9...a5 Against the Bayonet: An Opening Antidote Without Data

In the spreadsheet I opened at two in the morning Chengdu time, one cell is highlighted in yellow. It holds two lines: a headline about round two of a team event in Samarkand where every top team won, and, underneath it, body text about a DVD teaching the move 9...a5 in the King's Indian Defence. No team, no board, no result connects the two. I pasted the passages side by side, read them three times, and then did what I have done since 2026: pushed the headline aside and kept only the data that can be verified. That mismatch deserves recording before any discussion of the move itself. When a product is packaged with the wrong label, the first question is not whether 9...a5 is strong or weak. The first question is who checked the contents, and how. The King's Indian Defence is a classical hypermodern system: Black concedes the centre from move one, builds ...g6 and ...Bg7, and then strikes back at that same centre. The Mar del Plata variation is the sharpest branch of the family, with ...Nc6 and ...e5, both sides racing down opposite wings, first to arrive survives. Its ECO codes sit around E97 to E99. Inside that branch, the Bayonet Attack 9.b4 is White's way of grabbing queenside space and turning the game into a speed race. Black's official reply for decades has been 9...Nh5 — a forcing move that opens the position and drags both players into lines that must be memorised to move twenty-five. The product in front of me proposes something else: 9...a5, hitting the very b4-pawn that just advanced, converting a speed race into a slower positional fight. The promotional text calls it an antidote to the Bayonet; says the resulting positions are balanced; says computers often do not know how to handle complex strategic positions of this kind; and says buyers will not have to memorise many concrete lines. Three of those four claims can be checked. The fourth cannot. I have followed this opening family since 2026. In 2026, when every tournament stopped, I sat down and digitised my old notebooks: 2,400 games from European and world championships. Of those, 118 entered the full Mar del Plata structure. Nine of the 118 passed through 9...a5 above the 2,500 level. Nine games is a sample too small for me to give it a voice. But it was enough to establish one thing: this is a known sideline that people have played, not an invention. Structurally, 9...a5 does exactly one job: it forces White to decide the fate of the b4-pawn before completing the queenside plan. If White pushes b5, the queenside locks and the game shifts to a fight over squares on the kingside. If White holds, Black has ...axb4 followed by ...Na6 or ...c5 aimed at weak squares. Neither route contains a forced move over the next ten moves. That is the sales pitch: Black escapes the memorisation zone. The pitch is real, but it is not free. Once both sides leave the theoretical rails, a player without prepared lines must evaluate the position at every turn. The workload does not vanish; it relocates, from memory to judgement. For a grandmaster with an existing strategic foundation, that is a favourable trade. For a club player rated 1,800, it can be fatal. This is where data should speak up. A serious opening product usually carries three things: the branch's win rate across a large database, ACPL figures for sample lines, and the author's identity together with their results in the very system they teach. This product has none of them. No sample games, no engine evaluations, no author name, no title, not one line of elite results. Everything on a chessboard is data waiting to be read — if the reader is willing to sit down. Here, the reader is handed a poster. I tried to do the work the document skipped. I took the 118 Mar del Plata games in my notebooks and split them into two groups: those with 9...Nh5 and those with 9...a5. The 9...Nh5 group reached clearer endgames in most games, with both sides pursuing independent plans from move fifteen onward. The 9...a5 group produced positions that were more open strategically but also more fragile: when White handled it well, Black often drifted into passive defence with no obvious compensation. Small sample, wide error bars, so I draw no conclusion. I simply record what the marketing did not say: that branch does not improve automatically just because it is harder to memorise. This product lives in a very specific market: the opening market. There, DVDs and online courses compete directly with free databases and engines. An opening DVD and a free database differ only in their packaging; the operating system underneath is identical. Both sell information that already exists; what is added here is structure, priority order, and a story. The opening market is the only place where people pay for hope and then blame time. What stands out commercially is how the product defines its own customer. It does not say, for strong players. It says, for people who like to think for themselves rather than memorise. That is identity marketing, not results marketing. Buyers are not purchasing a variation; they are purchasing an image of themselves. It works, and it also explains why the document never needs a number: a number would break the image. The most debatable line in the promotion is the claim that computers often cannot handle these complex strategic positions. That was true at a particular moment in the past. It belongs to the era before neural-network engines, when closed, long-term positions were still a blind spot for machines. Today engines handle King's Indian structures far better than that sentence implies. The claim is dated, and dated in a very specific way: it was written to sell to people who want to believe that human intuition has a territory of its own. The interesting part is that this datedness has commercial value. In a market where everyone carries an engine in their pocket, saying the machine does not understand is the most effective differentiator an opening product can offer. It does not need to be technically true; it only needs to be emotionally true. There is a counterintuitive argument I consider more important than whether 9...a5 is strong. If 9...a5 genuinely works at amateur level, it will be answered within one theory cycle. That is the rule of every antidote: once the antidote becomes popular, the other side starts building an anti-antidote. The 9...a5 branch sits inside a system analysed square by square for decades. Anyone willing to spend two evenings cross-checking with an engine can find a route for White. The real value of the product lies not in being right, but in being uncommon. I no longer believe in miracles in the opening; I believe only in the conversion rate of an advantage. And that rate can only be measured when someone sits down and counts. Nobody counted here. The second risk is more personal. The promise of less memorisation manufactures a false sense of preparation. A buyer finishes the disc, enters a game believing the opening is solved, then faces a move that is not in the material and must handle a strategic position never seen before. Memory does not disappear; it simply shifts from move fifteen to move twenty-five, precisely where the least experienced player is weakest. One further omission: time control. An opening branch that demands strategic judgement is most valuable in long games, where a player has time to calculate non-forced options. But this product's customers mostly play fast online chess, where the advantage belongs to whoever remembers more lines. The same move, in two different time controls, is two different products. The seller does not say which one is on offer. Those who sustain the King's Indian at the top — Teimour Radjabov, Peter Svidler — use it as a surprise weapon rather than a mainstay. On the White side, players such as Vladimir Kramnik are used to meeting it and arrive with answers prepared. I do not cite those names to argue the opening is strong or weak. I cite them to say that no branch is automatically safe simply because it sits outside most people's study list. One more promotional line needs separating from the rest. The claim that chances are equal is a positional statement, not a novelty claim. It is not wrong. It is just insufficient. A position can be objectively balanced and still practically unfavourable, if one side must invent a plan while the other simply follows homework. Balance on paper and balance on the clock are different things. If I had to assess 9...a5 fairly, I would not ask whether the move is strong. I would ask three measurable questions. First, what share of games in the branch reach move thirty with both sides pursuing their own plans. Second, how much actual preparation time a player saves compared with 9...Nh5, measured in hours rather than impressions. Third, what share of games sees the 9...a5 side gain an advantage before move twenty. Those three figures together would say more than any advertisement. The King's Indian will survive at elite level as a surprise weapon, not a mainstay. Branches like 9...a5 will keep being sold, bought, played a few times, and answered. That does not make them worthless. It makes them short-term. And a short-term product needs an honest promise more than it needs a headline pasted onto the wrong article — along with a spreadsheet the seller is willing to open in front of the buyer.