Bayonet at Midnight: The King's Indian 9.b4 and the 9...a5 Answer

Three in the morning in Nha Trang, the surf reduced to a thin backing track behind the ceiling fan. I reopened the game on screen for the fourth time in one night and stopped, again, at the same move: 9.b4. In Western literature it is called the Bayonet Attack. The name was attached to a queenside pawn thrust, but it captures something true about how a chessboard treats a player: a close-range stab, no noise, no explosion, only a space opening on squares nobody had noticed before.

Last week a young coach from Da Nang sent me a link to a teaching product titled "A night unlike any other": a Black repertoire against the King's Indian Defence, Classical variation, Mar del Plata branch, with 9...a5 offered as the answer to the bayonet. The title is plainly a pun on a night bayonet assault, the military metaphor that chess marketing still uses to sell a perfectly ordinary move.

I spent two days taking it apart. The product contains almost no verifiable data. No named author, no publication date, no reference games, no engine statistics, no database. That does not make it worthless. It makes it an offer rather than a proof. The stadium is empty, but the applause of 2026 still echoes inside every old record — and on a chessboard the empty stadium is the hall whose lights are switched off after the players leave, with nobody left to explain the last move to whoever is still sitting.

Mar del Plata: a name given to White's pain

The King's Indian arrives indirectly: 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7. Black concedes the centre, then builds a surveillance wall over it with a bishop on g7 and a pawn chain waiting for its moment. The Classical branch the product targets runs 4.e4 d6 5.Nf3 0-0 6.Be2 e5 7.0-0 Nc6 8.d5 Ne7. White locks the centre and the game splits into two parallel wars: White attacks on the queenside, Black storms the kingside. The variation's name comes from the tournament in Mar del Plata, Argentina, in the early 1950s, where this structure was analysed and played so heavily that a city became a chapter title.

To a beginner it looks peaceful: pawns locked, nothing happening across four central ranks. To anyone who has sat with it, this is one of the most dangerous structures in modern chess, because both sides have signed a contract. Whoever opens their flank first wins. White has four moves to break through on the queenside. Black has four moves to break through on the kingside. Nobody has time to change their mind.

The bayonet 9.b4 and its price

9.b4 refuses the slow tempo. The b-pawn stabs into the queenside before Black can organise, aiming at c4-c5, hitting d6, and denying Black's knight any comfortable queenside square. The cost is equally clear: c4 and b4 are no longer protected by pawns, and White must live with Black poking at exactly those squares. More importantly, 9.b4 raises a question White must answer with action rather than thought: if Black thrusts ...a5 at once, does White take or push to b5?

Based on my experience watching these games, the bayonet became fashionable in open international events through the 1990s and entered textbooks as a standard variation. It is not an engine-era invention. It is the consequence of a simple realisation: in a position where both sides have committed to opposite flanks, speed on the queenside is everything.

Two doors for Black: 9...Nh5 and 9...a5

9...Nh5 is the main theoretical answer. The knight heads for f4, forcing White to redirect with 10.Re1 and 11.Bf1, after which Black launches ...f5, ...g5, ...h5. Its character is forcing, which suits an age in which everything is counted.

9...a5 is not forcing. It asks a structural question: how long does the b4 pawn intend to live? If White takes with 10.bxa5 the a-file opens and Black can recapture with a rook, turning the a5 pawn into an anchor for ...c5. If White pushes 10.b5 the queenside freezes, c4-c5 becomes harder to arrange, and Black gains time on the opposite wing.

The difference between the two moves is the difference between an order and an invitation. Notably, the product's own author concedes that objectively the chances after 9...a5 are equal. That concession matters more than everything else in the promotion. It means the product's value lies in preparation asymmetry and surprise, not in objective superiority.

Three branches, one strategic skeleton

The published source supplies no reference game, no database, no engine evaluation. What follows is a reconstruction of the strategic skeleton, explicitly flagged as informed inference.

After 9.b4 a5, White has three ways to handle it. Taking with 10.bxa5 opens the a-file and hands Black an active heavy piece on the rim, plus a concrete target structure for ...c5 or a knight rerouting via ...Nd7 and ...Nb6; in return White sometimes keeps an extra queenside pawn. Tempo matters more than material. Pushing 10.b5 closes the queenside, lets Black proceed with ...f5, ...Nf6, ...g5 and ...Rf7 without interference, and the cost to White only appears around move 25, when there is no queenside break left to force Black to turn around. Ignoring the challenge with 10.Bg5, 10.Nd2 or 10.Ne1 pulls the game back toward older main lines, each of which Black must also know.

Black's plan after 9...a5 can be summarised as: hold e5, wait for ...f5, reroute the e7 knight, use a rook to control the a-file once b4 disappears, and above all deny White a fixed target. Black plays with fewer forcing moves in exchange for keeping more options. 9...a5 should be understood as an established alternative, not a refutation.

