Ruy Lopez with d3: a White repertoire against the Berlin and 3...a6

Short answer

A White repertoire in the Ruy Lopez (1.e4 e5 2.Nf3 Nc6 3.Bb5) built around d3: 4.d3 against the Berlin and 5.O-O Be7 6.d3 against 3...a6. The guide uses Stockfish 19 to measure when c3 can be played, the two captures on e5 that lose, the bishop's retreat to c2 against ...Na5, the Nbd2–f1–g3 manoeuvre and Noah's Ark, and shows why avoiding the Marshall Attack is not, on evaluation, the reason to play d3.

The essentials in a few lines

• Against the Berlin, 4.d3 Bc5 measures +0.11. From there the timing of c3 decides: after 5.c3 d5, 6.exd5 keeps +0.21, while castling with 6.O-O drops to -0.26.
• After 4.d3 Bc5 5.Bxc6 dxc6, the capture 6.Nxe5?? loses: 6...Qd4 hits the knight and threatens mate on f2, and the engine measures -5.62. After 7.Nf3??, the game ends with 7...Qxf2#.
• With both sides castled, the e5 bait returns: after 5.O-O O-O 6.c3 d5, 7.Nxe5? measures -2.21, against +0.17 for 7.Nbd2.
• Against 3...a6, the line 5.O-O Be7 6.d3 b5 7.Bb3 d6 8.c3 measures +0.14. After 8...O-O 9.Nbd2 Na5, the retreat 10.Bc2 keeps +0.03; 10.Re1? allows 10...Nxb3 and drops to -0.26.
• The Marshall Attack is not bad for White: after 8.c3 d5 it measures +0.41, higher than the d3 lines in this guide. d3 buys positions without forced theory, not evaluation.

Why d3, and what it doesn't solve

d3 is usually explained as an escape: from the forcing lines of the Marshall Attack and from the Berlin Wall endgame. The second part is true, since the Berlin Wall ending never arises after 4.d3. The first does not hold up as an evaluation argument. Measured by us with Stockfish 19 at depth 22, the Marshall after 8.c3 d5 gives White +0.41, more than any of the d3 lines measured in this guide.

What d3 buys is time without obligations. In the Marshall, White has to know a long forcing sequence to reach that +0.41. Against d3, it is Black who has to decide what to do with a centre that won't open. Players who would rather understand a position than remember a sequence trade a little evaluation for a lot less memory.

This guide covers White's side of the board. What Black does against 4.d3, with the bishop on c5 and the ...d5 break, is in our Berlin Defence guide; here the focus is White's move order and the two moments when taking on e5 loses.

Against the Berlin: 4.d3 and the timing of c3

Against 3...Nf6, 4.d3 is how White keeps the queens on. The engine's preferred reply, 4...Bc5, leaves +0.11; the more passive 4...d6 leaves +0.25. The bishop on c5 is what makes the position concrete, because it points at f2 and it is the piece both traps in this guide rely on.

The move c3 prepares d4 and opens a retreat square for the bishop, and it is also an invitation to ...d5. A player who plays c3 without having decided what to do against ...d5 finds the answer at the worst moment. With 6.exd5 White exchanges the central pawn and keeps the edge; with 6.O-O White lets Black open the centre just as the king has castled, and the evaluation changes sign.

There is a slower order that delays c3 until after castling: 5.O-O d6 6.c3 O-O 7.Nbd2 a6 8.Ba4. With both sides developed, the position measures +0.05, a manoeuvring game rather than an advantage. If Black spends time on 8...Ba7 or 8...b5, about a tenth goes back to White, within the noise margin.

The e5 pawn you don't take

The e5 pawn looks free twice in this opening, and both times the hook is the same c5 bishop. It is the chessboard version of the banknote left on the pavement in a hidden-camera prank: the problem is not what you pick up, it is what you stop watching while you bend down.

