9...a5, the Mar del Plata and the Chess Player's War with Memory

The café on Nguyen Chi Thanh Street stays open until eleven at night. It was August, Da Nang was raining, and the boy sitting across from me was seventeen, with three windows open on an old laptop: an online board, an opening database, and a text file packed with tiny lines he called his "notebook".

He plays the King's Indian Defence. On move nine his opponent pushed the pawn to b4 — the Bayonet Attack. On screen his clock dropped from four minutes twelve to three minutes fifty-eight while he opened the text file, scrolled down, scrolled up, and scrolled down again. His finger stopped on a highlighted line, but his eyes had already slid elsewhere.

"I forgot move eleven, uncle."

I asked him again: do you know what you are actually playing? He looked at me as if I had asked the question in a foreign language. Then he said, with a honesty that stung: "I know the move. I don't know why."

Forty years ago, when I was a junior chess player in Chennai and then moved into tournament organising, I thought technique was everything. Now I sit in a damp café, watching a child who holds more information than a 1980s research institute, and I realise he is losing for a reason nobody taught him how to name.

That is why I stopped for a long time in front of a chess instructional product recently launched on the international market: a lecture series titled "Svitlana's Smart Moves – King's Indian Attack at the Olympiad". A video or disc product, led by a female player, promising King's Indian Defence players an answer to the Bayonet Attack. The title sounds impressive. But when I read the description closely, I found the story behind it far more interesting than the move itself.

I thought I was reading an opening report. It turned out I was reading an indictment of memory.

Context: an opening named after a city on the Atlantic

The King's Indian Defence is one of the strangest systems chess has produced. Black deliberately lets White build a huge pawn centre, then constructs a counterattack from a distance. Black plays Nf6, g6, Bg7, d6, then castles and plays e5 — a slanted wall, not a fortress. White has space. Black has structure. And the real war happens on two wings far apart, like two men standing on either side of a field, both setting fires.

The Mar del Plata Variation is the most ferocious version of that system. Its name comes from a tournament held in the Argentine city of Mar del Plata in 2026, where the structure first appeared at the highest level and forced analysts to rewrite their books. In this variation Black pushes the c-pawn to c5 and expands on the queenside with b5, Rb8 and a6, while White masses pieces on the kingside and plays g4 and h4. Both sides declare war on different fronts and both believe they will arrive first. There is no retreat. Only the question of who miscounts a tempo.

To reach that structure, both sides must pass through a near-forced sequence. White builds the centre with d4, c4, Nc3, e4, Nf3, Be2 and castling. Black answers with Nf6, g6, Bg7, d6, castling, e5. Then White plays d5, pushing the pawn forward, and Black reroutes the knight from c6 to e7. The structure is set. From move nine onward, each side must choose a stance.

And White has a famous option called the Bayonet: pushing the pawn to b4.

The move looks small. It is just a pawn advancing two squares. But as an idea it is a declaration: White refuses to let Black open the queenside freely. The b4 pawn obstructs a comfortable b5, forcing Black to solve a space problem on a wing Black had assumed was home ground. The Bayonet is the move of a player who understands he has more space and does not intend to trade that space for speed.

For decades the Bayonet has been one of White's main options in the Mar del Plata. That is precisely why it annoys Black players: it is not a novelty, it is a move analysed to the floor.

Black has two roads, and both have been walked before

The product description I read offers a very tidy contrast. On one side, 9...Nh5 — the fashionable move, the popular one, forcing the game open, making the lines compulsory and therefore memory-dependent. On the other, 9...a5 — a more classical move, with fewer forcing lines, resting on strategic understanding rather than recall and, according to the presenter, an "antidote" to the Bayonet.

Before arguing about right and wrong, I want to pause on the structure of that argument, because it is very familiar to me. I have heard it in football, in tennis, and now in chess. A new trend appears, takes over the market, forces everyone to memorise. Then someone arrives and says: go back to the older thing, that is where freedom lives.

