The 9...a5 Move in the Mar del Plata: Asymmetric Preparation and a Data Gap

Two forty in the morning in Guangzhou. The electronic board on my screen stops after move nine: 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. Beside the board sits an opening DVD just delivered to the newsroom, titled "A night unlike any other", promising that Black no longer needs to memorise 9...Nh5, that a single move, 9...a5, is enough to neutralise the Bayonet Attack.

I have heard that line before. People call it a shock; I call it data that has not been read yet.

That night I stayed at my desk, not to argue, but to check how much of that pitch could survive an independent verification. The result is this piece: a genuine defensive option, wrapped in a promotional frame that has never been tested.

Context: a city that gave the variation its name

Mar del Plata is not the name of a player. It is the Argentine coastal city where the 2026 tournament brought the King's Indian Defence to the highest level, and the variation has carried the city's name ever since. Svetozar Gligorić won that tournament, and he was one of the players who brought the plan of ...Nc6 followed by ...Ne7 into top-level practice. More than seventy years later the structural essence is unchanged: Black concedes a controlled centre, masses pieces on the kingside, opens the g7 diagonal, and accepts a race — whoever breaks through on the opposite wing first gets to talk about victory.

The Bayonet Attack, the move 9.b4, arrived much later. Through the 1990s, as leading players searched for tools to crack Black's closed setup by grabbing space on the queenside as early as move nine, 9.b4 became White's principal choice. It is not a pretty move; it is a purposeful one. It pushes the pawn to b4, prepares b5 to drive away the c6 knight, and forces Black to decide immediately: close the centre with 9...Nh5 and storm the kingside along classical lines, or take another road.

What makes this position remarkable is that both sides are playing against an invisible clock. At 2400 and above, the difference between two players is rarely understanding of the position; it is preparation time. White can arrive at the board with thirty moves memorised; Black must know three times as much. That is why the opening-product market exists, and why it is full of promises that are hard to verify.

Core: four layers of verification, and what the material does not say

Every opening product I receive passes through four layers. Layer one: does the proposed move exist in competitive practice. Layer two: if it does, how does it score among strong players, on a sample large enough not to be noise. Layer three: does the material acknowledge White's alternatives, or does it present only the one line it wants. Layer four: who is the author, and how many reference games are named.

For this product, the first two layers have answers. The move 9...a5 is real, has appeared at the top level, and belongs to the minority-but-legitimate group of options in the Classical King's Indian. The last two layers are empty. Not one reference game is named. Not one statistical figure is given. Not one engine evaluation appears. No author name, no publication date.

I checked three sources before writing this line. Source one is the product description. Source two is a forum post quoting the description verbatim. Source three is a video review quoting the forum post. Three sources, one origin. This is the trap I remind myself of every week: counting sources is not verifying origins. Three copies of the same advertisement are still one advertisement.

So where does the real value of 9...a5 lie? The answer sits inside the material's own admission: objectively, the chances are equal. A balanced move cannot be an antidote. The value of 9...a5 lies not in objective advantage but in asymmetric preparation and in surprise. That is a different argument altogether, and far more modest than the product's title.

Strategically, 9...a5 pulls play to the queenside before White can shut the kingside. It turns the b4 pawn into a fixed target, prepares ...a4 or ...Nc6-a5, and retains the option of returning to ...Nh5 later. Most importantly, it delays the choice of central structure, keeping both the queenside plan and the classical Mar del Plata plan in Black's hands. It is a move for someone who wants to play chess, not for someone who wants to look up a table.

But one thing the description omits must be said plainly. A Black repertoire built on 9...a5 still requires two things that cannot be skipped. First, a fallback for White's alternatives — 9.Ne1, 9.Nd2, 9.Bg5 and similar. Second, knowledge of the main line 9...Nh5, because after 9...a5 White can steer the game back into familiar structures. The preparation cost does not disappear; it relocates. The claim that only very few concrete lines must be remembered is the highest-value and least verifiable claim in the entire material.

Another gap: the product's scope is undefined. Does it cover only the 9.b4 branch, or the whole King's Indian after 7...Nc6? That changes the buyer's learning cost completely, and it is never stated.

Here I have to address the central argument of the promotional material: that computers often do not know how to handle the complex strategic positions arising from this move. That argument has an age. It was born in an era when breadth-first engines still stumbled in closed positions, where a move's value lies in structure rather than material. From 2026 onward, the neural-network generation of engines — Stockfish with NNUE, Leela Chess Zero — evaluates closed, manoeuvring positions far more reliably than the generation before it. Stockfish's NNUE build shipped in mid-2026, and the gap has only widened since. A balanced position such as the one after 9...a5 is no longer a machine blind spot. It is a human blind spot — and that is the genuine selling point. Age is the one variable that never lies, and the argument that computers cannot understand is twenty years old.

Numbers are asceticism: you must give up convenience before you can see the truth. Here, convenience is believing the title. The truth is: a legitimate minority option, sold with no supporting data, by an argument that has gone stale.

The contrarian angle: asymmetry cuts both ways

If 9...a5 wins because of surprise, it also loses because of surprise — or more precisely, because surprise runs out. This is the point opening-product sellers rarely make.

A minority move has value only for as long as opponents have not met it. Every top-level game is recorded, every database updated monthly. The life cycle of an opening surprise today is far shorter than in the 1990s, when information still travelled through print magazines. Whoever buys this product is not buying a permanent advantage; they are buying a window, and that window closes faster if the move genuinely works — because what works gets copied.

There is a deeper paradox. If 9...a5 really is balanced and awkward for White on strategic grounds, it will be mined and become a main line. It then stops being a surprise weapon and becomes part of general theory — and this year's DVD buyer will have to buy another one next year. If it is not mined, then it most likely stays a minority option for some technical reason nobody has openly demonstrated. Both branches lead to the same conclusion: the product's value lies in the moment of purchase, not in the nature of the move.

As for the club-player argument — suitable for those who prefer thinking for themselves over memorising long variations — it should be read correctly. It is market positioning, not a technical property. It targets players with little study time but plenty of ambition, a large and loyal group. For them the promise has real value. But it does not change the fact that a minority move still needs a supporting system behind it. The transfer market is where people pay for numbers that have not been verified; the opening-theory market runs on exactly that logic.

What would change my mind? A clear, falsifiable condition. If over the next twelve months the database of top-level games shows 9...a5 scoring for Black on a par with, or better than, 9...Nh5 across a sample of at least two hundred games among players rated 2400 and above, I will upgrade my assessment of the move. If the sample is a few dozen games, that is noise, and I will not write a hymn out of noise.

xG does not replace emotion; it explains why our hearts race. In chess, the equivalent of xG is the conditional win rate by rating band and the number of games in the database. It does not make a game less thrilling; it merely forces the writer to state what they are standing on. Based on my experience covering games across many tournament cycles, the opening products that last are not the ones that promise the most, but the ones willing to publish their reference games.

Takeaway: signals for the next cycle

I will track two signals in the coming cycle. First, the appearance rate of 9...a5 in top-level databases after twelve months — if it rises, the surprise story is being validated; if it stays flat, this was a product sold at the right moment. Second, and more important, whether the seller publishes a list of reference games. A repertoire can be sold by a title, but it only becomes theory when someone can name the games that proved it.

The move 9...a5 will remain on the board longer than the DVD that put it in a title. My job is to record both, and to record the date correctly.