The Bayonet Attack 9.b4 and the 9...a5 Move: Deconstructing a King's Indian Repertoire Product With No Verifiable Data

A very loud name, a very large hole

People called it a shock; I call it data that has not been read. The opening product titled “A night unlike any other” entered the market with a narrow and clear promise: 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 9.b4, Black can answer with 9...a5 instead of the main line 9...Nh5, and the player is not forced to memorise long variations.

What made me stop was the structure of the evidence, not the move. In the entire document I could access, every technical statement carries no source, except three points explicitly marked as the author's opinion. No player is named. No database is named. There is no engine evaluation for any move. No reference game. No publication date. No author credentials.

An opening package sold as an antidote to one of the most heavily analysed systems in modern chess, carrying not a single measurable quantity. That was the first anomalous signal — and for me the largest one. A product can be good. A claim cannot verify itself.

The title “A night unlike any other” almost certainly plays on the image of a nighttime bayonet assault, the direct metaphor of the Bayonet Attack. A pun, not an information headline. I record it as a signal of intent: sell the feeling first, sell the variation later.

Context: where the Bayonet Attack sits on the map

The Mar del Plata Variation of the King's Indian Defence is among the most hostile territory in modern chess. White closes the centre with 8.d5, then pushes 9.b4 to open play on the queenside. Black counterattacks on the kingside with ...f5, ...g5 and a rolling pawn advance. Both sides attack opposite wings, and in many lines a single tempo changes the result.

9.b4 is the Bayonet Attack. It became a frequent choice for players who want to avoid the theoretical race of 9.Ne1 or 9.Nd2. The early pawn push gains space and forces Black to commit to a pawn structure before completing development. For White, it converts the game from memory into understanding. For Black, it is a test of preparation.

The main line 9...Nh5 goes straight into contact. The knight steps to the rim, Black threatens ...f5, and everything becomes forcing: one move calls the next, one variation locks into another. That is why engines handle this branch relatively well, and why it devours memory.

9...a5 is different in kind. No contact, no forcing sequence, no immediate must-play chain. Black announces a plan: pressure the queenside pawn structure, keep the centre flexible, and wait for White to create a weakness. A move of this type only means something to a player who understands the structure, not to one who only remembers lines.

To read a commercial opening package correctly, I always separate three layers of value: objective soundness, surprise factor, and the preparation cost borne by the opponent. A product can win on the second and third layer while remaining neutral on the first. Confusing these layers is the most common mistake made by buyers of opening knowledge.

Where 9...a5 sits on the technical scale

When I break a variation into metrics, I am obliged to state which metrics have no data. For 9...a5, the result is as follows:

| Metric | Assessment | Comparison | Note | |---|---|---|---| | Sophistication | An existing sideline, not an invention | Versus the main line 9...Nh5 | Marketed as an antidote; in practice a repositioning of a minority branch | | Engine match rate | No data in the source | — | No engine statistics of any kind for any move | | Practical stability | Claimed high by the source | Versus classical time control | Plausible for a well-prepared Black; the “very few concrete lines” claim is likely overstated | | Key data | None | — | The largest verification gap in the entire document |

On sophistication, 9...a5 belongs to the established-alternative group, not to invention. The promotional material calls it an antidote to the Bayonet Attack. That framing obscures the fact that the move has long existed in theory. Repositioning a minority branch as a commercial product is ordinary market behaviour. What is not ordinary is claiming discovery.

On engine match rate, the source supplies nothing. No evaluation from any engine, no game count from any database, no average centipawn loss, no search depth.

On practical stability, the source claims a high level, arguing that the move is strategic and requires few concrete lines. That argument has a basis for a well-prepared Black player. But a basis is not evidence, and the quantitative layer is entirely empty.

On key data, the number that should exist is precisely the number that does not: no games, no win rate, no time usage, no author name.

The most important point is that the author himself concedes that objectively the chances are equal. That concession redefines the product's entire value. The advantage lies in preparation asymmetry, not in objective superiority. In other words, the product sells surprise, and it sells exactly that.

The transfer market is the place where people pay for numbers that have not been verified. The market for opening knowledge runs on the same logic, differing only in currency.