It is a good argument. It is also an argument that is easy to abuse.

First, 9...Nh5. Black brings the knight to the rim to free the f-pawn, or to control the key dark squares around the centre. The move sharpens the position. The lines become compulsory. When lines are compulsory, the side that remembers more usually wins. That is why 9...Nh5 became the choice of people with time — and with analytical help.

9...a5 is different. Black strikes directly at the b4 pawn that has just advanced. Instead of letting White build the queenside freely, Black asks the question on the spot: what do you intend to do with that pawn. The move does not open the game. It slows it down. It accepts that Black will have less space on the queenside, that Black's pawn structure will be torn in some way, and that in exchange Black gets a clear plan instead of a sequence to remember.

That is a real trade. And here is the point the product description does not state: 9...a5 is not a new move, nor a stronger one. It is a move with a lower maintenance cost — and people pay for maintenance cost, not for strength.

I have spent many evenings over more than thirty years watching chess events to notice one thing: the most heavily advertised moves are rarely the best ones. They are the easiest to sell. A novelty sells on curiosity. An old move repackaged sells on fear — the fear of falling behind.

Here, the product sells that fear very skilfully. It tells the Black player: you do not need to remember eighteen moves of a line your opponent studied three months ago. You only need to understand the position.

That is a true sentence. But it is true in a different way from how it is presented.

The biggest blind spot: the board abandoned mid-sentence

One technical detail in the description made me read it three times.

The move sequence given stops at White's seventh move: castling. Then the story suddenly jumps to 9.b4. Between those two markers the reader is forced to infer two moves that are never stated: Black plays Nc6 on move seven, White pushes d5 on move eight, and Black reroutes the knight to e7.

For someone who has played the King's Indian for twenty years, that gap is harmless. For the buyer, it is a hole. Because the entire value of an opening document lies in being checkable. When you omit the reference position — what the profession calls the tabiya, the point from which analysis is conventionally counted — you have taken away the reader's anchor.

I fell into a similar trap a very long time ago, and I remember it well enough that I wrote it into my first notebook. In 2026, when I was still organising chess events in Da Nang, I taught a group of students by reading variations aloud from books. I read correctly. They wrote correctly. And when they met an opponent who played the moves in a different order, everything collapsed. I had taught them a route without teaching them a map.

I thought I was teaching openings. It turned out I was teaching dictation.

A good instructional document must tell the story of the position, not just list moves. And here, when a commercial product aimed at buyers skips the very position it claims to discuss, I am forced to put a large question mark in the margin.

"Computers often don't understand these positions" — an expired sentence

The introduction uses one argument as a lever: that computers often do not know how to handle complex strategic positions like these, so humans still have room to win through understanding.

I have to say it plainly: that sentence belongs to a period that has passed.

A decade ago the claim still had merit. Closed, plan-based positions with few forcing lines were indeed the area where classical engines evaluated less accurately than in open positions. But the current engine generation, with neural-network evaluation, has changed that picture completely. Closed King's Indian positions, where plans matter more than calculation, are now evaluated at a level the analysts of the 2000s could not have imagined.

So the honest version of the argument has to be phrased differently. Not "computers don't understand". But: "your opponent is unlikely to have memorised a precise, engine-checked path to an advantage here".

That is a completely different sentence. It says nothing about engine capability. It says something about human habits. And it is far more accurate.

I understand why the old version keeps being used. It is emotionally appealing. It tells the amateur something they badly want to hear: that the machine cannot take everything, that there is still a corner of the field where humans are in charge. That is a comforting feeling. But a comforting feeling does not make a move stronger.

In my line of work I have learned an uncomfortable lesson: when a technical argument is delivered in emotional language, it is usually because the technical part is not strong enough to stand on its own.

Two contradictory sentences, and why the contradiction is the best part

One passage in the description made me laugh, in a good way.

At the same time, the document says two things. First: objectively, chances are equal. Second: things might quickly become dangerous for White.

