King's Indian Defence, the Bayonet Attack and the move 9...a5: reading a repertoire promise again

A night that was given a name

“A night unlike any other.” Someone glued that sentence onto the cover of a chess opening training product, and in St Petersburg, with late snow tapping the window frame on the fifth floor, I sat and read it the way one reads a bad poem. Nobody names an opening variation with poetry unless they want the buyer to hear steel before seeing the position. Bayonet — a blade. In military English it is the close-combat weapon fixed to the muzzle of a rifle; in chess it is the name researchers gave to White's 9.b4 in the King's Indian Defence. An infantry move that marches straight up, without hesitation, without mercy for the central structure.

That night I did not watch a game. I read a sales pitch. It contained three promises: that Black's 9...a5 is an antidote to the bayonet storm, that it is a plan-based approach rather than a memory-based one, and that it contains “very few concrete lines” to memorise. The last four words — objectively chances are equal — sat at the end like the small italic print on an insurance contract.

When the stadium holds no ball, memory starts scoring. For me that sentence holds at the chessboard too. An opening product with no verifiable data leaves only the writer's memory of it, and memory always favours the person doing the telling.

Context: why anyone goes looking for an antidote

The King's Indian Defence stands in modern chess history as one of the strangest defences humanity ever trusted. Black voluntarily concedes the centre, locks itself into a space disadvantage, and pays it back with a wave on the kingside. No other defence demands such faith from its practitioner: you let your opponent take space, and you believe that at some point their king will crack.

The classic move order runs 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. White has sealed the centre with d5, Black massed pieces towards the kingside. The structure is locked, and from here the game becomes a parallel race between two wings: White pushes c4-c5 and b4-b5, Black answers with f5, g5, g4 and sometimes h5-h4. Whoever arrives first wins; whoever is one tempo late loses the house.

The variation that carries this structure's name — Mar del Plata — is tied to the tournament held in Mar del Plata, Argentina, in 2026, where the two-wing race took its modern shape and became known worldwide. From then until now, nearly seven decades, the system has remained one of the heaviest theoretical battlegrounds in chess. Heavy in the sense of memory volume, not in the sense of move count.

The name Bayonet appeared when White looked for a way to dodge lines that had been memorised to death and chose 9.b4, hitting Black's queenside structure before Black could start the kingside engine. Through the 1990s the variation became a fashionable choice at the highest level; theory books had to give it a chapter of its own. The pressure shifted to Black: the question stopped being what Black wanted to play and became what Black had to do in order to survive the first twelve moves.

The traditional answer is 9...Nh5, lifting the knight from e7 to the kingside belt. The familiar main line usually runs 10.Re1 f5 11.Ng5 Nf6 12.Bf3 c6 13.Ba3, and from there the game enters territory every opening coach knows: one wrong move and the whole position appears to collapse. The price of that line is memory. It forces a club player to maintain a living library of variations, updated whenever a new game is published, and to accept that an opponent — someone with an engine at home — can overtake them in a single evening. There is also an older branch, 9...Ne8, approaching the kingside the long way round; it was the main line for decades before speed pushed it aside.

Core: what 9...a5 actually does on the board

The product I was reading proposes a third path: 9...a5. Formally, it is a modest rook-pawn push so simple that one has to look twice. The a7 pawn advances one square; it does not strike at the centre, does not open a file for heavy pieces, does not attack directly. But read in the grammar of a queenside structure, the move starts to speak.

The a-pawn advance is much more than a waiting move. It defines the queenside contest before White can organise it: if White wants b4-b5, Black answers a5-a4 and freezes the queenside with White's own pawn. If White wants to keep the tension on b4, Black can open the file with a5xb4 at the right moment, turning the a-file into a supply line, and keep the knight on e7 ready to jump towards d7, c5 or f5 depending on events. Black is therefore not forced into an early commitment. The e7 knight can still go to h5 as before, but it can also go to d7, or stand still for a few more tempi and let the kingside look after itself with f5 and g5.

That is the attraction of such a move: it buys time by posing the question first. White is forced to decide whether to defend b4, and every answer carries a cost of its own.

