The Bayonet Attack: 9...a5 and the Limits of Machine Memory

The board stops at move nine. White has just pushed the pawn to b4 — the Bayonet blade driven into the queenside of the King's Indian Defence. Behind that seemingly simple move lies a spatial map being redrawn square by square: White seizes space, threatens to open the c-file, and forces Black to choose between two roads. For decades, the near-default answer has been 9...Nh5 — the knight leaping to the rim of the board, accepting a chaotic fight in which both sides must recall exactly every step the computer has already mapped out. But a newly released training product proposes a different turn: 9...a5. A quiet pawn move, almost forgotten by the flow of modern theory. It opens no combination; it opens a question larger than one opening line: can a player still beat the machine by understanding, rather than by remembering?

To see why that move stirs debate, one must rebuild its coordinate system. The King's Indian Defence is a classic hypermodern system: Black concedes the centre to White and counterattacks from afar. The typical 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 — the Mar del Plata Variation, where the two wings charge at each other like armies leaving their fortresses undefended. White attacks the queenside, Black attacks the kingside, and the winner is usually the one who arrives a tempo earlier.

Within that structure, White's 9.b4 — the Bayonet Attack — is a sharp blade. White does not bother to shield, but pushes the pawn to b4 to drive Black's knight off c6, clearing the way for c5 and b5. The move breaks Black's plan: it denies Black the leisure to swing pieces to the kingside. Based on my experience following matches, most games at club and professional level are not decided by beautiful blows but by one side forcing the other off its familiar map. The Bayonet does exactly that: it pushes the opponent out of the safe zone of memory.

And this is where 9...Nh5 appears. The knight leaps to h5, threatens exchanges, opens the g-file, and accepts a position the computer has dissected to the bone. This is the road of the engine warriors: they need not understand why, only remember correctly. The trouble is that at a high level both sides remember equally, and the game becomes a memory contest rather than a contest of thought. I once timed a famous comeback — fourteen seconds from the corner kick to the decisive shot — and recognised the same thing in chess: the decisive moment lies not in the final move but in the gap before it appears. Fourteen seconds — enough to redraw the opponent's entire defensive map.

The new training product proposes 9...a5 as an escape from that spiral. The flank pawn move attacks the b4-pawn, forcing White to decide at once, while preparing slower plans for Black: ...c5, ...Be6, ...Nd7. No more forced lines, no more sequences to be recalled beat by beat. What is sold here is not a lightning blow but a strategic position where plan matters more than calculation. The same board, the same pieces, but the genre of the game has changed.

9...a5 does not aim to prove Black is better; it aims to change the genre of the game — from a memory contest to a contest of understanding. The accompanying argument is clear: the resulting position is objectively balanced yet practically dangerous for White, because White usually prepares by memorising forced lines rather than by understanding strategic plans. If Black grasps those plans, he can lead his opponent into waters where the map of memory no longer works. This is an argument about preparation psychology rather than about objective advantage — and that is its subtlest point.

Compared with 9...Nh5, the choice of 9...a5 trades speed for control. It demands no superhuman memory, but it demands patience and structural understanding. It is also not the only option: 9...Ne8 is a gentler retreat, while 9...Nf6 returns to a familiar structure. What is striking is that the very existence of such a product reveals a real demand: a section of King's Indian players feels squeezed by theory they are forced to memorise, and they are looking for an exit.

But here I must stop and read more closely. The introduction claims 9...a5 has very few concrete lines, and that computers often do not know how to handle the arising complex strategic positions. No quantitative data backs that claim. No engine match rate, no ACPL, no database statistics, no cited games. An argument without data is only an unverified hypothesis. Every formation is a hypothesis until the ball rolls — and here the ball has not yet been passed.

Moreover, the product's central premise is rotting over time. At a moment when Stockfish and Leela evaluate the strategic positions of the King's Indian Defence reliably, the claim that computers do not understand becomes harder to defend every year. I once spent 250 days during the pandemic coding 380 games and writing a 120-page report. In 120 pages of report, I found what a season never records: repetition. What I learned is not that computers are weak, but that humans tend to repeat themselves — and whoever understands that repetition will beat whoever merely remembers the moves.

So where does the real value of 9...a5 lie? It lies in the audience, not on the board. For strong club players — who meet the Bayonet less often than grandmasters, who want to understand plans rather than memorise — such a strategic line has genuine practical value. For professionals, where everyone has a database and everyone knows 9...a5, the surprise advantage vanishes. One move, two fates, depending on whether your opponent prepares by memory or by understanding.

The most telling thing about this story is perhaps its format rather than the move 9...a5. A static disc product represents a mature and declining distribution model. Opening knowledge today flows through subscription platforms and interactive databases, where content can be updated after the engine finds a correction. A disc cannot do that — a structural disadvantage, not the author's fault.

Meanwhile, in another corner of the European board, a school event took place in Tsaghkadzor, Armenia, where European school chess champions were honoured in a cultural celebration. Armenia is a chess nation, where the intellectual sport is taught in schools as part of the national curriculum. Such events make no hot news, but they feed the talent pipeline — something an opening product never touches.

Seen from afar, both stories speak of the same thing: chess is stratifying. One layer chases forced theory and machine memory, another returns to strategy and understanding. The move 9...a5 sits on the boundary between the two. It stands between two extremes — a revolution and a mistake. It is a proposal: choose to understand rather than to remember.

A spatial map never lies — it only exposes what we want to believe. If you believe the machine cannot grasp strategic positions, you will find in 9...a5 a weapon. If you believe the machine understands everything, you will find there an illusion. The answer lies in the next game, when your opponent sits down with an updated database and a prepared plan. Will he really step into the waters you want him to step into — or are you the one swimming in someone else's map?