Read separately, both are fine. Read side by side, they open a gap worth thinking about.

If the position is genuinely equal objectively, why is it dangerous for White? The answer lies in a distinction amateur chess usually skips: between objective evaluation and practical scoring probability.

Objective evaluation is what the machine says. Practical scoring probability is what happens on the board when two humans sit down, one prepared and one not, one with forty minutes and one with twelve.

The two are often very far apart. And the entire real value of the product lives in that gap.

I think this is where the copywriter accidentally stated the most interesting truth in the whole piece: this product does not sell an objective advantage. It sells surprise.

And surprise is an entirely legitimate commodity in chess. I have seen it decide bigger games than any engine evaluation.

But surprise has one fatal weakness: it depends on the opponent not knowing. And in an era when anyone can look up a line in four minutes, "the opponent doesn't know" is an increasingly fragile assumption.

A few years ago I covered a game in which a young Vietnamese player chose a defensive option that was, in theory, perfectly sound. He lost in thirty-two moves. Afterwards I asked his coach what had happened. He said something I copied into my notebook: "His opponent had prepared for exactly the option he thought nobody prepared for. The thing nobody prepares for is the easiest thing to predict."

That is the greatest risk of every "play less theory" recommendation. It assumes the other person is as lazy as you are. Usually they are not.

A small problem in the title, and why it matters

The product is called "King's Indian Attack at the Olympiad". But the content described is the King's Indian Defence, Mar del Plata Variation.

Those are different things. The King's Indian Attack is a White system beginning with Nf3, g3, Bg2 and d3, with a small centre controlled from distance. It is a distant relative of the King's Indian Defence, but a separate opening with its own pawn structures, its own plans and its own players.

It may simply be a series label, and this episode may cover something other than the series title suggests. That happens all the time in content publishing.

But for a buyer, the confusion has a cost. A King's Indian Attack player who buys the product because of the title will receive a document about the King's Indian Defence, and vice versa. In a market where most players have a small annual budget for instructional material, that kind of confusion is a real loss.

And there is a second, larger detail: the word "Olympiad" appears in the title and nowhere else. No game is named. No player is named. No year is named. No board is named.

I have spent years reporting on chess Olympiads, and I know the power of that word. A game played at an Olympiad carries a weight a club game does not. It tells the buyer: this has been validated at the highest level, under the greatest pressure.

But for that weight to work, someone has to point to the game. Otherwise the word "Olympiad" is only an echo.

I thought I was reading evidence. It turned out I was reading an echo.

Who actually buys this product

Setting aside the overreach in the advertising, I still think there is a group of players this product serves very well. And that group is not named in a single line.

It is the club-to-semi-professional player, aged thirty to fifty, with a job and a family and perhaps four to six hours a week for chess. They play the King's Indian because they love it — the idea of letting the opponent take the centre and then slowly squeezing, the feeling of playing an adult's opening.

But they are stuck. Every time they meet the Bayonet they have to open a database, and every time they open a database they see a new line they have not studied. They have no analytical assistant. They have no time to re-run every branch after each engine update.

For those players, a lower-forcing option has genuine value regardless of its objective assessment. Because what they are buying is not a stronger move. What they are buying is an evening not spent in front of a computer.

I have met many such people at chess clubs in Da Nang, Hanoi and Ho Chi Minh City. They do not need to become grandmasters. They need a plan good enough that they no longer feel left behind.

And here is what I would say to the people making these products: if those are your customers, say so plainly. Do not promise them an antidote. Promise them a less tiring road. The second promise is smaller, but it will not disappoint them — and in the instructional content business, a customer who is not disappointed is a customer who comes back.

What the product sells, and what it does not

There is a fact about the opening-document market that few people state: the move is free; the explanation of the move is what has a price.

9...a5 is in every database. Anyone can look it up in four minutes and see thousands of games, win-loss ratios and sub-branches. Nobody needs to pay to know the move exists.

