King's Indian: A DVD Sells 9...a5 as the Antidote to the Bayonet Attack

I opened the product page on a late weekend evening, after rewatching a few Bayonet Attack games to cross-check variations. The headline at the top read: "Samarkand R2 (open): Despite close calls, all top teams grab wins." The body underneath was about a DVD teaching the move 9...a5 against 9.b4 in the King's Indian. Two texts, two subjects, one display frame.

Nobody checked. That is the detail I am keeping before I discuss a single move. A headline about a team event in Samarkand, a body about an opening product sold to club players. When the metadata layer is already misaligned, every number beneath it has to be read again with suspicion.

I have worked in data long enough to know one thing: a source that breaks on the first line cannot be saved on the last. And I am old enough not to call this a small incident. It is a signal.

Mar del Plata and the crutch called Bayonet

The King's Indian Defense (1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6) is a hypermodern system: Black surrenders the centre early in order to demolish it later with ...e5, ...c5 and counterattacks on both wings.

The Mar del Plata variation — named after the 2026 tournament in Mar del Plata, Argentina — pushes that story to its sharpest point. Black puts a knight on c6, pushes the e-pawn, White advances d4-d5, and both sides sprint into a pawn race on opposite wings. Whoever arrives first wins.

For decades this was the weapon of players who like to calculate. Garry Kasparov is the name most tightly bound to it at elite level.

Then came the Bayonet Attack. The ninth move 9.b4 is a small but irritating knife: White pushes the b-pawn to gain queenside space, threatens b4-b5 to evict the knight, and turns the pawn race into a problem Black must answer with precision on every tempo.

From the mid-1990s, 9.b4 became a serious choice for White players who wanted to avoid heavy theory. Vladimir Kramnik, later Teimour Radjabov and Levon Aronian — on both sides of the board — kept the Bayonet in circulation at top events. Black's main reply is 9...Nh5, the knight going to the rim, accepting structural concessions to open the position.

That is why a product selling an "antidote" to the Bayonet with 9...a5 deserves a close look. Not because it is meaningless, but because the label on it is too pretty.

The evidence chain and the data gap

9...a5 is not a novelty. In the King's Indian theory tree it is a known sideline: Black pushes the a-pawn to challenge White's queenside structure, slows down b4-b5, and deliberately shifts the game from a pawn race into a closed, grinding fight for squares.

The difference between 9...Nh5 and 9...a5 is one of character, not of value. 9...Nh5 opens the position, forces concrete lines, and pushes both sides into a stream of moves that computers handle better than humans. 9...a5 does the opposite: it drags the position into a grey zone where planning matters more than a table of variations.

The product markets exactly that point. Its message compresses into one sentence: if Black knows what he is doing, things can quickly become dangerous for White.

Pleasant to hear. But this is where I stop and do my job.

In the description there is not a single metric. No ACPL. No Black win rate in practical games. No draw rate. No sample games. No engine evaluation. No author name, title, or record in the King's Indian.

A theory product sold to club players that does not include one complete game is a notable thing. Not because it is wrong. Because it has not been tested.

It took me three months to learn that a pretty chart is no substitute for a correct process. The correct process here has four steps.

The first step is frequency. If 9...a5 is genuinely the answer, it must appear in elite games at the 2700+ level often enough to produce a sample. Without frequency there is no trend, only anecdote.

The second step is symmetry. An antidote holds its value only until the other side reads it. The Bayonet is a branch analysed so thoroughly that every move has a name. A plan that appears within a few months will draw a response. That is the law of the theory market.

The third step is provenance. This product has the shape of a ChessBase-style Fritztrainer DVD — a teaching format that has existed for decades, with the characteristic "antidote to a popular line" register. It competes directly with free databases and with the engines themselves. In that competition, what sells is not the variation. It is the telling.

The fourth step is the technical premise. The message that "computers often do not know how to handle these positions" belongs to a period that has passed. Before AlphaZero in 2026, the argument had some force. After it, it does not. Neural-network engines today assess long-term King's Indian positions far better than the marketing sentence implies.

When the data does not lie, we are the ones lying to ourselves. An outdated sentence about computers does not devalue 9...a5. It only tells you who the product was written for, and when.

One more thing needs saying: the 9...a5 branch is not bad. Closed, square-by-square struggles suit players with a plan. If you are the type who wants to understand why you play a move rather than memorise thirty of them, 9...a5 is a sensible practical choice.

What I object to is the packaging, not the move.

There is a point worth putting on the table: 9...a5 takes away part of Black's motivation. The Mar del Plata is attractive precisely because it gives Black the right to attack the kingside. Choosing 9...a5 means trading an attacking weapon for solidity. For a weaker Black, that trade is usually a bargain. For a strong Black, it is avoidance.

I once sat in a meeting room in Shenzhen presenting a striker's numbers — 22 goals, but actual output nearly twenty percent below expectation — and the club changed its entire attacking system because of one page. A Chinese club taught me that data is not the destination; it is a walking stick.

That stick works on one condition: it supports you only when you know where you are standing. A DVD that sells a plan without selling evidence leaves the buyer to find his own footing.

That is why I read this description with the same attitude I once applied to football scouting reports: verifiable facts first, elegant prose second.

When an antidote becomes a target

This is the basic paradox of every theory product: its value is inversely proportional to its reach.

The day 9...a5 is known to a small group, it is a weapon. The day it appears in a widely sold DVD, it becomes a target. Your opponent buys the same product.

The transfer market is not a chess game; it is a synchronised routine performed by thousands of algorithms. That image applies to the theory market too: every antidote gets read, every line gets rebutted, and rebuttals arrive faster each cycle.

There is another test worth running: the "low theory" label. The Mar del Plata is by nature a sharp system. When pawns race on opposite wings, no line is absolutely safe. Telling a buyer they will not have to calculate is a hard promise to keep.

And this is where my doubt is deepest. A product promising "few concrete lines" for a system whose excitement comes precisely from concrete lines.

The data is a mirror; only the person willing to face himself sees the truth in it.

The real risk for the buyer is not losing one game. The risk is spending three months learning a plan, then discovering that your opponent has finished learning it too.

There is a second layer of risk, less often discussed. The buyer pays for a product that duplicates what is already free in databases and public online studies. The only originality left is the structure of the presentation and the author's voice. That is a legitimate way to earn money, but it is not original research, and the buyer should know what he is paying for.

I say this from the position of someone who has sold similar research. Three years ago I was on the other side of the table. That experience taught me that the difference between a good product and a poor one lies not in the number of variations, but in whether the buyer can verify it afterwards.

With 9...a5, the current answer is: no.

Signals to track in the next round

So I will not conclude about the move. I am scheduling four signals instead.

The frequency of 9...a5 in classical games among the 2700+ group. The frequency of 9.b4 in the opening tree. The identity of the product's author. And the original Samarkand report — because if a headline and a body can drift apart in one display frame, they can drift apart in ten others.

After 2026 I stopped trusting predictions. I only trust early-warning systems. With 9...a5, the early warning has a concrete name: whether it survives classical time-control games between two seriously prepared opponents.

If it does, this branch will have a cycle. If it does not, it returns to its old place: a practical choice for club players, not a revolution.

I will still buy the product. I will just buy it after I find out who wrote it.