The 9...a5 Antidote to the Bayonet Attack: The Blind Spot Named 'No Memorisation'

Mar del Plata, Argentina, summer of 2026. A tournament on the Atlantic shore, and a variation of the King's Indian Defence has carried the name of that city ever since. Black yields the centre, then closes it. Knight to c6, pawn to e5, knight back to e7, then the f-pawn tears the kingside open. There is nothing romantic in it. Only calculation, and a belief that defending is a slower way of attacking.

Seventy years later, I sit in a room in Saint Petersburg and open a text file a colleague sent through the internal feed. The headline is explicit: a second-round report from a team event held in Samarkand, Uzbekistan, with a note that despite a few close calls every strong team grabbed a win. The body contains no Samarkand. No teams. No pairings. No results. Only a promotional blurb for a chess DVD, championing Black's ninth move 9...a5 as an antidote to the Bayonet Attack.

I read it twice. Then I understood that what I was holding was not a broken article. It was a specimen. A specimen of the chess-theory information market, where headlines and bodies routinely stop belonging to each other.

And at the centre of that specimen sits one move worth taking seriously: 9...a5.

To understand why a ninth move needs an entire file to explain, we have to go back to the structure.

The King's Indian Defence begins with 1.d4 Nf6 2.c4 g6, then bishop to g7, pawn to d6, short castling. White builds a centre with e4. Black does not contest the centre directly; Black lets White take it and then attacks it from both flanks. That is the hypermodern spirit of the 1920s, and it is still intact a century later.

The Mar del Plata variation is the sharpest branch of that system: 6.Be2 e5 7.0-0 Nc6 8.d5 Ne7. White pushes the pawn to d5, locking the centre like closing a door. Black will play f5, open the kingside, and both sides will hurl themselves at opposite wings — White on the queenside, Black on the kingside. The ECO codes for this branch sit between E97 and E99.

The Bayonet Attack is White's ninth-move answer: 9.b4. White pushes the b-pawn, opens a path for the bishop to b2, prepares c5 and builds a front on the queenside. This is the system that has cost more than a few King's Indian players a great deal of sleep over the past two decades. It has a clear target, and that target is Black's queenside.

Black's main reply has long been 9...Nh5. The knight hops to the rim, threatens to jump to f4, prepares f5 and turns the game into a calculation race with no room for hesitation. That is the mainstream road. Books, databases and elite training sessions all run along it.

And 9...a5 is the side road. A known branch, not an invention. Black pushes the a-pawn, attacks the b4-pawn head-on, and forces White to decide about the queenside structure before Black opens the kingside.

That is the technical foundation. The rest is market.

The product advertised in that file has the markings of a ChessBase-style Fritztrainer: a DVD, an analyst, a promise. The promise is packaged in three claims — a sound position with very few concrete lines, a style for players who prefer thinking for themselves over memorising, and a warning that if Black knows what he is doing, things may quickly become dangerous for White.

Those three claims sound familiar. They are the language of a specific market: the market that sells answers to fashionable opening lines.

The transfer market is where people sell the future to buy hope. In chess, that market does not close in January and July. It stays open all year, and its currency is the move.

So what does 9...a5 actually do?

It must be said at once that the source material offers not a single sample game. No engine evaluation, no average centipawn loss, no win or draw rate, no author name, no time control. I stress this because in my trade a technical claim without data is a hypothesis, not a conclusion. Readers deserve to know whether they are reading a hypothesis or reading evidence.

But even without data, the structure of the idea can be dissected.

After 9.b4, White has just committed the b-pawn to a square Black can attack with a5. The move 9...a5 asks a direct question: will White take, or push on?

If White captures on a5, the queenside structure opens and Black gains an open file plus a clear target for bishop and rook. If White pushes to b5, Black can change direction, play a4 and then c5, hitting the centre from behind. Both routes lead to positions where short-term calculation matters less than the long plan.

Compared with 9...Nh5, the difference lies in the character of the game. 9...Nh5 opens the kingside and creates forcing lines where one wrong move can be punished instantly. 9...a5 postpones the kingside war and makes the queenside the first front.

The promotional text frames that difference as a pair of opposites: on one side, computer analysis becoming important again; on the other, sound positions with very few concrete lines. That framing deserves a longer pause.

It assumes two kinds of positions exist: the kind engines understand, and the kind engines do not. The second kind is presented as human territory, and therefore worth buying.

This is the weakest point in the entire marketing argument. The claim that computers often do not know how to handle complex strategic positions is a leftover from the pre-AlphaZero era, and in the neural-network age it is technically obsolete.

I remember that period clearly. Before 2026, search-based engines genuinely had blind spots in positions whose value lay in pawn structure and weak squares rather than material. Back then, saying an engine did not understand was correct. Professionals used it to talk about good and bad bishops, open files, positions that static evaluation could not capture.

That era is over. Neural-network engines today play the long-term positions of the Mar del Plata better than most human players. Moves that a decade ago were dismissed as no engine would choose now appear regularly in elite games, simply because engines chose them first.

So when a product advertises itself with the argument that machines do not understand, that is a marketing message aimed at a particular buyer: someone who wants to believe that human reward lies where machines are weak. The belief is comfortable. It is also no longer true.

None of this means 9...a5 is a bad move. It only means the reason people are invited to buy it is not a technical reason.

Beneath the marketing, the substantive claim is far more modest. The text says chances are equal. In the Mar del Plata, equal means both sides have complex play, both have chances, and both can go wrong. That is a claim about a position, not a claim about theoretical superiority.

No novelty is claimed here. No new move. Only a known branch, packaged more carefully, retold in a more persuasive voice.

And that, in the end, is this entire industry. An old move, a new cover.

Fairness to the product side is due. A side branch has its reason to exist.

