9...a5 in the King's Indian: A Reasonable Detour, With Unverified Faith Attached

There is one position I keep returning to this week. White has pushed the pawn to b4, Black keeps the pawn on a7, and everything depends on Black's ninth move. I turned off notifications, let the room go quiet, and pressed the same move over and over: 9...a5. No stands, no clock, no one whispering commentary. When the stands are empty, you hear the studs on the grass; in chess, it is the sound of wood on wood and the silence between two moves. In that silence, a DVD on the King's Indian Defence is being sold to Vietnamese players as an antidote. And in that same silence, I counted the gaps left behind by the sales pitch.

The product targets the Classical King's Indian, Mar del Plata branch: 7...Nc6 8.d5 Ne7, with White choosing the Bayonet by 9.b4. The pitch is compact: stop memorising 9...Nh5 — the main line that any prepared opponent already has open in a file — and play 9...a5, a rare move, so you can think for yourself instead of recalling lines.

Structurally, 9...a5 does what it is said to do. It strikes at the b4 pawn and forces White to decide on the queenside immediately: push to b5, or trade on b4 and hand Black the a-file. It pulls the game out of forced sequences and into manoeuvring, where the value of each piece is measured in squares controlled rather than in moves memorised. This is pitch geometry with 64 squares: no passes here, only diagonals opening and closing.

The buyer the product targets is legible enough: an improving club player with thin preparation time, using the King's Indian as a fighting weapon, losing games to opponents armed with the Bayonet.

But the source document has an editorial crack. It stops the move sequence at 7.O-O and jumps straight to 9.b4, skipping 7...Nc6 and 8.d5 Ne7. Without those two moves, the Bayonet does not exist. That is the flaw of a text written to persuade a buyer rather than to seat someone at the board. Alongside it sits an empty list: no author, no publication date, no game count, no engine output, no sample games.

The most telling claim is that “computers do not understand these positions”. That is a claim from an older era. Since neural-network evaluation arrived — Stockfish 12 in 2026 — and Leela matured, closed strategic positions are no longer an engine blind spot at the level the marketing implies. The real value of 9...a5 is not that the engine fails to understand it, but that the human does not need to remember much: it trades memory cost for judgement cost. That is a legitimate practical benefit, only mislabelled as an objective improvement.

The difference between those two labels matters. If 9...a5 is objectively stronger than 9...Nh5, the buyer has the right to demand an engine evaluation, an empirical score, a database citation. If it is merely cheaper in preparation, the buyer needs to know one thing: this purchases time, not advantage. And time expires. A rare move holds value only until the opponent catches up; the half-life of surprise is inversely proportional to how often the move appears on the board. The source says nothing about this, though it is the central technical risk of the entire product.

The claim that “objectively the chances are equal” belongs in the unverified drawer. Mar del Plata structures typically give White durable space and a queenside majority plan while Black plays for the initiative. That is a high-variance balance, not symmetry. At engine depth the sentence may be roughly true; stated without evidence, it is an opinion wearing the clothes of data.

Based on my experience following games, the benefit of this approach shifts sharply with rating. Below 2026 Elo, opponents rarely reach a prepared Bayonet, so strategic guidance beats memorised theory. Around 2026–2400 is the peak zone: opponents do prepare, and a rare move pays the largest surprise dividend. Above 2400, the rarity advantage erodes because opponents and their seconds already have the move in their files, and 9...Nh5 reasserts itself as the main line. Vietnam's strongest players, Lê Quang Liêm among them, sit far above that band, where a rare move no longer means much.

The biggest blind spot in the document is the other side of the board. White has at least three natural replies — b5, Ba3, Nd2 — and none is mentioned. A serious opening book spends most of its length on exactly those moves, because that is where the strategy is tested. The DVD spends its length on the appeal of the idea.

There is a second execution risk, and it is a very Vietnamese one: time. A judgement-based approach eats the clock faster than a forcing-line approach. In a 15-minute rapid game with increment, or a blitz tiebreak deciding advancement at an amateur event, Black will have to think harder precisely when there is no time left to think. That asymmetry goes unmentioned, and it is why I always tell women who are new to the game: test ideas like this in slow games first, do not take them into rapid.

And a credibility gap: no author, no title, no sample games from a named grandmaster. The genre norm is author credentials plus sample games. People say girls know nothing about tactics. So I write for them to read, and for them to know that a product without an author's name is not worthless — it is simply untested.

Tactics never lie; people read them wrong. If you intend to buy the 9...a5 idea, test it in a slow game with the clock running: if you spend more minutes than you would on the main line, the surprise is not free. Sometimes, to understand a game, you have to stand in silence longer than others can bear — and in that silence, demand data before you grant belief.