The Bayonet 'Antidote': Where Is the Data Behind 9...a5?

I stopped twice while reading the promotional copy for a new King's Indian Defence repertoire. The first sentence: 'Computers do not properly understand these closed, long-strategy positions.' The second, a few lines later: 'After 9...Nh5, computer analysis becomes important again.' Both sentences sit on the same page and cancel each other out. If engines cannot read these positions, why does the machine suddenly regain its eyesight the moment play returns to the main line? No answer is offered, because promotional copy carries no obligation to answer. For thirty-seven years my habit with chess claims has been the same: I read the advertisement as a document, not as an argument. A number that is not given sits in my ledger as zero, and zero supports nothing — it only occupies space. Here a great deal of space has been occupied and very little has been filled. The Mar del Plata is one of chess's cruellest two-front wars. The move order is blunt: 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. White has closed the centre, Black is preparing to break through with ...f5 on the kingside, White will infiltrate with b4, a4 and c5 on the queenside. Whoever arrives first wins. In the ECO classification this is the E97–E99 band. At move nine White reaches a junction. 9.Ne1 and 9.Nd2 are the patient roads: settle the pieces, choose the attack later. 9.b4, the Bayonet Attack, has become the modern main weapon — White grabs queenside space immediately, restricts Black's ...b6 and ...c5 resources, and keeps the b4–b5 lever in reserve. From club level to 2700 its frequency has climbed steadily for two decades because it lets White keep the initiative without taking unusual risk. Black has two replies. 9...Nh5 is the main line — the knight heads for f4, supports ...f5, and the game becomes sharp, concrete and calculation-heavy. 9...a5 is the alternative — Black questions b4 at once, bets on pawn structure before the race of pieces begins, and pulls the game out of the crowd of forced moves into the open field of planning. This product advocates that second road. The DVD description argues that 9...a5 is where a human can think and where memorisation is kept to a minimum. The publisher is not named anywhere; the series format suggests the ChessBase ecosystem, but that is inference, not fact. The presenter is identified only as 'Svitlana' — no title, no rating, no grandmaster norm, no statement that she has ever played the line in serious classical chess. The Olympiad appears in the title; which Olympiad, which year, which games, none of it is disclosed. I repeat an old accounting principle. In December 2026, in Chennai, minutes after Viswanathan Anand's second grandmaster norm, my first question to him was not technical but journalistic: what game does this claim rest on? The habit has not changed. From a desk in Hyderabad I also read chess as a market, and I have gone through Olympiad classical games score-sheet by score-sheet for years, because what I need is opening frequency, not results. The 45th Chess Olympiad in Budapest in September 2026 had 197 teams in the open section — no other team event has that depth. For any opening product this is the cheapest and best raw material available. It does not follow that every product built from it is good. The product's internal logic is coherent, and that deserves saying. Six steps: the Mar del Plata produces rich two-sided strategic positions; engines are uncomfortable in closed structures; therefore human players benefit; but the Bayonet disrupts the picture; 9...Nh5 drags play back into forced, computer-friendly lines; therefore choose 9...a5 and stay on human ground. No coach would object to the internal cleanliness of that chain. The problem is not the first step. It is the second. 'Computers do not understand these positions' is directionally right and temporally stale. That is a 2026-to-2026 experience, when evaluation leaned on material and static search and engine opinion wobbled in slow, plan-based structures. NNUE evaluation arrived in the mainstream with Stockfish 12 in September 2026, and engine reading of closed positions changed with it. Engines now return stable assessments at depth; their advice sometimes diverges from human plans, but that divergence is 'understanding differently', not 'failing to understand'. The advertisement is selling an era that has ended, using the weapon of that era. And the product concedes the point without noticing. If computer analysis becomes important again after 9...Nh5, then the engine remains the presiding judge of the whole board; 9...a5 is not an escape from the courtroom, it is a different room inside it. The real emptiness lies elsewhere. There is not one number in the entire description. No engine evaluation of a critical position, no average centipawn loss, no engine match rate, no database frequency for 9...a5, no results split at any level, and not a single named model game. A repertoire product claims to cover the opponent's options; this one discloses neither its depth nor its boundary. Twenty-odd years ago I left a data vendor's desk because the gap between what the model card showed and what it hid had become intolerable. I did not leave because I disliked estimates; I left because estimates were being sold as verdicts. Opening publishing is now doing the same work, with a DVD cover where the board used to be. The scope is smaller than it looks. In the Mar del Plata, White retains 9.Ne1 and 9.Nd2, both alive, both with independent branches and model games of their own. Before that there are 8.Be3 and 8.Re1, and at move seven Black himself can choose among 