Mar del Plata, the Bayonet Attack and 9...a5: An Antidote Without Data

Mar del Plata, the Bayonet Attack and 9...a5: An Antidote Without Data Chess pieces on a wooden board have a drier sound than plastic ones, and in a club in Hai Phong on a Saturday evening that sound ticks like a metronome. In front of me was a game that had reached move nine of the Mar del Plata Variation. White had just pushed the pawn to b4, the signature of the Bayonet Attack. Black faced two book options: bring the knight to h5, open the position and force calculation, or push the pawn to a5 and change the whole character of the game. At the same time, my inbox held a promotional document for an opening-theory product. Its headline referred to round two of a team event in Samarkand, Uzbekistan. The entire body below it discussed the move 9...a5 in the King's Indian Defence. Two unrelated texts had been pasted together with no line of explanation. So I set the board up again. I trust footage more than testimony, because footage does not lie. In chess, footage has another name: a game database. When I compared the marketing language against the database, the gap became clearer than any move on the board. The product's promise rests on three claims. The move 9...a5 is an antidote to the currently fashionable Bayonet Attack. It leads to sound positions with very few concrete lines, so the player does not have to memorise. And if Black knows what to do, things can quickly become dangerous for White. Those three claims arrived without a single line of data. No average centipawn loss, no win-and-draw rate for the branch, no sample games, no author name in the material I received. A product teaching an opening that offers no game to verify leaves the reader with only the seller's tone of voice to rely on. At forty-eight, after thirty-two years watching this field from a player's chair, a scorekeeper's chair and a documentary screenwriter's chair, I have learned something simple. The most expensive thing in a theory product is not the moves. It is the right to verify. Buyers pay to save time. If a product gives them no way to check it themselves, they are paying for reassurance, not for knowledge. Mar del Plata: a two-faced position The King's Indian opens with moves that look concessive. White takes the centre with 1.d4, 2.c4 and 4.e4. Black answers with 1...Nf6, 2...g6, 3...Bg7, 4...d6 and 5...0-0, letting White build a large pawn block in the centre before attacking that block with 6...e5. The philosophy is hypermodern: contest the centre by letting the opponent occupy it and then undermining it. The Mar del Plata is the fiercest version of that idea. After 7.0-0 Nc6 8.d5 Ne7, White seals the centre with the d5 pawn while Black commits everything to the kingside with ...f5, ...g5 and ...Nf6-h5. The variation takes its name from the tournament held in Mar del Plata, Argentina, in 2026, where Soviet players brought it into large-scale practical use. In Encyclopaedia of Chess Openings terms it sits around E97 to E99. What makes it distinctive is the two-wing contrast. White has space and a huge pawn mass on the queenside, but the white king sits directly beneath Black's coming pawn storm. Black has less territory, yet every piece points at a single target. That is why the variation was once a weapon for many top players, and why Vasyl Ivanchuk and Teimour Radjabov rank among its most committed high-level practitioners. The Bayonet Attack is White's most popular answer to that storm. Instead of waiting for ...f5 and reacting, White plays 9.b4 and opens the queenside first. The point is concrete: gain space, prepare c5 to drill into Black's centre, or push b5 to lock the queenside before turning to the other flank. The Bayonet turns the game into a two-way race. Against this structure, Black has two recorded directions. The move 9...Nh5 is the main line. The knight heads for f4, forces White to calculate immediately and opens a chain of forcing variations both sides must know. The move 9...a5 is the secondary branch. Black does not rush the kingside but strikes at the b4 pawn the moment it lands, asking a question exactly where White just opened a front. Both branches are played at every level. They are not equivalent in theoretical standing. The 9...Nh5 branch has accumulated thousands of high-level games and is treated as the main road. The 9...a5 branch exists, has practitioners, has wins, but has far fewer games and far less documentation. Anyone opening a database sees this within three minutes. What 9...a5 does to the pawn structure After 9...a5, White faces a choice in which no option is free. If White answers bxa5, the b4 pawn leaves its post, the a-file opens, and Black can bring a rook to a5. Black gives up a wing pawn but gains an open file, an active rook and a weakened white queenside pawn structure. In a position where Black is already planning to blow up the centre, an extra open file is real profit. If White answers b5, the queenside closes. White keeps the space and the b5 pawn as a wedge, but simultaneously shuts down its own attacking direction on that flank. From then on White has almost no meaningful queenside plan and the game shifts to the other side of the board. That is exactly what Black wants: a battlefield Black knows better. If White defends b4 with a piece, White preserves the tension but pays with a slower development move. Each option trades space for tempo, and in the Mar del Plata tempo is Black's