9...a5 against the Bayonet Attack: a strategic move or a repackaged theory product?

The board stops at move nine. White has just pushed the pawn to b4 — the Bayonet thrust. Black's clock starts, and in that silence most Black players take the road already stored in memory: 9...Nh5, the knight to the rim, the kingside cracking open, and the game racing into a chain of forcing lines that the engine has already ploughed through move by move. There is another move, less discussed, and it has just been packaged as a product: 9...a5.

From my experience of tracking King's Indian games, I still keep the habit of timing the clock at that exact instant — the silence before Black touches a piece. What I have learned from doing that many times has nothing to do with whether 9...a5 is strong or weak. It has to do with who is selling you the feeling that one move can erase the obligation to calculate.

Mar del Plata is the sharpest branch of the King's Indian. The classical sequence runs 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, and from there White has several approaches; 9.b4 — the Bayonet Attack — is the most favoured thrust of the past two decades. White gains queenside space, aiming for b4-b5 and c4-c5, opening files and freeing the rook. Black, squeezed for space, answers with a kingside campaign: ...f5, ...f4, ...g5, ...g4, pushing pawns all the way to White's queenside. This is a game of two opposite attacks, where time is measured in moves rather than minutes.

Black's main reply to 9.b4 is 9...Nh5. The move is forcing: the knight threatens ...f5 and eyes f4, obliging White to respond at once, and pushing both sides into lines memorised to the level of individual moves. That is why 9...Nh5 is regarded as the main road, and also why it has become a memory burden for both sides. 9...a5 is the secondary branch. It attacks the b4 pawn directly and steers the game into a more closed structure, where both sides must settle the queenside before the kingside blows up.

This is where the space map has to be redrawn.

When Black plays 9...a5, the b4 square comes under attack. White has several choices: trade with bxa5, hold with Ba3 or Bb2, or push b5. Each choice rewrites the queenside control zones in a different way. If White trades with bxa5, the a-file opens and Black usually recaptures with ...Bxa5 or ...Nxa5 depending on the position; the pawn structure becomes more symmetrical but White's queenside loses the b4-b5 spearhead. If White holds with Ba3, the bishop leaves the long diagonal to guard the pawn, and Black gains extra time for ...f5. If White pushes b5, Black concedes space but gains the a5 square as a base and the a-file as an entry route. In all three cases Black has achieved the most important thing: slowing White's queenside attack by exactly one tempo, enough for the kingside to enter the game.

The core point: 9...a5 does not create a theoretical advantage, it creates a different map. And on that map, what is stolen is not White's pawn structure but White's tempo.

Compared with 9...Nh5, the difference lies in the character of the lines. 9...Nh5 opens the game up: pieces meet quickly, moves are forcing, and mistakes are punished immediately. 9...a5 closes the game down: pieces operate at longer range, plans run longer, and mistakes are punished slowly — but still punished. What is notable is that in both cases the central pawn structure is nearly identical: White still has the d5 wedge, Black still has e5. What changes is not the material but the distance between the pieces and the moment they meet.

In the diagrams I used to rebuild for my own blog, I always found one common feature in Mar del Plata games that choose 9...a5: the number of moves before the two sides clash on the kingside rises markedly. The game loses its move-counting quality and shifts to square-grabbing. That is an environment in which a player who understands space but does not have computer theory memorised can survive — and that is exactly what the product is selling.

The product is advertised with a few headline claims. First, 9...a5 leads to sound positions with very few concrete lines. Second, 9...Nh5 opens the position, forces the lines and makes computer analysis important again, whereas 9...a5 leads to positions computers often do not know how to handle. Third, the chances are said to be equal, and if Black knows what he is doing, things quickly become dangerous for White.

Those three claims need to be separated and examined one by one.

The first is a positional claim, not a novelty claim. 9...a5 is a known move, not an invention. That it leads to few concrete lines is right in the sense that the game turns strategic, but few compared with what standard is never quantified. There is not a single metric in the package: no win rate, no draw rate, no average centipawn loss, no sample games with engine evaluations. For an opening product, that is a structural omission, not a minor detail.

The second is the most contentious and also the most dated. The claim that computers cannot handle long-term strategic positions belongs to the pre-AlphaZero era. Today's neural-network engines handle closed King's Indian positions far better than the marketing line implies. The phrasing "computers do not understand" is the marketing language of a bygone age, and it targets exactly the psychology of buyers who want to believe that human value lies in the regions machines cannot touch. That belief is not philosophically wrong. It is only technically wrong.

The third is directional. "Quickly dangerous for White" depends entirely on White's level of preparation. A well-prepared Bayonet player can still find play; equality is plausible, but equality is not the same as a solution. Across the hundreds of games I have coded for my long-form reports, I found what the results table never records: repetition. Moves that look new are often small variants of a structure that has appeared dozens of times before.

The biggest blind spot here is that the product presents a secondary branch as though it were the answer.

9...a5 is a known branch. Any White player who prepares seriously against the Bayonet has a ready response, because the move has sat in databases for a long time. Its real value is not novelty but practicality: it pulls the opponent out of a comfortable memory zone and forces him to think for himself. For a club player, that is a big advantage. For a well-prepared grandmaster, it is a small trap — and a small trap usually works exactly once.

There is another risk, and it concerns information. When I re-read the source package, the headline referred to a team-event round in Samarkand — the strong teams all winning — while the entire body dealt with an opening product against the Bayonet. The two are unrelated. A headline that does not match its body is an information-integrity signal, and anyone citing it must verify the original source first. In my work I have seen data packages spliced together like this, and the only way to handle them is to go back to the original. Without an original, any downstream analysis is just guesswork dressed up in numbers.

Then comes the human variable. The whole appeal of 9...a5 is that Black is freed from the pressure of memorisation. But where does that pressure go? It moves to White. With no forced sequence left, White has to decide at every move, and autonomous decisions burn clock. With a long time control that is a minor issue. With a short one, or in blitz, it can be a bigger problem than forgetting theory. There is a typical fourteen-second gap in every game — the silence after the opponent releases a piece and before you touch one. The 9...a5 branch stretches that silence across the whole game and turns it into a weapon for both sides.

And one more point: the product names no author, cites no tournament results for that author in the Mar del Plata itself, and offers no sample games with engine evaluation. An opening product that wants to be taken seriously needs at least those three things. Their absence does not prove the content wrong, but it places the content in an unverifiable zone. From my experience of tracking games, the products that last in club chess are always the ones bold enough to put sample games in the light so the buyer can check for himself.

What is worth keeping from this story is not whether 9...a5 is strong or weak. What is worth keeping is a verification question: when a move is sold as a solution, is it opening a new map, or merely shifting the pressure from memory to judgement? The answer only appears after the piece touches the board — and after the first game against a well-prepared opponent is over.

Every plan is a hypothesis until a piece touches the board. The space map never lies — it only exposes what we want to believe.