The cost not printed on the label

A Black King's Indian system cannot consist of 9...a5 alone. It must answer a minimal question: what if White does not play 9.b4? White has three major exits — 9.Ne1, 9.Nd2, 9.Bg5 — each opening its own body of theory, plus slower options such as 9.a4. A system advertised as "few lines" really has few lines only when the opponent agrees to walk through the door it opened.

This mirrors the transfer market I have covered for years. A signing fee for a free agent looks harmless on a bulletin because it never appears in the transfer-fee column, yet it is still money, still an obligation, still a burden on the wage bill, simply placed outside normal scrutiny. The hidden cost of a chess opening works the same way, and it covers at least three items: the branch after 9.Ne1, working knowledge of the main 9...Nh5 line, and the general skill of playing closed positions, which no video can install.

A seven-year-old promise about engines

One advertising claim deserves scrutiny: that computers often struggle with the complex strategic positions arising after this move. That claim belongs to a specific era, and the era has ended. In December 2026 DeepMind published AlphaZero's self-taught results; in September 2026 Stockfish 12 shipped with the NNUE neural network, fundamentally changing how engines evaluate positions. Before those milestones a traditional engine could genuinely struggle in locked, manoeuvring positions. After them the gap narrowed enormously.

I hold no engine data for 9...a5 and I will not invent a number. What I can state with high confidence is that "computers don't understand strategic positions" has lost its value as a sales argument. A cheap test exists: feed the position after 9...a5 to a modern neural-network engine, let it play both sides for several dozen games, and see whether Black is genuinely comfortable. If results drift toward White, the buyer knows what is being purchased: a surprise weapon for a specific game, not a foundation for a career. A record never grows old, it simply waits for someone who knows how to listen. An argument, unlike a record, does age — and it ages fast.

The buyer is never given an explanation on the spot

For years I have written about what I consider the central problem of modern sport: fans are left out of decisions. A referee reviews footage, reaches a conclusion, and the stadium hears no explanation. Outrage erupts not because the decision is wrong but because there is no channel for understanding.

The opening market runs on the same structure. The buyer receives a product with no reference games, no statistics, no author identity, no publication date. In chess those data exist and can be published. Their absence resembles a major match controlled from a closed room while the crowd only sees the result. I once travelled across Russia with a group of Vietnamese supporters during the 2026 World Cup, and what I remember is not a goal. It is a night in Kazan, a broken bus, falling snow, and an elderly Russian man driving the group to the stadium in an old car. Nobody asked him to explain anything, yet he talked the whole way. People need to be understood. Someone buying a chess course needs the same thing: not a promise of victory, but a clear statement of where they stand.

Why 9...a5 still deserves study

Even so, I would tell any improving player to spend time on 9...a5 — not because it is stronger than 9...Nh5, but because it teaches a skill that forcing variations cannot: how to play when nothing is compulsory. Studying a forced line is memorisation. Studying a strategic line is judgement. That is the difference between running a 400m from a pre-set split plan and having to invent your splits in a final where the favourite kicks earlier than expected.

The genuine strength of the move, described clumsily by its own marketing, is that it poses a structural question before a move-by-move question. It forces White to decide whether the b4 pawn is a structure or a move. Answer wrongly and the queenside falls behind, with no compensation, because Black's kingside is still coming. Every stadium I have walked through has a secret door, and the entrance is always a person. On a chessboard the entrance is always a move that theory called ordinary until somebody won with it.

Depth against breadth

Modern chess tempts players to dig extremely deep into one system and treat it as a single home. At the top level that once worked. Now opponents can look things up. When every game you have played sits in a database, depth stops being a secret and becomes data. Hence selective breadth: two or three systems known well enough to handle surprises, plus one known extremely deeply. 9...a5 belongs to the first group. It does not need to become a foundation. It needs to be available, at the right moment, in the right game. That is an investment in adaptability rather than probability.

How to test an opening system before buying

Define the scope. Does the system cover one move or a whole complex? If the seller does not say, treat the silence as the answer. Count the exits. With the King's Indian, start with 9.Ne1, 9.Nd2 and 9.Bg5. Look for reference games; a product that shows no evidence usually leads with promises instead. Then verify with a modern engine, letting it play both sides from the key position. None of this requires special skill, only patience — and patience is the one thing opening sellers never encourage in a buyer.

What remains after a move

I keep an old VHS tape from the 1980s, grainy beyond recognition on the back rank, not because the game was good but because at the end, after the handshake, the commentator lingered for two sentences about a move he found memorable. Nobody watched that part. I did.

9...a5 is that kind of move. It does not demolish Mar del Plata theory, it does not make Black stronger, it merely opens one more window — and that window may amount to nothing unless the player builds the rest by hand. The real lesson reaches beyond a single move. In a market where everything can be packaged, a player's value lies in the ability to redistribute their own weights: what must be memorised, what must be verified, what must be discarded even though the seller insists it matters. Chess and the running track agree here. No system runs for you. A system only helps you know where you will stop if you have to stop.

Tonight, if you open a King's Indian and push 9.b4, try answering the question Black is obliged to answer. If you can answer with a plan rather than a line, you have learned more than any product can sell you.