7...Nc6 (the Mar del Plata door), the older 7...Nbd7, and 7...a5 or 7...Na6. This product answers one move, one branch, one node. Being an antidote to the Bayonet and being a complete Mar del Plata repertoire are different things, and the copy fuses them into one sentence. That is not a flaw in the product; it is a flaw in how the product will be perceived — and perception is what sells. Two claims travel together in the description: the position is 'objectively equal', and it is nevertheless 'dangerous for White'. Logically there is no conflict. An equal position can be practically lethal; that is the definition of a practical weapon. But 'objectively equal' is a claim settled by engine numbers, and 'dangerous' is a claim settled by a results split. Neither appears. The pairing is a familiar marketing device: the first sentence removes the buyer's fear of purchasing a refuted line, the second raises the hope of scoring chances. Both sentences may be true. Being true is not the same as being shown. Using Olympiad games as source material is sensible: deep field, high stakes, public scores, zero acquisition cost. But team events carry a selection bias. A draw is often sufficient for the team, so players produce result-oriented, risk-averse chess; the same players press differently in a Candidates or a pure round-robin. A repertoire built only on team-event games must be cross-checked against individual elite events, or you will build hard conclusions on a soft sample. Empty stadiums did not kill home advantage; they made it visible. The Olympiad's name gives this product prestige, not strength. One further item is packaging rather than chess, and in my old trade packaging errors were caught first. The title says 'King's Indian Attack' while the subject is the King's Indian Defence. These are different systems — the Attack is 1.Nf3, 2.g3, 3.Bg2, d3, a slow squeeze; the subject here is 1.d4 Nf6 2.c4 g6 ... 6.Be2 e5 7.0-0 Nc6 8.d5 Ne7. In a search-driven market the error cuts both ways: Attack players may buy the wrong product, and Defence players may never find it. In my experience a mislabelled field predicts errors further down; here it is a signal, not proof — but signals are not free either. Now the part where I have to argue with myself. The strongest thing about this product is not the move; it is the audience definition: 'players who think rather than memorise'. That is a real segment, and it is badly served. Most repertoire material today is memory-heavy — long variation trees, engine move lists, play this, then this, then this. A club player with four hours a week experiences that material as weight, not support. A product that offers fewer moves and more understanding meets a genuine need, and that need will not expire. Which is exactly where the trap sits. 'Few concrete lines, strategy rather than tactics' sounds like relief and plays like a burden. The very positions the product calls free, strategic and non-forced are the ones that demand the most decisions under a ticking clock. In a forced line, a mistake is caught by the engine; in a loose position, a mistake can go undetected for years. The copy itself says these positions require long-term planning and tactical sharpness — a sentence that breaks the promise of ease on the same page. The buyer who reasons carefully will discover he traded memorisation work for judgement work. That is not a bad bargain; it is not the bargain advertised. The second trap is causation. Suppose buyers of this product score well with 9...a5 in club events. Is the win the product's? The cause may be selection — the player who buys such a product is already more committed and more studious. The cause may be the opponent's ignorance — White had not prepared for 9...a5. The cause may be something else entirely. In none of those cases is the product's contribution zero, but it is smaller than it will appear. A prior is only a story until the sample makes it pay rent. The third point will sting the publisher. Printing an antidote also hollows it out. The moment 9...a5 is explained in public, White can prepare specifically against it. The 'few concrete lines' advantage was an information edge, and information edges end at publication. This is a good surprise weapon; it is not a permanent edge. So what survives? A great deal, and I will not mark it down. 9...a5 is genuinely playable and respectable — not a discovery, but organised, explained existing knowledge, and organised knowledge has value. Drawing model games from the Olympiad is sound practice. Above all, targeting the thinking player is an insight the market rarely offers. The weakness is not the move. It is the evidence management. Four signals can be measured over the coming months. First, engine posture: whether evaluation of the critical positions holds across depths; a clear White improvement at any depth collapses the core claim. Second, appearances of 9...a5 in 2650-plus classical games; one high-profile win gives the line prestige, one loss takes it away. Third, White's ninth-move frequency: if 9.b4 gives ground to 9.Ne1 or 9.Nd2, the addressable market shrinks. Fourth, publisher disclosure — the presenter's title, the model-game list, whether a database is bundled, whether an errata channel exists. The market moves first; the model monk waits for the closing line to speak. The shelf life of this product will be set not by its advertising but by its update cycle. And the real test of 9...a5 is not whether a grandmaster plays it once — it is whether, three years from now, a prepared White player arrives with a specific plan and Black still holds the balance. Until that answer exists, this is a claim, not a tool.