9...a5: The King's Indian's Escape Route Against the Bayonet Attack

Wednesday, after lunch, I sat in a small chess club on Huaihai Road in Shanghai, watching two middle-aged men play a league game. There were four people in the room: the two players, a scorekeeper, and me. On move nine, Black pushed his a-pawn to a5. The board went quiet. It was the silence of a man who has just chosen his road and knows he will be walking the rest of it on foot. I remember the ticking of the old mechanical clock behind me, the lid of the teapot opening and closing, and the way White picked up his rook and put it down twice on the same square. Nothing exploded. No combination worth a gasp. Just a pawn on a5, a file about to open, and a position that had changed key. That game lasted sixty-two moves and was drawn with rook and three pawns apiece. I spent the whole afternoon thinking about an edge pawn. To any chess player, this is the territory of the King's Indian Defense, Mar del Plata variation, ECO E97 to E99. In short: Black concedes the centre, builds the g6 wall with a bishop on g7, plays d6 and e5, then counterattacks with ...f5. White answers with d5, locking the centre and opening two fronts on the flanks. The beauty of the system is that the two sides play different games on the same board: White attacks the queenside, Black attacks the kingside, and whoever gets home first wins. For decades White tried everything to slow the kingside down. The most annoying method is the Bayonet Attack: 9.b4. White pushes the b-pawn, grabs queenside space, clears the c1-a3 diagonal for the bishop, and prepares b4-b5 to cramp Black's pieces. It is not a gentle move. It turns the game from Black attacks, White endures into both sides attack, and whoever errs first dies. Against the Bayonet, the orthodox answer has long been 9...Nh5: the knight goes to the rim, ...f5 follows, and Black accepts a tangle of forcing variations. The alternatives ...Ne8, ...Nd7 and ...a5 exist alongside it but are rarely discussed. In the opening databases I checked, the overwhelming majority of top-level games still go ...Nh5, the main line, analysed down to move twenty. That is why I stayed with the file about 9...a5. The material arrived in a package whose headline described round two of a team event in Samarkand, Uzbekistan. The body was something else entirely: a promotional text for a chess opening product sold on disc, advertising 9...a5 as the antidote to the Bayonet. The headline and the body do not match. In my trade that is the most serious kind of error, the kind that forces you back to the primary source before you quote a single word. The material names no author, carries no publication date, and includes no sample game. I read on anyway, with a red note in the margin. 9...a5 does exactly one thing, and it does it very early: it asks White's b4-pawn how long it intends to stay there. White has three choices. Take on a5, opening the a-file for Black's rook and handing Black the ...Nxe4 idea against a centre that has lost a shield. Push on to b5, clamping the queenside but weakening his own pawn structure and leaving squares for Black's pieces to occupy. Or keep the tension with Ba3 or Be3 and wait to see what Black does first. All three roads lead to the same kind of position: slow, closed, full of pieces, short of forcing lines. One thing has to be said clearly, because advertising tends to bury it: 9...a5 does not create a theoretical advantage. It is a line that has been known, and played at various levels, for some thirty years. Its value lies elsewhere: it changes the currency you pay with. With ...Nh5, the game is usually decided in the first fifteen to twenty moves by memory. Whoever knows more, whoever updated his file more recently, survives the opening; the man who runs out of theory first is usually the man who loses. With ...a5, the down payment is far smaller, but the interest is harder: there are no lines to remember, only pawn structures to understand. You trade a memory test for a judgement test. For a club player with two hours a week to study openings, that is a trade worth weighing. It is also why this branch has survived so stubbornly without ever taking over. In the engine era, the King's Indian's share of elite games has fallen sharply: a closed, centre-conceding system is no longer a default choice when everyone has an engine in the analysis room. It survives as a surprise weapon, and its faithful are dwindling. Teimour Radjabov is one of the few still keeping that room lit at the highest level. Here, though, I part company with the document. The pitch rests on a familiar sentence: computers often do not know how to handle these complex strategic positions, so the human decides. That sentence was true around 2026. It stopped being true when neural-network engines arrived, because those machines judge closed positions, long-term plans and pawn structures better than most grandmasters do. Selling an anti-computer weapon with an outdated premise is pleasant comfort for the buyer, not a technical argument. The second blind spot is subtler. The whole chess world argues about whether 9...a5 is theoretically balanced, while the real question sits somewhere else: who actually pays for this. The buyer is almost certainly not a 2700 grandmaster. He is a club player who just lost a league game on move twenty-two because his opponent knew the book better. For him, the value of a dull line is simply staying at the board thirty minutes longer. When the seventeen-year-old prodigy does not yet know he is about to become a legend, he is still memorising move nine. The children in the training halls of Shanghai and Hanoi today learn openings by downloading a file of variations, not by reading books. A product that tells them to understand rather than memorise therefore sells something more than lines: a moral posture, in which the player who understands stands above the parrot. I do not entirely believe in that posture, but I understand why it sells. And the most romantic mistake of mine is to believe that chess pieces know how to write poetry. Pieces know nothing. People only write poetry about the silences between moves, and a pawn on a5 is a rather long silence. One thing is certain. Within a single theory cycle, if 9...a5 is used more often, White will find an answer. Opening history is the history of such loops: the antidote, the antidote to the antidote, and then an old line returning twenty years later as if it had never been buried. Anyone who buys an opening disc believing he has just purchased a permanent solution will be disappointed somewhere between three months and three years. What I took home from that afternoon on Huaihai Road was not wrapped in any single move. It was the moment when White, after twelve minutes of thought, pushed his bishop to a3 and began a plan he would never be able to explain in variations. The strength of lines like 9...a5 is that they give a player back the right to be wrong responsibly. An empty arena still holds the applause of those who have gone. That small club had nobody departed in it, but it had exactly one round of applause, from the scorekeeper, when the game ended in a draw nobody had predicted. I closed my notebook, walked out into the street, and thought about all the games that were never played. After three decades of holding a pen, I realise I have never written about sport. I have only written about people trying to remember, and people who decided to stop remembering so they could begin to understand.