most valuable asset. This is where I want to slow down, because it sits at the centre of the story. The move 9...a5 does not demolish the Bayonet Attack. It converts it into a different type of game. Instead of a two-way calculation race with forcing lines running fifteen moves deep, 9...a5 turns the game into a positional manoeuvring battle, where plans matter more than short-term memory. For a club player, that is a genuinely attractive offer. But converting the type of game does not reduce the calculation requirement. The pawn structure after 9...a5 is still the Mar del Plata structure: Black must still push ...f5, still push ...g5, still move pieces to the kingside, still accept a king standing on a pawn rank that is gradually opening. White can still stab with c5 into the centre, still bring a rook to the c-file, still open the b-file. No move in that chain is automatically safe. That is why the phrase "very few concrete lines" must be read correctly. It describes memory load, not thinking load. A branch with few forcing lines can demand more analysis hours than one with many. Players choosing 9...a5 do not save effort. They shift it from memorising to understanding. Three promises without data Back to the document. Three points can be verified and three cannot. Verifiable point one: 9...a5 exists inside the King's Indian, Mar del Plata, Bayonet branch. That is correct. It is a real move with precedent and high-level practitioners. It is not an invention, and the document does not claim otherwise. Verifiable point two: 9...a5 differs in character from 9...Nh5. The 9...Nh5 move opens the position and drags in a chain both sides can study in advance. The 9...a5 move keeps the position more closed and steers toward manoeuvring. That classification is accurate. Verifiable point three: both branches lead to objectively balanced positions. That is a reasonable positional statement, in the sense that any high-level King's Indian branch has at some point been assessed as balanced. The remaining three points sit outside what the document can support. There is no move-quality metric against the engine's top choice. There is no win, draw and loss rate for the branch. There is no list of sample games. There is no named player, tournament or year. There is no author name, no title, no personal record in the very line being taught. For a product sold as an antidote, that absence matters. An antidote is a strong claim. It says that after studying, the buyer will hold a reliable answer to a troublesome line. A strong claim needs strong evidence. The evidence here would be statistics, games and names. Without them the claim never leaves where it started: inside the seller's tone. I have been writing score sheets at club events since 2026, and I learned early that a mistaken score sheet cannot be fixed by speaking louder. The only fix is to pick up the sheet and replay every move. For a theory product, that score sheet is called a game database. Buyers have the right to open it before paying. One more commercial point deserves clarity. Modern theory products mostly sit in the lecture-video catalogue format, with a board interface, a variation tree and annotations, a familiar category in the German-language chess market. The format is not bad. It helps intermediate students who want a structured path instead of swimming through millions of games. But it carries a built-in weakness: it tells a linear story while the position itself branches. Students easily confuse the coherence of a lecture with the simplicity of a game. "Computers do not understand" is an outdated line The document contains a notable sentence: computers often do not know how to handle such strategically complex positions, so humans hold the advantage. That sentence was once true. It held in the era when traditional engines evaluated positions mainly through deep search and handcrafted evaluation functions, before neural networks changed the field. In 2026, AlphaZero showed that a neural network could judge long-term positions far better than the prevailing assumption. Afterwards, open-source engines such as Leela Chess Zero and neural-network builds of Stockfish raised the whole field's evaluation standard. Closed, strategically rich positions with long-term plans are precisely the category modern engines handle far better than they did a decade ago. That makes the sentence outdated, and worse, outdated with a sales purpose. It manufactures a fake enemy so the buyer feels aligned with humanity against machines. A theory product that sells by diminishing engines, while almost certainly having been built with engines and databases, carries a contradiction that is hard to overlook. I ran hurdles all my life; I was simply never timed. With opening theory I occupy a similar position: enormous analysis, most of it never validated by an official timed game. Without that timing mechanism, people mistake their confidence for accuracy. This product is the same. It contains no official game to be timed, so nothing restrains its confidence. A mismatched headline and what it costs The document's headline described round two of a team event in Samarkand, Uzbekistan, reporting that the top teams all won despite close calls. The body described an instructional disc on an opening. The two have nothing to do with each other. This may simply be a data-stitching error in the aggregation stage. For an editor, though, it belongs to the most serious class of