The Buenos Aires 2026 Headline and the 9...a5 Board: A King's Indian Preparation Report

Last Tuesday a file landed on my desk. The headline at the top read: 'A farewell triumph: Capablanca wins gold in Buenos Aires 2026.' Everything beneath it was a product blurb for an instructional DVD on the King's Indian Defence, Mar del Plata Variation, Bayonet Attack. The blurb contained no Capablanca. No Buenos Aires. No 2026. No player, no score, no time control.

I read the file twice. The first time as a reader, and I got lost. The second time as a sub-editor, with the habit that formed on the Hindu desk in 2026 and has never quite left me: two separate stories have been stitched into one copy, and the stitching is visible.

And that is where I repeated my oldest mistake. I watched the match before I watched the board — that was the mistake. The name Capablanca sent me straight to the 2026 Buenos Aires Olympiad, which really happened, which is on the record, and which ended days before the war began in Europe. I started assembling historical detail when the actual question was simpler: who put a Capablanca headline on top of a DVD blurb, and why.

That is the question every file deserves: who sent this text with this headline, and why.

Now take the packaging off, because the board underneath is real.

The move order 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 9.b4.

Black's seventh move, ...Nc6, is the Mar del Plata, named for the Argentine coastal city where the variation announced itself in the early 1950s. Look at the shape of the position. White's d5 pawn collides with Black's e5 pawn and the centre locks. The centre becomes a padlock. And where the centre locks, the game moves to the wings — the entire economy of the King's Indian rests on that padlock.

Black's old plan runs at the king: ...f5, ...f4, ...g5, ...g4, ...Nh5, then the flood. White's plan runs at the queen: clamp the queenside with b4-b5, keep the c4-c5 break in reserve, and push Black back before he can start the flood.

The ninth move, b4, is the Bayonet Attack. The name carries violence and the violence is earned: this pawn advances like a bayonet so that the queenside is settled before Black's kingside climb begins.

The main answer is 9...Nh5. Why the knight goes to h5 is easy to see on the board: it supports ...f5 and keeps the threat of landing on f4 after ...f4. This is one of the most heavily analysed lines in the entire King's Indian — forcing, concrete, and completely open to the engine's eye.

The DVD recommends 9...a5. That is not the main line. It is legitimate theory, not an invention, but it is not the main line, and that gap is the centre of the whole argument.

Watch what 9...a5 actually does. It stakes a claim on the queenside and prepares ...a4. Once Black plays ...a4, White's b4-b5 loses its bite, because even after reaching b5 the pawn has its support structure cut from behind. Black also keeps ...Rb8 and ...b5 in the drawer, an attempt to surround White on his own ground. The most interesting detail is that 9...a5 does not close off ...Nh5; that knight hop can still come later. The move does not shut a door. It keeps one open and opens another.

But there is a price, and the price should be said plainly. When Black plays ...a5, the kingside storm is postponed. Black's compensation in the King's Indian is dynamic — he pays for less space with an attack on the king. Postpone the storm and the compensation starts to thin. This is the point where theory and marketing take different roads.

Now to the argument that sells the whole recommendation, and which I think is the strongest argument in the file — even though whoever wrote it chose the weakest way to say it.

The argument is preparation economics. In a system whose value comes from a preparation imbalance, the question is not 'which move is best' but 'which move does the opponent have the thinnest file against.' 9...Nh5 is the main line, therefore the most analysed, the most engine-mapped, and any 2400-plus White has twenty moves of prepared file waiting for it. The file on 9...a5 is thinner. That argument is sound and it does not need a single engine number.

One thing needs saying clearly, because it is the most important part of this report. There is not one number anywhere in this material. No ACPL. No engine match rate. No database win-draw-loss percentages. No game cited. When a preparation product claims its recommendation is 'equal but dangerous' and cannot show a single game, the claim is incomplete — even if the move is right. I keep two notebooks: one for numbers, one for names. In this file both are empty. The numbers notebook because no numbers were sent; the names notebook because no player was named.