Technical level matters here, and I speak as someone who has followed professional chess since 2026, when I was still a tournament player and organiser before moving into chess media. 9...a5 is not an invention. It is an existing alternative with practitioners, documentation and lineage. In theoretical terms it is a minority but legitimate sideline. Packaging it as an “antidote” is repositioning, not discovery.

Based on my experience following games and the opening literature I can check, the sophistication of this proposal sits at the level of an established alternative. Compared with the main line 9...Nh5, it is not objectively superior. The presenter concedes as much with the phrase about objectively equal chances. Its entire value lies elsewhere: in preparation asymmetry and in surprise.

The transfer market is where people sell the future to buy hope. The opening market works the same way, except that what is sold is memory and what is bought is peace of mind.

The contrarian angle: the data gap is larger than the technical gap

The most striking feature of this product is that it offers nothing an outsider can verify. No presenter name with credentials. No named players. No publication date. No database cited. No reference game. No engine evaluation for any move. No engine match rate. No hit counts for the move. No win-loss statistics. No average centipawn loss.

In other words, this is a single-source commercial promotion, and every technical claim in it is unverified by definition. I record each technical assertion with its status: data pending verification. That is not an accusation. It is the professional rule of someone who has spent too many years at an editorial desk.

This leads to a more important question: scope. Nowhere in the material does it say whether the product covers only the 9.b4 branch or the whole King's Indian complex after 7...Nc6. Those two scopes imply completely different preparation costs. The buyer does not know whether they are buying a chapter or a book. For an opening product, that is a more serious information gap than whether the move wins outright.

Then comes the “very few concrete lines” claim — the most valuable and least verifiable statement in the entire pitch. Imagine a serious Black player building preparation around 9...a5. They still need two other blocks of knowledge. The first is White's alternatives to the Bayonet: 9.Ne1, 9.Nd2, 9.Bg5 and a few rarer tries. The second is knowledge of the main line 9...Nh5 itself, because after 9...a5 White may steer back into familiar structures — pushing pawns or regrouping pieces to drag the game onto ground they know better than Black. Without that second block, Black drifts into passivity exactly when they believe they have escaped.

Reducing the memory load is a legitimate goal. Claiming the approach needs almost nothing but understanding the plan is a deliberate exaggeration.

And here is the point that deserves the most words, because it is the heart of the matter. For a product presented recently, the argument that computers often do not know how to handle the arising complex strategic positions is a dated argument. It belongs to the era before neural networks changed how engines judge chess.

Recall the milestones. AlphaZero published its results in December 2026, with a system trained by self-play and no human chess knowledge. Leela Chess Zero appeared in 2026 as an open project following that road. Stockfish integrated NNUE-style neural networks from 2026 and quickly transformed how leading players use engines. The difference between older and newer engines in closed, manoeuvring positions is a difference of kind, not degree. Today's engines assess positions that once confused them with far greater reliability.

That turns the “machines cannot understand it” argument into a sales mantra rather than a technical statement. I rate it high confidence as a general claim about engine evolution, and medium confidence when applied to this specific position, because no evaluation is supplied to check it against.

One small detail deserves attention: the title. “A night unlike any other” is very likely a military pun — a night bayonet raid, tied to the name Bayonet Attack. If so, the title carries atmosphere rather than information. And a product sold on atmosphere needs the buyer to inspect the data even more carefully, because data is usually the only part that cannot be painted over.

Positioning matters too. The line about suiting players who like to think for themselves rather than memorise long variations is aimed at club and improving players — the group that accounts for most sales of any opening product. The elite cannot abandon forcing main lines on aesthetic principle: at that level you need to know the forcing branch to survive round four of a three-week event. The product does not promise that, and does not need to. It only needs to make the buyer feel they have found a private road.

At Luzhniki, an old bird understood that landing is also a way of flying. At the chessboard, choosing a less forcing approach is a legitimate way of playing — provided the person choosing knows what they are trading away.

Takeaway

The real value of an opening product lies not in the move it sells but in whether it teaches the buyer to ask the right questions. A pitch with no presenter name, no reference game, no publication date and no engine evaluation can still be useful — as long as the reader understands they are receiving a hypothesis, not a confirmation. If some night you sit down with 9...a5 in mind, remember that what you hold is not an antidote. It is a door into a room you have never measured. The only trustworthy thing in any opening pitch is the count you make of the squares yourself.