5 White gets a large advantage with 5.Qb3.
• 3.cxd5 c6 4.dxc6 Nxc6, the Tan Gambit: Black usually follows with ...e5 and an exchange of queens, but White keeps a small advantage in the queenless middlegame.
• 3.cxd5 g6, a tricky try to transpose to the Grünfeld Defense if White plays 4.Nc3; after 4.Qa4+ Black cannot regain the pawn and lacks compensation.
• 3.Nc3 ignores the provocation and usually transposes to the Queen's Gambit Declined (3...e6), the Slav (3...c6), the Queen's Gambit Accepted (3...dxc4) or the Grünfeld (3...g6).

Strengths and weaknesses

Strengths

• After 3.Nc3 Black can transpose to sound main-line defences.

Weaknesses

• In the main lines after 3.cxd5 White keeps an advantage, large after 3...Nxd5 4.Nf3 Bf5 5.Qb3.
• Its own inventor abandoned it, and experienced players no longer use it.

Who should play it

Because it is considered dubious, the Marshall Defense is at most an occasional surprise for Black. White players should know 3.cxd5 Nxd5 4.Nf3, the most testing answer.

The numbers

On Lichess White does well at every level: 52.7% against 43.5% under 1400, 53.3% against 42.6% at 1400–1799, 53.7% against 41.4% at 1800–2199 and 54.5% against 39.2% from 2200 up. There are only a few hundred master games, too few to draw firm conclusions.