Sicilian Dragon: a Yugoslav Attack repertoire with opposite-side castling

Short answer

The Yugoslav Attack arises from 1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 g6 6.Be3 Bg7 7.f3 Nc6 8.Qd2 O-O 9.Bc4 Bd7 10.O-O-O, when both sides castle to opposite wings and every advancing pawn opens lines against one's own king just as much as the opponent's. This guide measures the main tabiya and Black's most-tested reply to the h4-h5 advance with Stockfish 19, showing that White's edge here depends on precise move order, not on a line that "wins by itself".

The essentials in a few lines

• After 10.O-O-O, White's edge runs from +0.45 to +0.67 depending on Black's reply: 10...Rc8 (+0.45) and 10...Rb8 (+0.48) are practically equivalent, and only 10...Ne5 (+0.67) concedes more than half a pawn.
• The sequence 11.Bb3 Ne5 12.h4 h5 is the most played reply to White's advance: it stops the h4 pawn before it can force open the h-file, reducing White's edge to +0.32.
• In that position, the thematic 13.Bh6 is the worst of the three candidates: it measures +0.08, against +0.30 for 13.Bg5 and +0.25 for 13.Kb1 — trading the defending bishop by reflex hands back three quarters of White's edge.
• White's edge in the Yugoslav Attack never exceeded two thirds of a pawn in the measured positions — the reputation of a line that "wins by itself" does not hold in the evaluation; it holds in the practical difficulty of defending precisely as Black.
• Trading the dark-squared bishop with Bh6 is thematically central to White's plan, and that is exactly why it tends to be played too early: in the position measured here, securing the king first with Kb1 is worth three times as much in evaluation as the thematic trade.

The tabiya after 10.O-O-O

The position after 1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 g6 6.Be3 Bg7 7.f3 Nc6 8.Qd2 O-O 9.Bc4 Bd7 10.O-O-O is the tabiya that defines the Yugoslav Attack: both kings have castled to opposite wings, and every advancing pawn opens lines against its own king as much as against the opponent's.

The three most-tested black replies — 10...Rc8, 10...Rb8 and 10...Ne5 — measure, respectively, +0.45, +0.48 and +0.67 (Stockfish 19, depth 22). The first two differ by three hundredths of a pawn; what changes between them is the order in which Black prepares ...b5 and whether the f6-knight remains available to jump to e5 sooner or later.

This is a position where memorising moves without understanding the race is costly for both sides: a lost tempo in White's attack tends to cost more than in closed positions, because the opponent is racing in the opposite direction at the same time.

The h4-h5 advance and Black's reply

11.Bb3 removes the bishop from the diagonal before ...Ne5 can hit it, and 11...Ne5 is the natural follow-up: the knight occupies a strong central square and prepares ...Nxc4 or ...Ng4 depending on White's play. The 12.h4 advance is the start of the declared race — the h-pawn heads for h5 intending to trade on g6 and open the h-file against Black's king.

12...h5 is the most played answer, and it works for a simple reason: it stops White's pawn before it reaches h5 on its own terms, delaying the opening of lines. The cost is that Black's h5-pawn itself becomes a target — nothing in the position stops White from attacking it with Bg5 or simply trading heavy pieces on the h-file later.

In the resulting position, the three natural candidates are not worth the same. 13.Bg5 (+0.30), keeping the tension, and 13.Kb1 (+0.25), removing the king from the a7-g1 diagonal before anything else, preserve the edge White had; 13.Bh6 (+0.08), the thematic trade of Black king's defending bishop, is the one that hands it back. The move most quoted in introductory material is the weakest of the three here, and the gap is not depth noise: it holds when measured at 26.

Trade the bishop on h6, or defend the king first?

Trading the dark-squared bishop on h6 is the most quoted move in introductory material on the Yugoslav Attack, because it removes Black king's natural defender and prepares mate on the h-file after a heavy-piece trade. It is a real plan, but treating it as automatic ignores that White's own king is exposed on the queenside too.

When Black has already started their own expansion with ...a5 or threatens ...Ng4 against the e3-bishop, delaying Bh6 to play Kb1 first costs nothing at all: the evaluation rises from +0.08 to +0.25, and the white side is still left without an immediate tactical target.

The practical lesson for anyone playing the Yugoslav Attack for the first time: learning the thematic move sequence helps, but playing by fixed pattern without checking the opponent's concrete threats is the most common mistake in this kind of position — and here the engine is blunt: the prettiest move in the position is the one that hands the edge back.

Frequently asked questions

Openings mentioned in this guide

• Sicilian Dragon (B70)
• Sicilian Defense (B20)
• Sicilian Najdorf (B90)