9...a5: A King's Indian DVD, Olympiad Model Games, and the Price of Human Judgement in the Engine Age

On the first floor of an old building in Balewadi, Pune, on a cold December evening, a thirteen-year-old boy sat in front of a laptop. The menu loop on the screen ran twenty seconds long — no music, just a chessboard, and one move being reset over and over. In his notebook, in pencil, he had written: 9...a5. Underneath it, three arrows, running through the middle files towards the queenside.

His coach sat two chairs away. The tea in the mug on the table was going cold and she said nothing. When he restarted the same video for the fourth time, she said: 'Now put it on the board.' He did. He recited twenty-two moves out loud. On move thirteen his opponent played something else, and the boy stopped. The only sound left in the room was the ceiling fan.

My habit is to write the room before the result. Since 2026 every script of mine opens with a scene note, and the analysis is built on top of it. There was no scoreboard in this room, because this story does not have one. This is not a match report. It is a report on what was running inside that laptop — a DVD in the 'Svitlana's Smart Moves' series, presenting Black's 9...a5 against the Bayonet Attack (9.b4) in the Mar del Plata branch of the King's Indian Defence as a practical alternative.

Everything available about this DVD is in the language of product copy. The presenter's full name, her title, her rating, the publisher, the total runtime, the number of model games — none of it is on record in the material available. What is on record is this: a series, an opening, a move, and a claim.

First, the ground it stands on. The Mar del Plata form of the King's Indian 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. Black has conceded the centre and built a dense, withdrawn wall of pieces instead. The plan ahead is clear: Black breaks on the kingside with ...f5, usually accompanied by ...Nh5 and ...Nf4, while White advances on the queenside with c5 and b4. Both sides attack on their own wing, and whoever arrives first wins. In ECO classification this sits in the E97–E99 band.

White's modern main weapon here is the Bayonet Attack, 9.b4. It does not attack Black's plan directly; it puts sand in the foundation. White takes space on the queenside, ties Black's pieces down through b4-b5, and forces Black to decide whether to complete the kingside attack or repair the house first. In my reading, the Bayonet's strength is not its threat but its delay — it pushes Black behind schedule at precisely the moment Black needs the most time.

Black's classical answer is 9...Nh5. The knight goes to the rim, points at the kingside, and opens the path to ...f5 and ...Nf4. This is the kind of chess in which every move is calculated — White plays something concrete, Black plays something concrete, and the outcome usually goes to the side that spent more hours in front of an engine screen.

This is where the product makes its move: 9...a5. Black pushes a pawn towards the queenside, applies direct pressure on b4, and leaves his own house slightly more open. The pitch arrives in three parts — the move is objectively equal, it is practically dangerous for White, and it is for players who prefer thinking to memorising.

I have to lay out that chain of reasoning, because the product copy builds it in full and its weakness is hidden inside the same sequence. Step one: the Mar del Plata produces rich, strategic positions and genuine mutual attack. Step two: computers are weak in such positions. Step three: therefore human players benefit. Step four: but the Bayonet spoils that beauty. Step five: and after 9...Nh5 the game collapses back into forced, computer-friendly lines. Step six: so play 9...a5 and stay in the human domain.

The logic is internally coherent, but its second step is the weakest link today. 'Computers cannot handle these closed, strategic positions' was true between roughly 2026 and 2026, not after 2026. Since the arrival of NNUE-evaluation engines, engine handling of locked structures and slow manoeuvring positions has improved dramatically. No 2600-level player now believes an engine loses its way in a ...f5 and ...Nf4 position. The product is selling an older, honest insight without contemporary evidence for it.

More striking still, the copy collides with itself. On one page computers are said to be weak in these positions; on another it concedes that after 9...Nh5 'computer analysis becomes important again.' The second sentence cancels the first. The truth is that 9...a5 is not an island the engine cannot reach; it is a move whose theoretical depth has not yet been fully mapped — and that is a statement about the limits of current coverage, not about evading the engine. The distinction is small, and the entire marketing rests on it.

The third point matters most to a buyer: this product is not a Mar del Plata solution; it is one branch of a large tree. At move nine White also has 9.Ne1 and 9.Nd2, both still popular, and at move eight there are 8.Be3 and 8.Re1. A Black player who adopts the Mar del Plata still needs separate answers to all of them. An anti-Bayonet cure and a complete Mar del Plata repertoire are two different things, and product copy can make them look like one. That is not a flaw in the product; it is a flaw in the packaging.

The fourth point is an old device in chess publishing: 'objectively equal, but practically dangerous for White.' The pairing calms two buyer fears at once — 'I hope I haven't bought a bad line' and 'I hope there is something in this for me.' Both statements can be true. But this description contains no engine evaluation, no game counts, no results split, no named model game. Without evidence, the pairing is reassurance, not analysis.

The fifth point touches my own habit. I have called any number of Olympiad rounds, and I like the stage — deep field, real stakes, games played in the open. The Olympiad is a sound sourcing choice for model games, but it carries a hidden selection bias the buyer should know about. In team matches a draw is sometimes in the team's interest, and the intensity of play is not always what it would be in a Candidates or a pure round-robin. A repertoire built mainly on team-event games should be cross-checked against individual elite events.

And the sixth point is the one that stings. The title says 'King's Indian Attack.' The move order given describes the King's Indian Defence. The King's Indian Attack is a completely different system — 1.Nf3, g3, Bg2, d3 — in which White builds the same wall Black builds in the Mar del Plata, with the colours reversed. It looks like a small slip, and in a search-driven market it is expensive: it attracts the customer who plays the King's Indian Attack and repels the one who plays the King's Indian Defence. Neither buyer will be satisfied by that title.

Now the thing I believe most and the copy never says. The product promises fewer concrete lines, more strategy, less memorisation. That sounds like relief. But a position with no forcing sequence is exactly where judgement is taxed hardest. After 9...a5 there is no order of moves that tells you what the next six moves are. You decide each one yourself, and the clock runs at the same speed as before. For a less experienced player that is not relief but load — the position where nothing is pre-decided is where small errors accumulate and then wash away in a single move. The product promises less calculation and, on another page, acknowledges that these positions demand long-term planning and tactical sharpness. Those two sentences cannot sit on the same page.

The second thing no opening product ever tells its buyer: publishing a cure begins to end the cure. A move sold as 'the answer to the Bayonet' invites White to come prepared against it. The selling point is said to be 'few concrete lines,' and that is precisely the advantage that dies first, because the moment the move becomes public it becomes a reason for White to sit down and prepare.

Look further out, though. The real story is not the move; the real story is that a product like this sells. The engine has pushed chess towards a world where every position has a best answer and the skill lies in finding it. Inside that world a market persists which says: give me the territory where the best answer is not yet settled. That is not chess's weakness; it is chess's need. A game that never knelt to the engine has kept a space for human judgement, and people pay for that space.

One memory keeps returning through all of this. Kolkata, a December evening in 2026, Anand and Carlsen four feet apart, and the hall going silent before a move. There was no DVD playing on any screen there. But the question was the same as in that Balewadi room — when the memorised line runs out, what is left? Anand's five world titles and Gukesh's 7.5–6.5 over Ding Liren in Singapore in December 2026 are both answers to that question, and both routes run through the same Tamil Nadu relay I keep writing about. In that relay the coaches sit in the dark, and the result arrives in the light.

I do not have that boy's notebook. I do remember the moment he closed the laptop and looked at the board. His coach said nothing; she simply picked up a pawn and placed it on a5, as if to say — this is yours now. The product did its job: it gave him the move. Thinking the move through is still pending, and there is no DVD for that work.