The Bayonet Attack, 9...a5, and an Outdated 'Engines Don't Understand' Claim

On move nine, the white pawn leaves b2 and steps to b4. For most of the people sitting in the playing hall, it is just a push for space. For anyone who has followed the King's Indian for a few decades, that touch is a declaration: White has chosen the Bayonet, and the game is about to turn into one of the most thoroughly dissected branches of modern chess.

I have stayed behind on many evenings after games like that. Back then I was sitting in the stands, with no idea that I would one day walk down to the board through words. When the roar of the hall falls silent, I hear the true breathing of the game: the sound of a player weighing two roads.

This time, the second road is called 9...a5.

The King's Indian and the split in tempo

The King's Indian begins 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Nf3 O-O 6.Be2 e5. Black gives up the centre from the start, builds a hypermodern setup and counterattacks later. Its sharpest branch is the Mar del Plata, where Black puts the knight on c6 and the pawn on e5, and both sides launch campaigns on opposite wings. In my own classification, the system sits in the ECO range E97 to E99.

White's ninth move, 9.b4, is the Bayonet Attack. The b4 pawn is not there to capture anything immediately. It gains queenside space, prepares b5 to chase the black knight off c6, and declares that White intends to play actively rather than wait. It is the fashionable branch, popular with young players precisely because it forces Black to remember a great deal.

Black's orthodox reply is 9...Nh5. The knight goes to the rim, eyes f4 and prepares ...f5 to open the kingside. After that move the position heats up at once: White pushes b5, Black pushes f5, and both sides hurl themselves into forcing lines where a single lost tempo can cost the whole game. That very forcing character makes it computer territory.

Then there is 9...a5.

How 9...a5 changes the nature of the game

The move 9...a5 has been known for a long time; it is not a new invention. The a5 pawn attacks the b4 pawn head-on and forces White to decide immediately: take, defend, or push on. White may choose 10.Ba3 to keep the pawn structure intact, or simply 10.bxa5 and accept a completely different type of position.

What matters is that this move changes the character of the game. After 9...Nh5 everything revolves around concrete calculation. After 9...a5 the centre closes, the pieces do not rush at each other, and the story shifts to pawn structure. Black gains the a5 square as a springboard, can use the a- and b-files, looks to trade cramped minor pieces, and only then considers ...f5. White keeps the space but must answer a hard question: what to do with the b4 pawn now that it has become a target?

A training product in DVD format has been promoted around 9...a5, presenting it as an antidote to the Bayonet. The sales message is clear: instead of memorising the forcing lines of 9...Nh5, a player can enter a strategic battle with very few concrete lines, where chances are equal. The product targets players who prefer to think for themselves rather than memorise.

For me, the most interesting part is how the argument is constructed. The material sets up two worlds: 9...Nh5 opens the position, forces the lines out and makes computer analysis important again; 9...a5 leads to sound positions with few concrete lines, where a human can fend for himself. That is a psychological argument far more than a theoretical one.

Based on my experience of watching games, the approach is not unreasonable at all. In team events, where every board can decide the match, an option that depends less on memory is a genuine gift. The problem is that it has been inflated into too large a promise.

An old premise behind a new promise

That promise rests on a dated premise. The material claims computers often do not know how to handle complex strategic positions of the King's Indian type. That is pre-AlphaZero language. Since the AlphaZero paper appeared on arXiv on 5 December 2026, and since the neural-network and NNUE engine generations that followed entered analytical practice, engines handle long-term strategic positions far better than that sentence suggests. The move 9...a5 creates no no-go zone for an engine.

Beyond that, 9...a5 is a secondary branch. It does not overturn 9...Nh5, and a well-prepared White player still has play. The claim of equal chances is a positional judgement, not evidence. The material offers no sample game, no evaluation figures and no named author. For anyone who reads data professionally, an argument missing all three should be treated only as the starting point of verification.

One detail about the source caught my attention. The material I read circulated under a headline about a team-event round report in Samarkand, Uzbekistan, while the body discussed this opening product. The headline and the content do not match. For anyone intending to cite it, that is a signal to trace the original before relying on it.

Strategy is like people: you have to know who you are before you try to become someone else.

What remains after all of it

The Mar del Plata King's Indian still lives at the highest level, but mostly as a surprise weapon. There, names such as Nana Dzagnidze and Nino Batsiashvili, Georgian women who treat the King's Indian as part of their identity, show that the system is far from obsolete for someone who knows how to use it.

The real value of 9...a5 lies elsewhere. It teaches a student to read pawn structures rather than to memorise. A secondary move, sold with a dated argument, can still be useful if the buyer knows what he is buying. And the thing worth tracking is the database: if 9...a5 starts appearing in classical games at the 2700 level, its premise will be confirmed or refuted on the board itself. The board has never stopped waiting.