This is where a line I wrote in my own diary after the 2026 National Premier comes back: 'The number said exhausted; the player said fine. I believed neither.' There is a second edge to that sentence, and it applies here — when the number is absent altogether, you cannot trust the number, and you cannot trust the man making the claim without one.

Now to the claim that is technically weakest and commercially brightest: that computers often do not know how to handle the arising complex strategic positions.

That sentence was true. Roughly a decade ago. In the era before neural-network engines, when Stockfish's strength came from hardware depth, engines genuinely wobbled in closed King's Indian structures where the compensation was long-term and positional. The NNUE-class engines and the Leela-style networks ended that era. These structures are evaluated with real confidence now. A sentence that once read as inside knowledge now reads as nostalgia sold as analysis.

But — and here I will speak for the seller, because fairness is a habit — there is a defensible version of the claim. An engine can evaluate a position correctly and still fail to hand a player a plan. Zero point zero zero means nothing to a club player; what he needs to know is where the rook goes, when ...b5 comes, and when to sit still. An evaluation is a number; a plan is a sequence — and in the King's Indian the sequence costs more than the number. That is where the product's real value sits, not in 'the computer doesn't understand.'

Now the part no seller writes.

The first problem is the time control. The DVD's whole promise is that you should stop memorising and start thinking. That is a beautiful sentence in a ninety-minute classical game. Club chess does not have ninety-minute classical games. It has ten-plus-five, fifteen-plus-ten, and very often blitz. The moment you trade a forcing line for a strategic one, you shift the load from memory to the clock — and the clock is exactly where the club player bleeds points. Preparation for 9...Nh5 hands you fifteen moves on autopilot. 9...a5 asks you for a decision on almost every move. Ask yourself which gift is for whom.

In Kolkata in 2026 I charted Anand's rapid games and found he spent 38 per cent of his clock between moves 14 and 20 — precisely the window where theory ends and planning begins. The same study shows that window collapsing under faster controls, with the player falling back on what he already knows. A preparation product selling itself against memorisation owes the buyer a line about which time control its method survives. That line is missing. — Root: Kolkata 2026 — Anand

The second problem is the shelf life of the argument the sale rests on. A preparation imbalance is a wasting asset. The value of 9...a5 comes from the opponent's file not containing it. The day that line appears on a DVD, the imbalance goes into mass distribution. A repertoire choice that draws its strength from being hidden cannot keep that strength once it is hidden no longer. That is not a fault in the move; it is the limit of the argument being used to sell the move.

The third problem is the one I opened with, and I will not call it a player's error. I will call it a process error. A Capablanca headline sits on top of a King's Indian blurb and nothing bridges them. It may be a technical fault on an aggregated newsletter page where two items merged. It may be a deliberate click-frame. Either way the message that travels is this: a publisher whose headline does not speak to its own copy should have the proofreading checked before the analysis notebook is trusted. I will not name anyone, because the problem is not a person. It is the habit of borrowing a historical name to move a product.

And the historical name is genuinely worth borrowing, which I should also say. The 2026 Buenos Aires Olympiad took place, across August and September, and Capablanca sat at first board for Cuba. Days after the tournament closed, war began in Europe; several teams went home early and many players stayed on in Argentina. The empty hall taught me that silence has a heart rate too. That September the hall's pulse belonged to a radio, not a clock. It is a real story with its own article in it. It just does not attach to this DVD, and stitching the two into one file damages both.

What to watch next, because the report does not end here.

The next time this line appears in a classical tournament, look at three things. First, the rating of the player using it — 2600 or 2100. Second, when White played b5 and when Black played ...a4. Third, and most important, the clock: did Black spend as long thinking after move 20 as he would have in the main line. I have an old habit here — half the data was in the corridor, not on the board. Walk into that corridor and ask people where they learned the line. If the answer is 'a DVD,' you will know exactly how fast the preparation imbalance is evaporating.

We measured everything that season except what it meant to share it. Preparation products make the same error: they measure moves, not transfer.

The question is not whether 9...a5 is a good move. The question is whether, when a move needs a borrowed headline to sell, the analysis is doing the work or the advertising is. And that question will be asked again, with the next DVD, and the next bad stitch.