The engine claim has aged out

Age is the one variable that never lies.

The document's central premise — that computers often do not know how to handle the arising complex strategic positions — belongs to the pre-2026 engine generation. Since engines moved to neural networks and NNUE, the evaluation of closed, manoeuvring positions has become far more reliable. The anti-computer argument has therefore lost most of its weight.

The remaining value is real but belongs to a different category: humans under time pressure still misjudge closed positions. Two kinds of advantage need to be separated clearly — advantage over the machine and advantage over the human. The document merges them, and that is where the reasoning softens.

Based on my experience tracking games, I learned this lesson the expensive way. In 2026, when I analysed a major match through data rather than emotion, I was dismissed on the grounds that numbers cannot tell a story. That article went on to exceed two hundred thousand reads. Since then I apply one rule: every conclusion must carry sourced data, and every data point must survive the question of provenance. For this document, that question has no answer.

The product's scope is undefined

A decisive question remains unanswered: does the product cover only the 9.b4 branch, or the entire King's Indian complex after 7...Nc6?

That difference determines the real preparation cost. If it covers only 9.b4, the buyer must still build his own response to White's alternatives such as 9.Ne1, 9.Nd2 and 9.Bg5. The buyer also still needs knowledge of the main line 9...Nh5, because White can steer the game back toward familiar structures after Black plays 9...a5.

If the scope is the whole complex, the document should list every covered branch, along with the branches deliberately excluded. Either way, the same demand follows: the source must be explicit.

The real preparation cost is therefore far larger than the claim of “very few concrete lines”. That does not make the product useless. It only means the promise and the delivery do not match.

Three sources, or one source counted three times?

My discipline is to cross-check three independent sources before concluding. But quantity is not independence. If all three sources trace back to one original number, I have one source, counted three times.

With this document, the count returns something unusual: the number of independent sources is zero. No engine data, no database, no reference games, no author profile. The only three points with any attribution are the author's own opinions — that is, the single primary source of the document itself.

That is why I can neither confirm nor refute 9...a5 on technical grounds. I can only classify it: a minority branch, theoretically grounded, presented with a verification level of zero. Classification is the job of data. Judgement I leave to the board.

The counter-intuitive angle: surprise mistaken for superiority

If 9...a5 works, the effect most likely does not come from the strength of the move but from the opponent never having faced it. Correlation is not causation. This is the point every commercial opening package wants the buyer to skip.

The consequence is that the product's value decays over time. The more people buy and play 9...a5, the more White players prepare for it, and the thinner the surprise advantage becomes. This is a general property of any commodified minority sideline: the seller collects once, the buyer carries a long decay risk.

I must also state the falsification condition for my own assessment, because a conclusion without a falsification condition is merely a belief. My assessment — that 9...a5 is a legitimate sideline but not an objective advantage — would be wrong if one of three signals appeared. First, a large database showing Black's win rate after 9...a5 is meaningfully no lower than after the main line 9...Nh5. Second, an engine evaluation stable across depths without degradation. Third, a published top-level game of convincing technical quality. Until then, I stand on the cautious side.

Takeaway: signals to watch in the next cycle

Numbers are asceticism: you must give up convenience before you can see the truth.

Before paying for any opening package, I check three things. First, scope: does the source state clearly which branches it covers and which it excludes. Second, data: are there reference games, database rates, engine evaluations at a specified depth. Third, authorship: does the seller have a name, a playing or coaching record, something that can be publicly cross-checked.

For a club player who wants to diversify his opening play, 9...a5 may still be a worthwhile addition, provided he accepts carrying the remaining preparation cost himself. For someone looking for an antidote, the document has not produced the evidence required to justify that word.

The larger lesson sits in the market. Opening knowledge is being packaged and sold as a consumer product, while what buyers actually need is the ability to verify. In the next cycle I will watch two signals: the number of real games appearing in large databases once this branch spreads, and whether the next wave of opening packages ships with verifiable data or continues to run on feeling. When a very loud name carries no number at all, the correct reading is that it is not a conclusion but the starting point of an investigation.