After 5.Bxc6 dxc6, White has given the bishop for the knight and the e5 pawn has lost its defender. 6.Nxe5 wins the pawn and removes from f3 the knight that was controlling d4. 6...Qd4 makes two threats at once, the queen on the e5 knight and mate on f2 with the c5 bishop's support, and White can only meet one: 7.Qe2 (-5.81) and 7.Be3 (-5.84) lose the knight, and 7.Nf3?? is mated next move.

With both sides castled, the bait is smaller and the mechanism is different. After 5.O-O O-O 6.c3 d5, 7.Nxe5? Nxe5 8.d4 looks like it solves everything with a fork of the c5 bishop and the e5 knight, and it does win the piece back: the engine's line continues 8...dxe4 9.dxc5 Qe7, and it is the position left behind that measures -2.21. The right move is 7.Nbd2 (+0.17), keeping the tension.

Against 3...a6: 6.d3 and the bishop's c2 square

The first decision is when to play d3. 5.d3, straight after 4...Nf6, is playable and leaves +0.06. Delaying it to 6.d3, with White castled and Black's bishop already committed to e7, is this repertoire's order, because it takes away Black's option of putting the bishop on c5.

On move eight, 8.c3 (+0.14), 8.a4 (+0.07) and 8.a3 (+0.08) sit in the same band, and evaluation does not choose between them. What settles the choice of c3 is structure: it gives the b3 bishop the c2 retreat, prepares a future d4 and takes d4 away from Black's knight.

That c2 square is what Black's knight tests with ...Na5. The b3 bishop is White's best piece in this structure, and letting it be exchanged for a knight that went to the rim costs three tenths: 10.Re1? Nxb3 11.axb3 measures -0.26, against +0.03 for 10.Bc2. With the bishop on c2, Black continues with ...c5 and the position turns into a manoeuvring Closed Ruy Lopez.

The Nbd2–f1–g3 manoeuvre, and the lesson of Noah's Ark

With the knight on f1 the choice passes to Black, and only one of three natural replies keeps the balance. 12...Nc6, bringing the rim knight back to the centre, leaves +0.01. 12...h6 leaves +0.21 and 12...Bf8 +0.26. Two tenths is enough to say the position has better and worse moves, and too little to say anyone is losing.

The reason for all this care over the b3 bishop has a name. In Noah's Ark, an old Ruy Lopez trap, the bishop is trapped on b3 by a chain of black pawns. In a line with an early d4, not the order used in this repertoire, 3...a6 4.Ba4 d6 5.d4 b5 6.Bb3 Nxd4 7.Nxd4 exd4 8.Qxd4?? allows 8...c5 and ...c4, and the engine measures -4.78; with 8.c3 the position is 0.00.

The lesson applies to the whole repertoire. When Black plays ...b5, ...c5 and ...c4, White's bishop needs somewhere to go before the chain closes. With d3 and c3 on the board, c2 is always available, which is why c3 in this system is worth more than the hundredths the engine gives it.

Who this repertoire suits

The price shows in the numbers. The d3 lines measure less than the Marshall and less than the Berlin Wall endgame, which our Berlin Defence guide measures at +0.24. Choosing d3 means choosing the kind of game you will play, not the evaluation on move fifteen.

For study, three things are memorised and the rest is understood. Never 6.Nxe5 after Bxc6 dxc6. Play c3 only with the answer to ...d5 ready. Bring the bishop to c2 when Black's knight reaches a5. Everything else, the knight's route to g3 and the timing of d4, comes out of the structure, and it is the same logic as the slow Italian.

For 1...e5 players with Black, our open games repertoire guide shows the other side of these positions, and the guide to building an opening repertoire explains why a system with recurring plans pays off more than a collection of memorised lines.

Frequently asked questions

Openings mentioned in this guide

• Ruy Lopez (Spanish Opening) (C60)
• Marshall Attack (C89)
• Italian Game (Giuoco Piano) (C50)