The Bayonet Attack, 9...a5, and the Price of an Unprescribed Antidote

1:47 in the morning.

I open the file. Fourteen information points. Not a single game cited. Not one engine evaluation. No author's name, no date. Just one promise repeated in different shapes: the move 9...a5 will pull you out of the lethal theoretical thicket of the Bayonet Attack in the King's Indian Defense.

I used to trust emotion, until a number knocked on my door at 3 a.m. Tonight it knocked two hours early, and it was not a pretty figure. It was a void — a void sitting exactly where the evidence should have been.

The document's headline reads: "Samarkand R2 (open): Despite close calls, all top teams grab wins." That is a round-two report from a team event, most likely held in Samarkand, Uzbekistan — a country with a serious chess tradition. The body of the document is about an instructional DVD on an opening. The two texts have nothing to do with each other. They were stapled onto the same page.

That was the first thing I marked in red. Not because it is interesting, but because it tells me that anyone citing this document — including me — is standing on unverifiable ground.

But before throwing the whole thing in the bin, I want to ask myself the question I always ask when I sit down in front of a fresh dataset: if all I had were the board and the numbers, what would I see?

---

Where the real board sits

The King's Indian Defense — KID to anyone who plays it — is one of the most stubborn opening systems in modern chess. Black concedes the centre from the very first moves, builds a fortress around the king, then strikes back with a counterattack on the queenside. It is a wager on time: you hand your opponent space in exchange for a solid structure and a clear plan.

The Mar del Plata variation pushes that wager to its extreme. Black pushes the pawn to e5, places the knight on c6, and both sides launch a race on opposite wings: White storms the queenside, Black storms the kingside. Whoever arrives first wins. There is no fallback. This is the kind of position where one wrong move can become a loss inside fifteen moves.

Then comes the Bayonet Attack.

The Bayonet Attack is White's answer 9.b4 — a move that is more conceptual than tactical. White is not trying to win with a single clever blow; White expands on the queenside, drives the black knight back, and turns a two-wing race into a race where White has more time. For more than two decades, this has been the weapon that made plenty of Black players strike the KID off their repertoire entirely.

Black's mainstream answer is 9...Nh5 — pushing the knight to the rim, threatening exchanges, opening files, creating forced lines. That is the main road, dissected move by move, to the point where at elite level both players often sit down with the first twenty-five moves already packed in their heads. Players such as Teimour Radjabov, Judit Polgar and Boris Gelfand have repeatedly used the KID as a primary weapon, and they are the ones who have paid the price of preparing this system down to the bone.

And this is where the document enters. It proposes a different road: 9...a5.

Before going further, I need to be clear about the market this product is selling into.

Opening-theory publishing is one of the most stable and least glamorous revenue streams in chess. It does not sell glory. It sells reassurance. The buyers are club players, junior coaches, amateurs rated between 1500 and 2200 — people who spend thirty minutes each evening learning a new system, hoping that what they bought will give them back time and confidence at the board.

That is why products like this are always packaged with two words: "the antidote." An antidote to a fashionable opening. A clean answer to a dirty problem. And the problem with the document in front of me sits exactly there: it sells an antidote but never hands over the prescription.

---

Fourteen information points, and what they leave out

Let us walk through what the document actually provides, because that is the entire dataset we have.

Point eleven states plainly: 9...a5 is a known branch, not a novelty. Point nine confirms 9...Nh5 remains the main move against the Bayonet. So where does the product's value lie?

It lies in the promise of a change in the character of the game. After 9...Nh5, the position opens, the lines become forcing, and computer analysis matters again. After 9...a5, the marketing says, you enter sound positions with very few concrete lines.

Split that promise into two halves and weigh each one.

The first half: "very few concrete lines." This is a testable claim. You open a database, count the reasonable moves within the first ten moves of each branch, and put the two numbers side by side. But the document supplies no number. That does not make it false. It only means it has not been demonstrated — and in the opening market, a product that cannot demonstrate its own difference is selling belief, not theory.

The second half: "sound positions," "equal chances." This is a much larger claim. Remember one thing: in the Mar del Plata, even on the 9...a5 branch, Black still concedes the centre and still has to withstand an attack on the kingside. Positional equality does not mean comfort. No branch of the Mar del Plata removes calculation. The Bayonet is a sharp weapon — that is true. But 9...a5 does not blunt it. It only changes the angle of the grind.

Point five is the line that stopped me longest: "computers often do not know how to handle the arising complex strategic positions."

That sentence was true in 2026. It is false in 2026.

Here I need to be blunt, because this concerns the credibility of an entire field. The pre-AlphaZero generation of chess engines evaluated positions by adding material, adding space, adding pawn-structure bonuses, then searching a finite horizon for the optimal move. In long-term KID positions — where a reasonable move today only becomes good twenty moves later — that generation genuinely struggled.

But neural-network engines have changed the picture completely. They are trained to assess the position, not merely to count pieces. They play closed positions, manoeuvring positions, structures the previous generation called murky, with a precision no human player can match. Saying "computers don't understand" these positions in 2026 is dated language kept alive because it sells.

