9...a5 and the Narrow Door of the King's Indian Defence: Notes from the Sideline of an Analysis Room

From the sideline, I see the whole game.

In the winter of 2026, in a small analysis room on the fourth floor of a chess club in Moscow, I sat at the back row, behind a young player. On the table there was only a wooden board, a notebook, and a mechanical clock set at an angle. The game followed a familiar order: 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Nf3 0-0 6.Be2 e5 7.0-0 Nc6. White played 9.b4. That is the Bayonet Attack. Black spent twenty-seven minutes on move nine, then placed the knight on h5.

I wrote the time down. Twenty-seven minutes for a move I had seen hundreds of times. The player hesitated not because he did not know the move. He hesitated because he knew exactly that after 9...Nh5 came a forced sequence of eighteen to twenty-two moves, where every deviation is punished immediately. He was weighing whether he had enough memory for that evening.

Three weeks later, in the same room, another player handled Black and answered with 9...a5. He took four minutes. The game lasted sixty-one moves and ended in a draw that both sides felt was chess, rather than a memory test.

The distance between twenty-seven minutes and four minutes, between two people sitting next to each other in the same room, is the subject of this article.

Context: a system built to concede and then reclaim

The King's Indian Defence belongs to the hypermodern family. Black deliberately lets White occupy the centre with 1.d4 and 2.c4, then builds a fortress on the last two ranks with Bg7, d6 and e5. Black does not contest the centre with pieces at the start. Black waits for White to finish the structure, and then attacks that structure.

The Classical branch leads to the Mar del Plata system, where Black places the knight on c6 instead of keeping it on f6. This is a major fork. With the knight on c6, Black prepares c7-c5 and opens the queenside, while retaining the option of f7-f5 on the kingside. The central structure becomes a locked chain: White holds d4 and e4, Black holds d6 and e5. Neither side can break the chain without paying.

When the centre is locked, the game splits into two parallel races. White pushes the queenside with c4-c5, b4-b5, brings the rook to c1 and then c2, the knight to a4 or e1, and finally breaks with b5-b6 or c5xd6. Black pushes the kingside with f7-f5, g6-g5, f5-f4, sends the knight from f6 to h5 or e8, the bishop to e6, the queen to e8 or f7. The race is counted in moves. Whoever opens a wing first wins.

This is why Mar del Plata was once called the opening of counters. Not counting pieces. Counting moves.

And this is why the Bayonet Attack appeared as a response.

The Bayonet Attack: a ninth move that changes the whole clock

9.b4 is a move with a very specific purpose. White does not merely gain queenside space. White changes the tempo of the race.

In classical Mar del Plata, White usually pushes b4 after completing development, after placing the rook, after preparing rook c2. Black has time to push f5, then g5, then f4, and to create the threat first. The Bayonet compresses that entire process. White pushes b4 while Black has not yet placed the bishop on e6, while the black queen does not know where to go, while the black knight has not chosen a square.

The consequence is that Black loses the right to choose tempo. Black must decide immediately about the kingside structure, because the queenside is already under threat. If Black is slow, b5 arrives, then b6, then an open file.

In other words, the Bayonet is a move that steals time. And when time is stolen, a Black player has two choices: accept the exchange in deeply worked theory, or look for a road fewer people travel.

9...Nh5 belongs to the first category. 9...a5 belongs to the second.

9...Nh5: the crowded road and its price

Placing the knight on h5 has been Black's main move for decades. The idea is clear: the knight leaves f6 to clear the way for f7-f5, while keeping the option of jumping to g7 or f4. Black opens the kingside at maximum speed.

In return, White gets a queenside that is almost unobstructed. After 10.c5, White opens the c-file and begins the process of breaking the d6 structure. Black must constantly calculate how many moves remain before the c-file becomes a highway.

The complexity of 9...Nh5 lies in the fact that it allows no margin for error. White can play 11.Nd2, 11.Ba3, 11.Re1, 11.b5, or 11.Ne1. Each choice leads to a different structure, and in many cases the difference between winning and losing lies on move seventeen or nineteen, when both sides have run out of theory.

That is why the Black player spent twenty-seven minutes. Not because he did not know 9...Nh5. Because he knew all too well that after 9...Nh5, the game would be decided by memory rather than judgement.

