The 9...a5 Move in the Bayonet Attack: A Backup Plan or a Solution to the King's Indian Defence?

That night I stayed at the chess club on Phú Lợi street until nearly midnight, after the plastic chairs had been stacked and the keyholder stood waiting at the door. One game was still unfinished on the last board: a seventeen-year-old boy with the white pieces against a man thirty years his senior. The boy played 9.b4 — the Bayonet Attack — and then went quiet. The man went quiet too, his eyes fixed on a single point on the board, his clock ticking from minute seven to minute seventeen.

I knew what was happening in his head. He was searching his memory for 9...Nh5, but his hand would not put the piece down. Across the board, the boy was strangely calm — he had not memorised enough to be afraid.

At home, on my desk, there was an opening training video titled "A night unlike any other". It promised exactly what the man outside needed: an escape from memorisation.

The King's Indian Defence is a system in which Black concedes central space in exchange for a position that can explode at any moment. The sequence that led to the game on that board was fixed long ago: 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Nf3 0-0 6.Be2 e5 7.0-0 Nc6 8.d5 Ne7. The branch is called Mar del Plata, after the tournament held in Argentina in 2026, where the approach was first gathered into a system with a name of its own.

From there White has several approaches, and 9.b4 — the Bayonet Attack — is the most direct: pushing the queenside pawn to gnaw at Black's structure and disrupt the stability Black is building. The name "bayonet" describes the move accurately: it does not attack the king, it attacks the foundations.

The most common reply, regarded as the main line, is 9...Nh5, followed by ...f5, ...Nf4 and ...g5 — a race that drives straight at the white king. This is the kind of position calculated move by move, where opening memory has direct value, and where chess engines have scoured every corner.

The move 9...a5 takes a completely different road. Instead of a race, Black strikes at the very pawn structure White has just pushed forward: ...axb4 will open the a-file, weaken the b4 square and turn the queenside into a long-term bargain. The position slows down, the pieces stop charging at each other, and both sides enter what I still call "the forest without a map".

I have a habit of studying openings by replaying old games. Sixty days of listening to recorded footage, and I imagined the roar of empty seats — because in a slow position, the only thing that echoes is the sound of a piece touching the board.

The first thing to state clearly: 9...a5 is a genuine backup option in theory, not a new discovery, and certainly not a refutation of the main line. The promotional material itself concedes that "objectively, chances are equal". That concession matters more than the rest of the introduction combined. If chances are equal, the product is not selling objective superiority. It is selling two other things: asymmetry in preparation and the element of surprise.

Second, the material claims that computers often struggle with complex strategic positions. That phrasing has a date on it. It was born in an era when chess engines assessed closed positions through crude, error-prone evaluation functions. From 2026, when reinforcement-learning engines appeared, and especially from around 2026, when compact neural networks became the standard, the ability to assess static, manoeuvring positions changed completely. If computers now understand those positions far better than a decade ago, the greatest advantage of a "few lines, more strategy" option narrows precisely where it is advertised as strongest. Some of that value remains, but it is smaller than the seller wants you to believe.

Third, and this is the highest-value and least verifiable claim: that the player does not need to memorise many variations. A repertoire built on 9...a5 in practice still needs at least two layers of fallback. The first is White's alternatives to 9.b4 — 9.Ne1, 9.Nd2, 9.Bg5 and similar moves. The second is knowledge of the main line 9...Nh5 itself, because after 9...a5 White can perfectly well steer the game back into familiar structures. The cost of preparation does not vanish; it moves from memory to judgement. And judgement trains more slowly, is harder to check and harder to share.

Fourth, the scope of the product is undefined. The material does not say whether the video covers only the 9.b4 branch or the entire King's Indian complex after 7...Nc6. For a buyer, those are two very different costs, and the ambiguity favours the seller.

Fifth, and this is the largest gap: no player is named, there are no database figures, no engine evaluations, no reference games, no publication date and no information about the author. Almost every information point is unsourced, except three attributed to the author himself. Every technical claim in the source material sits in a state of data pending verification.

Sixth, the target customer is clearly positioned: the club player who prefers to think for himself rather than memorise. That is a sensible commercial position, but it announces its own limit. Elite players cannot abandon main lines for aesthetic reasons — they pay for it in points.

Finally, the title "A night unlike any other" is very likely a military pun: a bayonet raid by night, gesturing at the name Bayonet Attack itself. The title is a marketing hook, not information.

Based on my experience following games at amateur events and replaying top-level games many times over, I see a pattern: players who choose a "few lines" option usually do not win because the option is better, but because the opponent is forced to think for himself in a position he has never had to think about.

The blind spot of collective memory is that we equate surprise with advantage. Surprise has a shelf life. A move that has just been packaged as a commercial product begins to lose its surprise at that very moment — the first buyer has an edge, the thousandth does not. That is the paradox of every opening repertoire sold on a disc.

The second blind spot is the illusion that strategy is cheaper than memory. It is not cheaper; it is more expensive in a different way. Memory is an investment paid up front, checkable and transferable to a student. Judgement is an investment paid in instalments, with no final statement, and it usually shows up only after you have already dropped half a point. Shifting the burden from memory to judgement does not make the burden lighter — it only makes it harder to see.

The third blind spot belongs to the identity of the King's Indian Defence itself. The system grew up on two-wing races, where both sides charge and the one who is a tempo slow loses. Abandoning that characteristic is a trade-off, not an upgrade. It may be good for a particular player at a particular stage, but it is not free.

And the final, subtlest blind spot: when a product tells you that you do not need to memorise, what it is really selling is permission to stop studying. A chessboard holds more than pieces; it holds the calluses of a life.

The man at the club that night eventually placed 9...a5 on the board. The game drifted into a slow, balanced position, beautiful in an old man's way: nothing exploded, only small moves piling up. He no longer calculated as fast as he once did, but every move was a story not yet told. He lost on time around move forty-something.

The game ended, yet the echo of that move still rings in my memory. And I keep asking myself: if 9...a5 buys you only thirty extra minutes of thought in a balanced position, what is really worth buying — a new line, or a new way of studying?