It is an old form of marketing: selling the feeling that humans retain a private territory machines cannot touch. It feels good. It also is no longer true.

Point six offers a different, subtler signal: the product targets "players who like to think for themselves, not memorizers."

Note the structure of that sentence. It does not tell you how strong 9...a5 is. It tells you who you are if you buy it. This is identity-based selling, a classic and effective technique: rather than convincing you the product is good, convince you that the buyer of the product is a smarter kind of person.

And it works. Because most amateur players hate memorizing. They hate the feeling that their opponent has more preparation time. They hate losing in the first twenty moves to a line they have never read. A product that promises to free them from that pain will always find buyers, regardless of whether the moves inside are strong or weak.

---

The risk is not in the move

Here I want to put a balance scale next to a spreadsheet.

With any opening product there are three kinds of risk. The first is theoretical risk: the branch being sold can be neutralized by the opponent. 9...a5 is a known secondary branch, meaning it has already been analysed. Any well-prepared White player can open a database and find ten ideas against it. An antidote that has been countered is no longer an antidote.

The second is expectation risk. The buyer pays because they believe they will escape calculation. Then they sit down, play 9...a5, and discover they are still inside a Mar del Plata game — still calculating, still deciding, still under pressure on the kingside. That gap between the promise and the experience is what drives people out of chess.

The third is economic risk. This product competes directly with free databases and free analysis tools. Its real value is not in the lines — anyone can look those up — but in the way the story is arranged. That is a legitimate business model, but it needs to be stated honestly.

The last point, and the heaviest: there is no author's name. Fourteen information points, and not one line tells you who analysed this branch. In chess, author credibility is not decoration. It is data. A player who has played the KID at a high level has the right to speak about the KID. Someone who has never handled the black pieces in this system is selling a summary.

When there is no name, you are not buying knowledge. You are buying a voice.

And there is one more layer to the problem, sitting in the mismatched headline itself. In journalism, a headline about a Samarkand team event stitched onto a body about an opening DVD is not a small error. It is a sign that the text passed through at least one layer of automation with nobody checking. For a casual reader, that hardly matters. For anyone citing it, it matters absolutely, because a source that is not internally consistent cannot be a source.

---

The twist: what this document gets right

By now I have been fairly critical. And I stand by it: no data, no conclusion. A number that is not cited is not evidence; it is a decorated blank space.

But there is one thing this document gets right, and it gets it right by accident.

It does not promise you will win.

Twice, across fourteen information points, it admits that chances are equal. There is no claim that "White collapses after this move." No claim of a decisive advantage for Black. For a promotional product, that is rare restraint. I have read hundreds of opening advertisements, and most of them promise far more.

I light a candle for data. But I always let the flame of emotion illuminate the question. And this candle is throwing light somewhere other than where I expected.

The problem is not the move 9...a5. That move is legitimate. It exists, people play it, it has its own logic. The problem lies elsewhere entirely: this is not a product about technique. It is a product about people.

What buyers purchase is not a move. They purchase a way out of anxiety.

That anxiety is real. I have sat in that seat — twenty-five years ago, when I still believed I could catch up with opening theory by staying up later. I was wrong. Nobody beats a machine that reads faster than they do. But nobody becomes a better player by trying to do that either.

What this document is really selling, without saying so, is release from the feeling of falling behind. That is a real need, and a real need always finds a supplier. The only question is whether the supplier is honest.

Here, the honesty is split in two. The first half — the restraint about equality — is honest. The second half — the story about machines failing to understand humans — is not. And buyers remember only the second half, because that is the half that makes them feel good about themselves.

When the stadium is empty, the true value of a person begins to speak. Here there was no empty stadium. Only one man in front of a screen at nearly two in the morning, trying to tell the difference between a chess move and a sales pitch.

---

What would tell me I am wrong

I do not want to end with a verdict. Verdicts belong to courts, not to people reading spreadsheets. I want to end with a list of signals I will track over the coming months, signals that will tell me whether I have read this badly.

First: the frequency of 9...a5 in classical games at 2700 Elo and above. If it appears more often within a year, the document was right before I was. If it stays quiet, it is a club-level product and nothing more.

Second: the number of Bayonet Attack games played in open events. If 9.b4 remains fashionable, this product's window stays open. If White players shift to another weapon, the antidote becomes an answer to a question nobody is asking anymore.

Third, and most important to me: whether anyone publishes the name of this product's author alongside that person's own results in the KID. Because between a signed analysis and an anonymous one, the distance is not fame. The distance is accountability.

There are players who get forgotten, but data never forgets them. And there are products remembered for a long time, which data has never once confirmed.

I will reopen this file in six months. If by then 9...a5 is still a footnote in a database, I will know that tonight's lost sleep was not wasted.

And if it appears on the board at an elite event, I will be the first to rewrite this piece. Not to apologize. But to add one more point to the data column.