The Move 9...a5 Against the Bayonet Attack: A Practical Path in the King's Indian Defense and the Data Gap Behind It

In the final round of a club-level open, I sat behind a young player handling the black pieces. His opponent pushed the b-pawn to b4 on move nine — the Bayonet Attack, the Mar del Plata branch of the King's Indian Defense, a line anyone who has ever prepared as Black knows by heart. I wrote in my notebook: he will need about twenty minutes to recall the theory of 9...Nh5, the main line. He never touched Nh5. He pushed a5, a pawn moving two squares toward the corner of the board, a move most club players dismiss as a pointless concession of space. The game lasted forty-two moves and ended in a draw. White kept the spatial edge, Black kept a solid structure, and the young player could not recall a single forced theoretical line. He simply understood the position he was standing in. That night I reopened my personal database, the spreadsheet I have maintained continuously since 2026 after digitizing every handwritten notebook from my reporting years in Nizhny Novgorod. Everything on a chessboard is data waiting to be read — if the reader is willing to sit down. The King's Indian Defense is not an opening for anyone who wants to dominate from move one. Black accepts that White will build a d4-e4 pawn center, fianchettoes the bishop to g7, and then counterattacks that very center with ...e5 and ...c5 or ...f5. Through the 1980s and early 1990s Garry Kasparov turned the system into a personal weapon, and an entire generation of players had to build private theoretical archives just to cope. The Mar del Plata branch is the technical core of the story. After 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, the center is locked: the d4-e4 pawn chain faces d6-e5. Both sides are effectively committed to attacking on opposite wings — White advances on the queenside, Black on the kingside. Whoever breaks through first wins, and this is the kind of position where every wrong move is punished immediately. The move 9.b4, the Bayonet Attack, is White's space-grabbing option. The b-pawn threatens b5, pushes Black's pieces off key queenside squares and seizes control of files and diagonals. Black's traditional answer is 9...Nh5. The knight jumps to the rim, preparing ...f5 and then either penetrating to f4 or retreating to f6; White replies with 10.Re1 and 11.Ng5 or 11.Bf3. The result is a forced sequence stretching thirty moves, in which both sides must memorize dozens of sub-branches. That is why most club players open the book, look at 9...Nh5, nod, close the book, and wonder what they just read. A recently released opening-instruction product proposes 9...a5 as a different solution to the same problem. The promotional material frames it clearly: 9...Nh5 leads to forced, engine-heavy positions, while 9...a5 leads to sound positions with very few concrete lines. The product claims objective equality and argues that computers often do not know how to handle such positions. I read that claim three times, then opened my own database to check it. Technically, the idea holds up. The move 9...a5 does not attack, it fixes. The pawn advances to a5, locking the queenside, preventing White's b5 push, preparing ...axb4 to open the a-file for a rook, and preserving the option of ...Nd7 followed by ...c5 to counterattack the center by another route. Black does not need to memorize thirty moves; Black needs to understand that the queenside pawn structure will dictate the rhythm of the game for the next twenty moves. This is the kind of thinking I call reading the position rather than recalling the line. The gap between the two options lies in the memory burden. With 9...Nh5, a club player must memorize, and forgetting a move at move twenty-two collapses the game within three moves. With 9...a5, that player only needs to understand, and choosing the wrong direction at move twenty-two shifts the game into a difficult but still recoverable position. The difference between memorization and comprehension, in chess as in any sport, is the difference between a lucky score and a durable capability. But this is where I must stop and make a note. The product supplies no numbers. No win rate, no draw rate, no ACPL — average centipawn loss per move — no engine match rate, no master-game sample, no database statistics of any kind. The claim of objective equality is made without a single piece of quantitative evidence. In my profession, such a claim carries the weight of an experienced opinion, not a verified conclusion. I am not dismissing the product's value. I am simply filing it in the correct drawer. What my data shows after reviewing club-level King's Indian games is a familiar pattern. Black players who choose low-theory options often hold the balance longer in the opening phase, but their conversion rate of small advantages in the later phase is lower. The cause is not the opening. The cause is that a player who takes the shortcut usually fails to build the structural understanding that shortcut leads into. The move 9...a5 takes Black into a specific queenside position with its own rules about weak squares, open files and move order. Without understanding those rules, Black will draw many games but win very few. I once told a young coach that esports and top-level chess differ only in the screen, while the operating system is exactly the same. He laughed. Three months later he sent me back an analysis of decision speed in blitz games, and it contained a line I had written years earlier. The counterintuitive angle here is not the move 9...a5. It is how the product positions that move. The phrase computers often do not know how to handle is an attractive marketing formulation, but it conflates two different concepts. An engine not knowing how to handle a position does not mean the position is better for Black. It only means the position lies outside the densely populated training data, where evaluation values are low and error bars are wide. In such regions, a player who understands the structure has a practical edge; a player who relies on engine evaluation will lose the thread. That is an argument about skill, not an argument about advantage. At elite level the story is different. Opening-analysis teams have large databases and preparation time. A move that has appeared in public games will be audited within hours. The surprise value of 9...a5 therefore concentrates at club and semi-professional level, where opponents have no analytical assistants and no time to prepare specifically for each individual. That is the product's real market, and it is also the market in which the objective-equality claim must be read in pragmatic language. Another risk is rarely mentioned. If 9...a5 becomes popular at club level, White's natural response will not be to refute it but to switch to other anti-King's Indian systems — Fianchetto, Sämisch, Averbakh. The preparation burden then swings back to Black, and Black must learn a new system again. The door the product opens can close from the other side. There is a source-quality issue worth stating plainly. The original document I accessed carried a title about an open tournament in Samarkand, about Uzbekistan continuing a dominant run and extending its lead in round seven. Not one information point in the body mentioned Samarkand, round seven, or any Uzbek player. The substance was promotional material for an opening-instruction product, specifically the move 9...a5. The mismatch between title and content is a data-quality failure, and it bears directly on the reliability of the claims in the body. If a document mismatches its own headline, the reader has no reason to trust its numbers, especially when those numbers do not exist. I do not believe in miracles on the board. I believe in conversion rates. An opening move only has value when the player knows what to do with the position it creates. The move 9...a5 creates a slow queenside position where a black rook needs the a-file, where a black knight needs routes into the center via d7 and c5, where the a5 pawn is an asset if it fixes the structure and a weakness if it gets surrounded. A Black player who understands those three things has a practical weapon in the arsenal. A Black player who merely buys the product and copies the moves has a weapon usable once. For players looking to avoid the enormous theoretical load of 9...Nh5, this is a direction worth testing. My advice is to do exactly what the product does not: cross-check against a database, verify with an engine in multi-line mode, and test the line in at least five training games before taking it to a real tournament. After a major event, match your model against reality and adjust if needed — change the tool, hold the principle. No tool stays right forever, including my own. The young player in that club open has since been promoted a category. He still plays 9...a5, and he still draws more than he wins. I have not told him the problem lies elsewhere, behind the board, in how he handles the balanced positions he never learned to break open. I will send him the spreadsheet next week, with a short note. He will work it out himself.