9...a5 in the King's Indian: The Bayonet Attack, a Low-Theory Promise, and the Missing Data

I replayed the game with the sound off, rewound to the position after move nine, and left the frame there for a long time. White had just pushed the pawn to b4. Black, instead of lifting the knight to h5 as usual, pushed the pawn to a5. One square changed hands, but the character of the whole game changed with it: the queenside closed, the centre locked, and both sides settled into a quiet build-up that can only be understood through structure rather than through individual moves.

That day I also read a few pages promoting an opening DVD. The product sells 9...a5 as an antidote to the Bayonet Attack, the 9.b4 line that young players are using heavily. What made me stop was not the move but the way it was sold: “few concrete lines”, “no need to memorise much”, “for players who like to think for themselves”. A very attractive promise, and a very hard one to verify.

One sentence in it caught my attention more than the rest: computers often do not know how to handle these positions.

Sometimes, to understand a game, you have to stand in silence longer than other people can bear. I decided to spend the time rereading the whole argument, checking it against what I know about the Mar del Plata, and looking for where the evidence actually sits.

Context: a move that lives inside a structure

The whole story sits inside the King's Indian Defence: 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. Black lets White take the centre and strikes back with ...e5 and ...f5; White masses on the queenside and attacks there. This is the Mar del Plata, named after the Argentine coastal city where the line was forged in the 1950s, falling roughly in the ECO E97–E99 range by my own classification.

White's ninth move, 9.b4, is the Bayonet Attack. White pushes the b-pawn to hit the c6-knight, gains queenside space, prepares to open the b-file and slows Black's counterplay. The orthodox reply for decades has been 9...Nh5, followed by ...f5, sometimes ...Nf4. Those are forcing lines: both sides must calculate concretely, and one tempo lost can ruin the pawn structure. That is why the Bayonet became a line you must memorise, and 9...Nh5 a line you must analyse.

9...a5 goes another way. Black does not send the knight to the rim but attacks the advanced b4-pawn immediately. If White pushes b5, the structure closes on the queenside and Black gets the c5-square for a knight, the a-file, and a clear target. If White trades, Black opens the a-file and plays on the queenside half of the board. This is a line that has been in theory books for a long time, not a new invention.

I want to draw the geometry on the board here, because talking in shapes helps readers who never went through a chess school see the problem. After 9.b4 a5, if White pushes b5, Black owns a chain of squares: c5, a4, b6, and an a-file that gradually opens later. That chain attacks nobody immediately, but it creates a corridor for Black's pieces to walk through. After 9...Nh5, by contrast, Black creates a spike on the rim and forces White to answer with a concrete move right away.

The difference between the two moves is not strength. It is time. One asks a question at once; the other waits.

Analysis: three things you need, and all three are missing

The three data points you need to judge any opening DVD are the match rate with the engine's top choice, the ACPL (average centipawn loss), and a sample of practical games at high Elo. For this product, all three are empty. No engine evaluation, no sample games, no database statistics, no author's name, no record for that author in this very variation.

The only thing on offer is qualitative description: sound positions, equal chances, few concrete lines. Those are positional claims, not theoretical ones. In other words the product does not claim a novelty; it claims usability. Such a claim cannot be entirely wrong, but it also cannot be checked against any quantitative data.

The real value of the product lies in its guidance and its choice of variation, not in the move 9...a5.

The point I care about most is the human-versus-machine framing. The promotional text sets 9...Nh5 against 9...a5 in a particular way: the main line opens the position, forces the lines, and makes computer analysis important again; 9...a5 leads to sound positions with very few concrete lines, where computers often do not know how to handle things.

That last sentence is the language of the pre-AlphaZero era. Today's neural-network engines handle the long-term strategic positions of the King's Indian far better than the sales line implies. Using it as a selling point is using a dated premise, and I say this as someone who has spent years comparing machine assessments with human assessments inside this exact family of lines.

The third point is the audience. The product targets players who like to think for themselves rather than memorisers. That is a real segment: club players rated roughly 1800 to 2200 who do not have time to memorise hundreds of 9...Nh5 lines. For them 9...a5 is genuinely useful. It offers a plan instead of a sequence, a structure instead of a life-or-death line. But for top grandmasters that value falls away, because White only needs to prepare once to neutralise the whole idea.

Put differently: 9...a5 is a legitimate sideline, not the main line. The main line is still 9...Nh5. Anyone buying the product expecting to solve the Bayonet Attack once and for all is buying an expectation larger than what theory confirms.

So if someone genuinely wants to use 9...a5, the right way is not to memorise the entire branch. The right way is to understand the three pawn structures that can form on the queenside, to know where the knight goes once the a-file opens, and to accept that White always keeps a plan. Opening theory at club level does not work like insurance; it works like a map. A wrong map gets you lost, but a right map still will not carry you home if you cannot read the road.

Contrarian angle: the blind spot is off the board

The biggest blind spot in this document is not a chess move.

The headline describes a team-event round report from Samarkand — “R2 (open)”, “despite close calls, all top teams grab wins” — while the entire body is a promotional blurb for an opening DVD. Two unrelated texts have been spliced together. For an ordinary reader the error is harmless. For anyone rewriting or citing it, it is a warning sign: once the top of a document is wrong, the odds that the body has been cut and pasted are not small. I cannot verify a product from a document that contradicts itself.

Second, a promise has to be read with the right ruler. The sentence “if Black knows what to do, things might quickly become dangerous for White” is a conditional description, and its condition — Black knowing what to do — depends entirely on how well White has prepared. A claim that cannot be wrong is a claim with no predictive value.

Third, the “few concrete lines” pitch is itself suspect. The Mar del Plata is sharp by nature; no branch removes calculation. Even with the queenside closed, White still has b5, still has the plan of pushing c5, still can open the b-file. You do not remove calculation from the King's Indian; you move it from concrete lines to positional judgement. Two different kinds of difficulty, not a liberation.

And here I want to speak plainly: a sideline marketed through identity psychology is usually selling buyers a picture of themselves rather than a move. Buyers are not buying 9...a5; they are buying the feeling of belonging to the group of clever players rather than the memorisers. That feeling is not wrong, but it has no correlation with the quality of a line.

The theory stage has no grandstand, so people have to applaud themselves. Tactics never lie; only the reader misreads them.

What to check in the next game

There are four things to track, and all four are observable in public data.

First, whether 9...a5 appears in classical games above 2700 Elo. If the branch really is an antidote, it will show up there within two theory cycles. Second, the frequency of 9.b4 in the opening tree: if the Bayonet Attack stays popular, the product keeps its shelf life; if it fades, both the problem and the answer disappear with it.

Third, the author's name and their practical record in the Mar del Plata itself. A specialist with results in the line is evidence; an anonymous name is not. Fourth, the original of the Samarkand report, to establish where the text was spliced and whether the DVD blurb was cut from its original context.

Based on my experience following games, sidelines like this usually have short lives: a wave of interest, a few months, then White finds an answer and everything goes quiet. What is worth keeping is not 9...a5, but the way it illustrates an old rule: a move does not create an advantage by itself, it only creates an opportunity. And that opportunity belongs to whoever understands the structure better, not to whoever buys more DVDs.

People say girls know nothing about tactics. So I write for them to read.