King's Indian, Bayonet Attack: 9...a5 Is a Reasonable Sideline, and the Data Is Still Missing

"A night unlike any other" is the title of a commercial opening library for Black in the King's Indian Defence, Classical system. I read the promotional page on a Friday evening, with Moscow outside the window below minus ten. The promise is compressed into a single move: 9...a5, following 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 9.b4. The headline evokes soldiers fixing bayonets for a night assault, which is the marketing pun on the Bayonet Attack. I set the advertising aside and built the position on the board first.

The Bayonet Attack 9.b4 appeared regularly on top-level boards from the early 1990s and became one of the heaviest tests for anyone playing Black in the King's Indian. White opens a striking arm on the queenside, preparing b5 and c5, gaining space and weakening his own d5 in a controlled way. Black's classical answer is 9...Nh5, the knight stepping to the rim, clearing the path for ...f5, ...f4 and ...g5, turning a closed centre into a two-wing battle. That is the line every openings manual is obliged to present, and the line engines can check down to the last percentage point. The Mar del Plata variation, whose name is tied to the tournament held in the Argentine city of that name in 2026, survives because of its duality: White holds the attack, Black holds the structure and the hidden counterplay.

9...a5 follows a different logic. The a-pawn strikes directly at the base of b4, keeps the centre tension intact and postpones a fixed pawn structure. Black lets White choose the resolution: bxa5 opens the a-file, b5 locks the queenside, or White keeps the tension and accepts ...axb4. The reward lies in plans engines cannot force into a chain of only moves: ...c5 hitting d5, ...Nc4 planting a piece on a strong white square, and a ...f5 that arrives later but better prepared. From the sideline, I see the whole game. I use that sentence for football, yet it holds here too, because while every eye drifts to f4, the real story sits on the a-file.

One thing must be stated plainly: 9...a5 is an established sideline, not an invention, and certainly not a refutation of the Bayonet Attack. The promotion calls it an antidote, yet in its own conclusion the author concedes that objectively the chances are equal. A move leading to equality cannot be a solution to an opening line; it can only be a reorganisation of the problem. The product's real value sits elsewhere: asymmetry in preparation and surprise against an opponent who has never faced the structure.

Second comes the data gap. The promotion names no player, cites no reference game, offers no database scoring for 9...a5, no engine evaluation, no publication date, no author credentials. For an opening product sold to serious players, that is a substantial shortfall. I keep a spreadsheet for every King's Indian game I watch, a habit formed after a costly lesson: my first major analysis was savaged by readers. The first piece got pelted with stones. Data never takes offence.

The claim about very few concrete lines is the most valuable and least verifiable point in the whole presentation. A repertoire built on 9...a5 still has to solve two problems. First, White may decline 9.b4; 9.Ne1, 9.Nd2, 9.Bg5 and similar moves lead elsewhere, and the buyer needs answers. Second, after 9...a5 White can steer the game back toward familiar main-line structures, which means the Black player still has to know 9...Nh5. In other words, the supposed low memorisation load may simply shift the cost from one branch to another. The scope is undefined as well: does the library cover only the 9.b4 branch, or the whole King's Indian complex after 7...Nc6? That changes the buyer's workload entirely.

The counter-intuitive part is the engine argument. The promotion states that computers often handle the arising complex strategic positions poorly. That argument belongs to the pre-2026 engine era and is dated in the age of neural networks. Neural-engine evaluations of closed, manoeuvring positions are far more reliable than those of the previous generation; their blind spots have moved to different, less structured positions. This does not strip 9...a5 of value, but it changes what the value is: the buyer is not receiving ground computers cannot tread, but ground opponents rarely prepare. That is the value of surprise, and surprise is only worth something once.

Market positioning deserves a read too. The promotion says the material suits players who prefer thinking for themselves over memorising long variations. For improving club players that is a fair offer: they need plans more than move strings. For professionals, abandoning a forcing main line on principle is unthinkable, because their opponents need not know whether 9...a5 is good, only that it exists.

Based on my own tracking of King's Indian games, the worthwhile next step is to open a database, filter games featuring 9...a5, count Black's score by rating band, then compare it with 9...Nh5 under the same filter. If the gap is under five percentage points, 9...a5 deserves a place in any King's Indian repertoire. If it is wider, it remains a one-shot weapon. Systems do not lie, but they can only be heard when the data is thick enough, and this presentation gave me not a single line of it.