Queen's Gambit Declined: Marshall Defense (Nf3) – How to Play as Black

The Marshall Defense (2...Nf6) immediately challenges White's centre, avoiding the more passive 2...e6 lines. After 3.Nf3 e6, Black has a compact pawn chain (d5 and e6) that makes it hard for White to break through easily. Your position is solid — not flashy, but very hard to crack. The main idea is to complete development, keep the centre closed if it suits you, and look for ...c5 breaks once your pieces are ready. Because the evaluation is only +0.30 in White's favour, you can play this line with confidence at club level: small positional advantages rarely decide games without a mistake.

By far White's most common move here is Nc3, appearing in over 2.4 million games — far ahead of any other reply. White scores 51.8% with it — nothing scary. The second most popular is Bg5 (1.18 million games, White 50.5%), pinning your knight and often leading to classic Queen's Gambit Declined positions. e3 (880k games, White 49.6%) and cxd5 (346k games, White 49.6%) both score under 50% for White, meaning Black actually has slightly better results against those moves. The engine's top choice is g3, though White only scores 54.6% with it despite the computer's preference — a reminder that engine evaluations and human results don't always match.

Stockfish recommends g3, preparing a kingside fianchetto. After g3, the best continuation is Bb4+ Bd2 a5 — you immediately put pressure on White's undeveloped position by checking on b4, forcing Bd2, then pushing a5 to gain space on the queenside. This line is sharp but principled: you develop, gain space, and avoid passive play. White's fianchetto setup aims to control the centre from a distance, while your ...a5 plan says you're not letting White have all the fun on that side of the board. Practice this exact line in the drill below.

The FACTS highlight Bf4 as an inaccuracy in this position, losing roughly 0.7 pawns of advantage (the best move was g3). This is a classic case of White playing a seemingly natural developing move without noticing something better. If your opponent plays Bf4, you should be pleased — the engine says they've slipped. While the concrete response depends on the position, the key takeaway is that not all natural-looking moves are equal here. The statistics back this up: Bf4 scores only 50.4% for White in 237,837 games, right around average, but the engine knows White could do better. Stay alert for these small edges.