errors. Headlines are the most read, most shared and most quoted part of any text. When the headline says one thing and the body says another, everything built on that document carries an embedded defect. A writer who trusts the headline narrates the wrong event. A writer who trusts the body describes the wrong tournament. Neither checks the origin. One detail is worth noting. Uzbekistan has a strong chess culture, regularly hosts international events, and team competitions across Central Asia still take place. A round-two report from such an event could well exist. It simply has nothing to do with 9...a5. The fact that the two texts were glued together shows that the content pipeline passed through an automated stage with no final human check. This is why I still keep the habit of rewinding and cross-checking every source. I rewind Sea Games 29 footage to find what everyone else forgot. This time, what was forgotten sat on the first line of the document. Questions to ask before paying If a Vietnamese club player read this material and considered buying, several steps come first. Open a database and look up the 9...a5 branch. See how many games it contains, at what level, with what results, and whether any appear in high-rated events. This takes minutes and costs nothing. Run a modern engine at a reasonable depth on the opening moves of the branch and check whether the evaluation is stable or swings. A branch whose evaluation swings sharply between nearby moves demands deep strategic understanding, and that requires a different study design from memorising variations. Find the author's name and check their record in this exact line. Someone teaching 9...a5 should have games of their own there. If not, the buyer is learning from someone retelling another person's theory, and that should be known in advance. Compare against free sources. Public opening research with full variation trees and reference games is widely shared on learning platforms. If a paid product cannot supply what a free source already has, its value lies in presentation rather than content. In Vietnam, top players such as Le Quang Liem and Nguyen Ngoc Truong Son have long prepared openings with databases and engines. Youngsters in the training system now receive the same tools far earlier than my generation did. In that environment, a theory product only stands if it answers one question: what does it add that a database does not. The best answer I have seen is structure. A good study path, a sensible grouping of position types, a set of exercises that lets the student self-test. Those are worth paying for. A single move is not, because free sources never run short of single moves. The biggest risk in this branch One type of risk deserves a direct statement. Any antidote can be neutralised once the opponent analyses it. The Bayonet branch has been researched at the highest level for decades. A move with precedent like 9...a5 has almost certainly been examined by elite players looking for a reply. Buyers may be studying a branch whose answer has long existed on the other side without their knowledge. Alongside that sits the risk of study time. A club player spends twenty hours on a secondary branch, then discovers that the regular opponents at the club never play the Bayonet Attack at all, and none of that work is ever used. This is the most common risk and the least discussed in promotional material. One further point concerns how this industry sells. Theory products often target a specific self-image: the independent thinker, the player who dislikes memorising, the player who wins through understanding. That is selling identity rather than content. I have seen the same mechanism elsewhere in sport, where money follows image and symbolic commitments rather than real investment in structure. Selling identity is always easier than selling a skill, because identity needs no verification. What chess writers should do Asked how a newsroom should handle documents of this kind, I would require three things before publication. First, state the document's origin and publication date. Second, state the author's name and title, or state explicitly that none is given. Third, separate the verifiable part from the promotional part and label each. This does not make an article less compelling. It makes it more trustworthy. Vietnamese chess readers increasingly look things up themselves. They will find the exaggerations, and when they do, they will remember. Trust in a small community is built slowly and lost quickly. There is a small but valuable detail in this whole affair. The move 9...a5 is reasonable. It does not need to be sold with inflated promises. It stands on its positional merit: a different way for Black to play, demanding understanding rather than memory, suited to a particular group of players. Had the document said only that, this article would not exist. The problem is that people wanted more. They wanted an antidote, a solution to an opening problem, a guarantee. And when people want a guarantee, they usually receive a headline better than the content. Every documentary is a track: the audience sees the finish line, I live at every starting block. An opening product works the same way. The buyer sees 9...a5 on the cover. A writer like me has to live through every move behind it, every game, every line of data, and answer the question no advertisement wants to answer: how many people actually tried it, and how far did they get.