The Quiet Arithmetic of 9...a5: Pricing a Sideline Against the Bayonet Attack

The clock read 21 minutes. In September 2026, during the 43rd Chess Olympiad in Batumi, one page of my notebook held this position — 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 Bayonet Attack. For White this is comfortable ground: a closed centre, space on the queenside, a king-side storm still several moves away. Then Black played 9...a5. The engine bar barely moved. The arithmetic on the other side of the table changed completely.

That night I wrote in the margin: tactics are spells with pressing triggers; the wizard reads the fine print. The fine print of 9...a5 is that it claims nothing in the evaluation. That is exactly why it is uncomfortable.

The King's Indian Defence is a transaction. Black surrenders the centre and buys, at interest, the right to attack it later. The Mar del Plata Variation is the sharpest version of that bargain. The centre locks at 8.d5, and then two storms run on two wings: White expands with b4-b5 and c4-c5 on the queenside, Black arranges ...f5, ...g5, ...Nh5, ...Nf4 in front of his own king. The Bayonet Attack, 9.b4, is White's most direct instrument: it takes permanent queenside space, prepares the c4-c5 break, and slows Black's habitual kingside expansion by a tempo. Whatever is fashionable at this moment is almost always the shadow of elite practice, and the Bayonet is no exception.

Opening theory is a fashion cycle, and somebody is paid at every turn of it. Elite practice creates fashion, fashion creates demand for an antidote, and demand for an antidote creates the instructional-media market. The source I am working from is a product of exactly that market: a repertoire course recommending 9...a5 against the Bayonet. It names no player, no rating, no sample game, no engine evaluation, and leaves the presenter unidentified. So I keep two things apart here: an opening idea, and a marketing claim.

In 2026, filing my first report from the press room at Purana Paltan, I learned that you read a press room by counting what nobody asked. In chess journalism one of those unasked questions is: who pays for preparation? Niaz Murshed became South Asia's first grandmaster in 2026; Bangladesh has produced no new grandmaster since Enamul Hossain Razib in 2026; the federation's annual budget sits near Tk 11 lakh; and when Standard Chartered's school chess programme ended in 2026, nobody filled the gap. The ability to buy a foreign course and a preparation pipeline are not the same thing. A notebook in Batumi can hold a World Cup if you fold it right; an annual budget of Tk 11 lakh leaves nothing worth folding.

9...Nh5 is the main line. It is forcing and tactical: after 10.Re1 f5 11.Ng5 Nf6 12.Bf3 c6 13.Ba3 a long, nearly forced sequence unfolds in which Black's kingside storm and White's queenside counter-attack run side by side. Memory is close to decisive there; whoever retains more lines stands better.

9...a5 works on a different argument. The move creates no forcing sequence — that is both its chief profit and its chief risk. The pawn on a5 questions b4 directly. White has two natural answers, and each builds a different future.

First answer: 10.bxa5. The b-file opens, White's queenside majority shrinks, and Black can later drop a rook onto the a-file. As long as the pawns stay fixed, White's extra space is really a target list — a2, b2, c4, d5, four locked pawns, four candidate weaknesses. In the endgame Black's rook enters down the open a-file; White may be ahead in the king-and-pawn race and behind in the rook ending.

Second answer: 10.b5. The queenside closes, White keeps permanent space, and Black gets a free hand on the kingside. White's only real lever is then the c4-c5 break — and if it does not come at the right moment, White owns big space and a small plan.

The trade-off sitting between those two answers cuts both ways. 9...a5 is a tempo slower than the main line; Black spends one move on the queenside before starting his kingside storm. In exchange he renders White's memorised warehouse useless. In a preparation war that is a large gain: when your opponent arrives at the board carrying a 23-move line and you leave it on move four, the clock starts playing for you. In 2026 at Purana Paltan I diagrammed 23 positions from a single ninth-round game; my editor cut all 23 and kept 400 words of result. Since then I have not opened with a result — I look first for the square where the game turned. 9...a5 is that kind of square-defining move: b4, a5, and later the four squares of the a-file.

What is missing beside the idea deserves equal clarity. There is no sample game, no ACPL, no win-draw rate, no engine evaluation. "Objectively equal" and "dangerous for White" cannot both hold, unless you read them practically: the result is theoretically balanced and practically uncomfortable for the unprepared White player. Every opening claim is a hypothesis; the engine evaluation and the database are its p-value. This file contains no p-value. A claim with no database beside it is not analysis, it is advertising.

The heaviest technical claim in the source is that computers often cannot handle these complex strategic positions. Standing in 2026, that claim no longer holds. NNUE-based Stockfish and Leela Chess Zero evaluate strategic manoeuvring, pawn breaks and king endings superbly; a closed King's Indian centre is not a blind spot for them.

What is revealing is that the source contradicts itself: it argues engines cannot understand these positions, then concedes that once you return to the main 9...Nh5 lines, computer analysis matters again. The honest version is probably this — the 9...a5 lines are less forcing, so a single engine move cannot refute you there. That is a statement about human practical discomfort, not engine incapacity. The difference is not small: the first is bad information, the second is a useful insight.

And the marketing archetype being sold here — understand, do not memorise — is evergreen among adult improvers. The product probably works, because it targets the player exhausted by memorising the 9...Nh5 tree and tells him there is nothing to memorise. But where there is no preparation budget, buying a course on understanding does not build a pipeline. That limit is not the course's fault. It is the budget's.

What to watch in the coming months: the frequency of the Bayonet Attack in elite databases, and NNUE evaluations of the critical 9...a5 lines. If the Bayonet's fashion fades, the antidote's market price fades with it — because this is not a refutation, it is a tenant of a cycle. And I trust a system that can explain its own failure in the 89th minute; that explanation for 9...a5 — what happens if White plays every best move — has not been written yet.