In opening theory, the value of a move does not lie in whether it is the main branch. It lies in whether your opponent has prepared for it. A player who has spent hundreds of hours on 9...Nh5 can be utterly surprised by 9...a5, and in a game with a tight time control that surprise is worth more than a few percentage points of engine evaluation.

I have watched enough tournaments to know that side moves often win not because they are stronger, but because they arrive when the opponent least expects them. My experience from the years when I organised events and recorded games tells me something simple: in modern chess, the psychological weight of preparation has its own mass, and that mass cannot be measured in centipawns.

The problem lies elsewhere.

The more widely a side branch is advertised, the faster it becomes a main branch. A move sold to thousands of buyers will appear in thousands of club games, junior events and online sessions. White's specialists will review it, find an answer, and within one theory cycle — months to a year — the surprise is gone.

That is the biggest risk buyers are not warned about. The DVD promises a weapon. But a weapon is strong only while it remains a secret. Once it is printed and sold, it becomes a target. The last buyer in that chain receives the least value, because by then everyone knows the move.

History repeats this pattern. A side branch appears, causes surprise, wins a few notable games, gets written up, and is neutralised by a prepared answer. David Bronstein was one of the figures who brought the King's Indian Defence into major tournaments in the mid-twentieth century, and the system has run through this exact cycle many times over the seventy years that followed.

The blind spot in the collective memory here lies elsewhere, and it is subtler.

When amateurs — even fairly strong ones — look at a King's Indian board, their memory is dominated by legendary games of the twentieth century. They remember the Mar del Plata as a romantic war: two wings exploding, kings torn open, both sides defying danger. It is a beautiful memory. It stopped reflecting reality long ago.

In today's elite events, the King's Indian is no longer a main weapon. It appears as a surprise option, used once, then returned to storage. The reason is simple: the system concedes space and invites the opponent into a fight where computer theory is far ahead of human theory. A player who prepares at home with an engine can enter the game with an advantage that comes not from talent but from data.

Garry Kasparov repeatedly used Indian systems as part of his arsenal. Magnus Carlsen, over more than a decade at the top, has hardly ever used the King's Indian Defence as a principal weapon. The distance between those two eras is not a distance between people. It is a distance between tools.

That decline is not bad news. It is simply a fact that needs stating, because it decides the true value of the product being sold.

And it creates a paradox: if the King's Indian is narrowing at the elite level, the real value of a DVD teaching 9...a5 does not lie there. It lies at club level, among players who want an unusual weapon to make life difficult for opponents in amateur events. That is a real market, with real buyers and real demand. It is simply not the market the advertising language is aiming at.

The second point needs saying plainly: the promise of very few concrete lines is almost certainly overstated.

The Mar del Plata is one of the sharpest structures in the entire opening canon. Its essence is a calculation race on opposite wings. No ninth move erases that essence. 9...a5 can change the order of priorities, can postpone the kingside for a few tempi, can produce positions where the plan matters more than a forcing line — but it does not turn the Mar del Plata into a slow strategic struggle of Nimzo-Indian character. Calculation is still there, waiting its turn.

An amateur who buys the DVD expecting no memorisation will discover within months that memorisation is still required, merely in a different shape. Expectation drifts from reality, and that drift is typical of every opening product on the market.

Now comes the part I consider most important, and it has little to do with chess.

The file in my hands carries a Samarkand headline and a Mar del Plata body. The two have nothing to do with each other. If I were a young and hurried editor, I would take that headline, attach it to an article about a team event in Uzbekistan, and publish. Nobody checks. Nobody objects. The site still has a post and the metrics still have a pageview.

That is the real risk of the theory information market. Not wrong moves. Texts that no longer know what they are about.

In my trade this error is becoming more common. Automated aggregation platforms gather headlines, descriptions, advertisements and comments into one place. The result is files like the one I am reading: a team-sport headline, an opening-theory body, and nothing linking them except an algorithm.

For readers who want to learn chess, the consequences are concrete. They lose time. They trust a claim with no data. They pay for a side branch presented as an invention. And most importantly, they are never told who the author is, which games that author has played, what results that author has achieved in the very system being taught.

That is the first question I always ask when reading an opening product: who did the person teaching me this move beat with it?

The file does not answer.

Luzhniki night, and the old bird understands that landing is also a way of flying.

I wrote that line about a football match, but it applies to chess in exactly the same way. 9...a5 is a formal retreat: the a-pawn leaves its starting square only to attack from the flank, seizing no centre, opening no kingside yet. To a beginner it looks like a small move. To someone who reads structure, it is a controlled landing in order to fly on the next tempo.

The strength of a retreating move lies in the fact that opponents rarely prepare for it. That is its entire value. And it is also its entire weakness, because once prepared for, opponents are no longer surprised.

When the stadium no longer holds a ball, memory starts scoring. When a chess position is pressed onto a disc, theory starts selling. Between those two sentences there is a gap, and the writer's job is to stand inside that gap rather than on one side of it.

The question I want to leave behind is not whether 9...a5 is stronger than 9...Nh5.

The question is whether the value of a side move lies on the board or on the cover of a disc.

From nearly fifty years of watching and recording, my answer is that most of the value lies on the cover. A side branch gains its edge not because it is better than the main line, but because it is rarer. That rarity is an asset that can be diluted, and every time a DVD sells well, it is diluted a little further.

If you are a club player looking for a weapon to collect points in amateur events, 9...a5 may be worth a few evenings of study. If you are an elite player looking for a theoretical edge, that move is not for you — what you need is an updated database file and a current engine, not a DVD.

And if you are a news reader, the one thing I ask is that you look at the headline before you trust the body. A headline about Samarkand, a body about Mar del Plata, and in between a reader slowly losing the ability to tell what is worth believing.