9...a5 Against the Bayonet: A DVD, an Open Board, and the Numbers That Aren't There

A Tuesday night at a club in Brooklyn. Under the rating list pinned to the wall, a fifteen-year-old sits down across from me and says the Bayonet frightens him. He has Black in front of him, his opponent has already played 9.b4, and he has eleven minutes on the clock. A grey-haired member standing nearby says, without pausing, Nh5, that is the main move. The boy nods, and in his eyes is the thing I have seen in club rooms for thirty years: he remembered the move, not the week that follows it. I asked him what he actually wanted to play, and he said: something I don't have to memorise.

That same night I opened the listing for a new DVD — Svitlana's Smart Moves, the Mar del Plata branch of the King's Indian, 9...a5 against the Bayonet. The promise was clean: fewer concrete lines, strategy over recall, release from the wage of memorisation. And the listing was missing everything I went looking for: how many model games, how many hours of video, which Olympiad, which player, which engine assessment. Brooklyn had frozen that final years ago, but the notebook kept breathing. I scrolled down and found that the title says King's Indian Attack while the moves inside are King's Indian Defence. Two different opening systems, one cover.

The argument cannot be judged without the Mar del Plata. In the King's Indian Defence, Black concedes the centre early — 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 — then strikes with ...e5 and launches a pawn storm on the kingside. After 7...Nc6 it becomes the Mar del Plata that has troubled White for six decades: Black goes ...Ne7 and ...f5, White takes space on the queenside, and the whole game turns into an accounting of two attacks running in opposite directions. Standard classification places it in the E97–E99 band. White's modern main weapon is exactly the Bayonet, 9.b4: the pawn advances to slow ...f5, pressurises along the c-file, and trades structural beauty for initiative. Black's normal reply is 9...Nh5 — the DVD's own description concedes this. Which means the product being sold is not the main road but the lane beside it.

The sales argument is textbook and tidy: the Mar del Plata produces rich, long-strategy positions; computers mishandle them; therefore the human player gains; but the fashionable Bayonet disrupts that comfort; and the main reply 9...Nh5 returns the game to the forced lines where engines are strongest; therefore take 9...a5 and stay inside human territory. The chain is internally coherent, and there is a genuine tension underneath it — the tension between computer-checkable forcing play and opaque manoeuvring.

The problem starts at the weakest link: the claim that computers do not understand these positions belongs to the pre-2026 engine. Neural-network evaluation (NNUE) improved engine handling of closed, slow, manoeuvring positions to the point where "the engine is blind here" is a convenient memory rather than a technical fact. And the description cuts itself within a paragraph: one passage says computers cannot handle these positions, another says that after 9...Nh5 "computer analysis becomes important again." The second sentence does not soften the first, it cancels it. What is really being sold is not engine incapacity but an escape route from the engine — a more honest and much less romantic pitch.

The second ingredient is an old and effective marketing move: "objectively the position is equal" plus "it is dangerous for White." Both can be true at once; equality does not mean ease. But behind neither claim is a single number — no engine evaluation, no database frequency, no results split, no named model game. The buyer receives two promises, reassurance and upside, with no evidential bridge between them.

Scope is the most overlooked fact: this product answers one White branch, not the whole Mar del Plata. Beyond 9.b4, White retains fully independent and still popular routes in 9.Ne1 and 9.Nd2, and earlier deviations in 8.Be3 and 8.Re1. A Black player who adopts the Mar del Plata needs separate answers to all of them. This product opens one node of a large tree. That is not a flaw in the product; it is a flaw in how it is marketed — an anti-Bayonet module is not a Mar del Plata repertoire, and buyers routinely purchase the first believing they bought the second.

Drawing model games from the Olympiad is sound in principle. The field is deep, the stakes are high, the games are public — cheap and prestigious raw material. But there is a selection bias hidden inside it. In a team match, a draw is sometimes sufficient for the team, and play in that climate becomes less critical and more result-oriented. A repertoire built mainly on team-event games should be cross-checked against individual elite events. In the listing, "at the Olympiad" is a prestige label: no edition, no year, not one named game.

The format's lifespan belongs in the same accounting. Static media — a DVD, fixed video, a file that does not update — stands in a market of subscription, continuously updated, database-linked courses. A published antidote carries a second cost: the moment 9...a5 becomes popular, White can prepare against it specifically, and the "few concrete lines" advantage is the first thing to wear away. That is the fate of every weapon printed in an open book.

And yet the product has one real strength, and it lies not in theory but in segmentation: players who want to understand rather than memorise. That demand genuinely exists and the market leaves it badly orphaned. Most repertoire products serve lines; very few teach how to think inside the line. The boy in Brooklyn did not need a move, he needed a map for the following week.

This is where the ledger opens, because sentiment always bills you later. I trust the pause more than the punch line, and the pause in this listing is the absence: no average centipawn loss, no engine match rate, no game count, no results percentage. A product whose entire value proposition rests on "this line holds up" should at minimum state how many games it examined and on what basis it drew the conclusion. I packed forty pages for Riyadh and learned there to listen between moves — a stack of paper is not preparation; preparation is whatever you can be audited on.

The counterintuitive conclusion is that "few concrete lines" is not a cheaper purchase for the buyer, it is a more expensive one. In a forced line, the error belongs to theory and you can turn back to the book. In an opaque strategic position, every move is your decision, the clock is your bill, and the loss is your burden. The description itself says these positions demand long-term planning and tactical sharpness — the very sentence that undercuts the promise of ease. A product that advertises convenience and delivers difficulty gets its first backlash not in reviews but at board four of a club night.

The title error is the small, telling part of the same story. The King's Indian Attack is a separate system — White's 1.Nf3, g3, Bg2, d3 construction — while the subject here is the King's Indian Defence, in which White plays d4. In a search-driven market this is more than a slip of words: it calls in the buyer who does not want the product and hides it from the buyer who does. If a catalogue misnames its own subject, how much trust does the rest of the editing deserve?

The picture is this: a practical answer to the most fashionable branch of the Bayonet, built on reasonable chess reasoning, supported by not one number; a narrow scope that damages the buyer if he misreads it and helps him if it is stated plainly; and an audience that is genuinely hungry and that this product recognises more honestly than most of its competitors. The verdict for me will come not from the listing but from the database. The day 9...a5 appears in a 2650-plus classical game and holds, the move will have earned its place without a price tag. Until that day it is a hypothesis with a label on it. The question is not how much the buyer pays; the question is whether he is buying a move or a method — and the listing does not price those two things separately.