Capablanca's 2026 and the King's Indian 9...a5: The Gap Between a Chess Headline and Its Product

At the very top of the page sat one line: Capablanca wins gold in Buenos Aires 2026. Beneath it were roughly three hundred words. In those three hundred words the name Capablanca did not appear once. Neither did Buenos Aires, nor 2026, nor any player. What appeared was an advertisement for an opening DVD — the King's Indian Defence, the Mar del Plata Variation, the Bayonet Attack, and a recommendation of 9...a5 for Black.

My job is to reconcile the headline with the ledger. I have done it for nineteen years — first on a risk desk in Ahmedabad, where the days belonged to football sheets and the nights to an unwanted chess book; now with the same habit, asking for source, timestamp and sample size behind every claim. On this page the reconciliation fails. Elo ratings came into use in 2026; no rating coordinate stands behind a title won in 2026. On the other side, the DVD's world contains no match score, no database win-draw percentage, no engine match rate, no ACPL. What exists is a recommendation and the language wrapped around it.

And yet, and this is the first lesson, a solid chess argument is buried inside the advertisement. The headline was wrong; the reasoning underneath was not entirely hollow. The rulebook had a footnote, and the footnote had a price.

In the classical branch of the King's Indian the move order is fixed: 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. That 9.b4 is the Bayonet Attack. White pushes the pawn to b4, takes space on the queenside and converts Black's central pressure into a permanent structural claim. Black's answer has historically pointed in one direction only: a storm on the king's side. ...f5, then ...f4 and ...g5, and if White errs, ...g4 with the knight swinging back for a direct assault on the king. The whole appeal of Mar del Plata lies in that imbalance — White has space, Black has time and a fast attack.

Inside that structure Black faces two roads, and the DVD stands on them. The first is 9...Nh5, today's main line. The knight goes to h5, supports ...f5 directly, and pulls the game into sharp, forcing branches where the price of every move surfaces a few moves later. The second is 9...a5, which places a brake on the queenside, slows the pace of White's b4-based expansion, and leaves Black to decide when and in what order the kingside storm begins.

The difference between the two is not technical but cognitive, and that is the centre of the whole matter. The real deal between 9...Nh5 and 9...a5 is a trade of memory load against judgement load, not of equality against inequality. In 9...Nh5 you must understand fewer moves but remember them almost verbatim; forget one and the position sours, because the opponent is playing from the same file. In 9...a5 you must remember fewer moves but decide in every position; the punishment is not immediate, it accrues quietly over ten moves.

From here the argument is economics more than chess. The practical value of 9...a5 lies not in being the engine's first choice but in reducing the return on White's preparation. A well-prepared White carries deep files on 9...Nh5 — from club level to grandmaster level that line has been repeated more than any other, so everyone is watching it. On 9...a5 the file is thin. This is a distribution-based argument, not an evaluation-based one. I write the shape of the distribution before the prediction, because the distribution tells you where the opponent is prepared and where he is not.

I learned that lesson in Tbilisi in 2026, when I built my first FIDE World Cup model on classical Elo alone. Levon Aronian took the title in the rapid tiebreaks and the post-mortem showed the model had undervalued rapid specialists by roughly forty rating points. Since then I have never trusted a single input. The 9...a5 recommendation repeats that error if read alone — it is one input, not the whole model.

In London in 2026, for the Carlsen-Caruana world championship, I priced 'all twelve classical games drawn' at 6.8 percent, because three Elo points separated the players and both carried decisive-game rates under 25 percent in elite round-robins. All twelve were drawn. 6.8 percent is not small. It is just quiet. That experience taught me to demand the sample before the claim — and this DVD is being sold without a sample.

Here the second gap opens. No database figure is offered for 9...a5 — no win-draw ratio, no average centipawn loss, no engine match rate, no game citations. 'Roughly equal but dangerous' is an evaluation sentence that requires three things to prove: which engine and which version, what sample of games, and which rating band. All three are missing. When a product does not send its ledger with its claim, the buyer must build that ledger himself — and that ledger is built from his own lost games.

There is a human part to this, and it must be written attached to a number. The club player who chooses this line on the promise of 'understanding' usually plays fast time controls — blitz and rapid, thirty seconds a move. Judgement-heavy positions are most expensive exactly where the amateur loses most points. The DVD's entire premise — think for yourself — is weakest under the clock. That is not the player's weakness; it is the product's design limit.

Now the claim that is the weakest part of the whole advertisement: 'computers often do not know how to handle these complex positions.' That sentence was true — roughly before 2026. With the arrival of neural-network engines (NNUE-class Stockfish and Leela-style networks), evaluation of closed King's Indian structures became far more reliable. To say today that 'engines get lost here' is marketing from an older era, not technical fact. What remains true is narrower: in human-preparation terms the benefit of 9...a5 is real, because the opponent's file is thin. Calling it an antidote is a different matter from proving it.

The second uncomfortable point is that low forcing cuts both ways. A position with few concrete lines also offers Black fewer winning chances when a strategically competent White sits across the board. White can keep the queenside closed, play on space, and strip away the dynamic compensation without which the King's Indian framework is hollow. In other words, this line is good against players who play from the book and bad against players who read the board. No product tells you which category its buyer will fall into.

The third thing every opening product hides: the decay of practical value. The moment 9...a5 enters an opponent's file, its original reason — the preparation asymmetry — is gone. It is an asset that wears with use, and the DVD's price does not account for the wear.

My own habit is simple. I keep the log so the log can keep me honest. In 2026 the Online Olympiad final collapsed into disconnections and was declared a joint title, and that week I did not argue — I pulled connection logs and clock-loss records from all twelve boards and reviewed them game by game across four days. The output was a rule: online chess never gets full weight. That same year I re-graded 41,000 archived games into my own database, because old data asks questions too. No opening claim passes me unless it carries version, sample and condition.

The spreadsheet does not flinch when the crowd roars. This DVD guarantees nobody a win and should not; it sells a line that asks less of your memory and less of your pocket, but more of your judgement. The player who understands that bargain is not cheated. The one who buys it as 'Capablanca's legacy' is paying for a byline that was never on that page.

In the months ahead, whenever an opening product sells 'a way around the engine', ask three questions: which engine and which version, what sample of games, and in which rating band the line's performance was measured. Take seriously the product that answers all three — even if its headline is not about Capablanca. And the one that does not, put in its real place: an advertisement whose footnote is still to be written.