What people pay for is the answer to the next question: after playing 9...a5, what should I be thinking.

That is a much harder question. A database does not answer it. An engine does not answer it either, because an engine does not tell you where to place a knight over the next fifteen moves so that you have a plan rather than a chain of reactions.

And this is where I see real value in the product, even while its advertising irritates me. A clear plan for Black in the structure after 9...a5 — knowing the queenside will be compressed, knowing pawns will be torn, knowing you must play on open files and weak squares rather than on space — is knowledge no database packages neatly.

But I have to say the reverse as well. If a Black player buys this product, memorises the first fifteen moves of the 9...a5 complex, and then meets a Bayonet player who genuinely understands the structure, he will be at a disadvantage in exactly the type of position the product promised would favour him.

That is the paradox of every "play less theory" recommendation. It only works if you truly understand. If you use it as a way to avoid studying, it will betray you precisely when you need it most.

The life cycle of an opening recommendation

I want to say something about time, because I have lived long enough to watch many opening recommendations born, flourish and fade.

An opening document has two kinds of shelf life. The first is the shelf life of the move: if the move has existed for twenty years and is still played at the top level, it will exist for another twenty. The second is the shelf life of the assessment: if the document says "this is equal" without an engine evaluation attached, that sentence expires very quickly.

9...a5 is the first kind. It is one of the two classical pillars of the Bayonet. It will not disappear.

But "objectively, chances are equal" is the second kind. In classical theory the Bayonet is generally regarded as one of White's principled attempts at an edge. Saying it is equal is a claim that requires data — and this document supplies none. No engine evaluation. No reference game. No statement of which sub-variation of 9...a5 is being recommended, even though those sub-variations diverge sharply in structure.

A sensible buyer would do this: buy the product for the plans, then check the first fifteen to twenty moves against a current engine and database. The product is a plan library. The engine is a verification library. Neither replaces the other.

What I thought, looking back at the boy in the café

Hưng, the seventeen-year-old, does not have a memory problem. He has an excellent memory. He remembers hundreds of lines.

His problem is that he is trying to remember something he does not understand.

I sat with him for another forty minutes after that game. I did not teach him a single move. I asked him one question: if you threw the notebook away, what would you do with the queenside. He thought for a long time. Then he said: "I'd push the a-pawn, open the file, and bring the rook."

That was a correct answer. Not a book's answer. The answer of someone who had seen the board.

I thought I was teaching chess. It turned out I was teaching how to look.

And that is exactly what every good instructional product must do — not hand the player more moves, but hand them a way of seeing strong enough that they can generate their own.

In that respect, "Svitlana's Smart Moves" has the right idea: teach plans instead of sequences. But it has wrapped that right idea in promises it cannot prove. And in a market where buyers are increasingly demanding, a right idea wrapped badly will be valued below its true worth.

Looking ahead

I think the chess instructional market is entering a phase I would call the gardener's phase.

Previously, the seed seller won. Sell them a new move, a new line, a secret. Now the seeds are free everywhere. What people need is someone to teach them how to grow.

An instructional product over the next ten years will not win by offering a move nobody knows. It will win by offering a process: here is how you build an opening system, here is how you verify it, here is how you update it every year.

And in that process, the question is no longer "is this move strong" but "how much time does this move cost to maintain".

That is a very different question. It is not exciting. It does not generate good headlines. But it is the question millions of amateur chess players actually need answered.

I thought I was writing about a move. It turned out I was writing about time.

And perhaps that is what I want to leave at the end of this piece, after walking through every branch of the Mar del Plata: in chess, the most expensive thing is not the best move. The most expensive thing is the move you can still remember at eleven at night, after eight hours of work, with four minutes on the clock and your opponent having just pushed the pawn to b4.

If a document helps you be present at that moment with a plan instead of a fear, it is worth its price. But demand honesty from it: it is a plan, not a miracle. And verify the rest with your own head — because that is the one thing no database contains.