I spent six weeks in 2026 reviewing hundreds of games in this system. I noted four hundred and twelve situations in which the Black side fell into difficulty after following correct theory to move twenty. In most cases, the deviation did not come from failing to understand the position. It came from running out of time to understand the position.

An opening that consumes the clock in the early phase consumes defensive capacity in the later phase. That is a loan with interest.

9...a5: the strategic fork and the logic behind it

9...a5 does something very simple in form. It attacks the b4 pawn that has just advanced.

But its real value lies elsewhere. By forcing White to decide immediately about the b4 pawn, Black removes White's freedom to choose tempo on the queenside. White has four options: keep the pawn with a3, push on with b5, exchange with bxa5, or ignore it and switch plans. Each option leads to a different type of position, and in all four cases Black has a stable structure to play from.

The key word is stable.

In the Mar del Plata system, Black is usually dragged into a battle where every move can be refuted by a line analysed to move thirty. 9...a5 changes the type of battle. It turns the game from a speed race into a structural fight.

After 9...a5, Black usually continues with 10...c5 or 10...Be6, depending on White's response. If White plays 10.bxa5, Black can answer 10...Qxa5 or 10...Nxa5, and the queenside opens in Black's favour because White's a-pawn has left the a-file. If White plays 10.b5, Black can push 10...Ne8 or 10...Nd4, and the central structure becomes denser. If White plays 10.a3, Black pushes 10...c5 and begins hitting d4.

In every branch, Black keeps an intact pawn structure and a clear plan. That is the real value of the move.

Black's concrete plans after 9...a5

I recorded four main plans that Black typically pursues.

First, pushing c5 to hit the centre. When White keeps the b4 pawn, White's c-file is blocked by the c4 and d4 pawns. Black can use the c-file for the rook after exchanging pawns. This is active defence.

Second, bringing the bishop to e6 and then c5, or the bishop to d7 and then b6. In this structure, the black bishop has more diagonals than when it is locked on g7.

Third, bringing the knight from f6 to e8 and then d6 or c7, creating a multi-layered defensive block around the king. This plan is especially effective when White has pushed b5 and the queenside is locked.

Fourth, pushing f7-f5 late, after stabilising the queenside. This is the biggest difference from 9...Nh5. In 9...Nh5, Black must push f5 immediately. In 9...a5, Black can choose the moment. Black waits for White to commit on the queenside, then opens the kingside.

That delay has enormous tactical value. In chess, the side that goes first in a race usually has to accept that the opponent knows exactly what it will do. The side that goes second but chooses the moment retains the element of surprise.

That is the entire argument for 9...a5: trade speed for the right to choose the moment.

White's options and the unanswered question

If 9...a5 were simply a good move, it would have become a main line long ago. The fact that it remains on the edge of theory shows that White has options strong enough to keep a small edge.

The first option is 10.b5. White accepts a locked queenside in exchange for space and a fixed target on c5. In this structure, White usually switches to hitting d6 with c5 or transferring the knight to a4 to attack c5.

The second option is 10.a3. White keeps the b4 pawn and prepares b5 under better conditions. In return, White loses a tempo, and Black has time to push c5 and set up the defence.

The third option is 10.bxa5. White opens the a-file but also opens the way for the black queen. In many games, White chooses this when looking to simplify.

The fourth option, the least common, is 10.Ba3. White ignores the queenside and switches straight to attacking d6. This is the branch Black must prepare most carefully, because it turns the game into a genuine tactical fight.

In all four branches, the objective assessment usually tilts slightly towards White. That is something any opening product advertising for Black must acknowledge. The honest question is: how much, and in what way.

The data problem: what is not said

I keep a spreadsheet with twenty-seven columns for every game I watch. It includes a column for the share of moves matching the engine's top choice, a column for average centipawn loss, and a column for thinking time on each opening move.

When an opening product is presented, I look for those three figures first.

In the material I have on 9...a5, all three are absent. No engine match rate. No average centipawn loss. No sample size. No player named.

That does not mean the move is wrong. It means the claim that 9...a5 is technically superior to 9...Nh5 has not been demonstrated with data. It is an assertion, not evidence.

For a product aimed at club players, that may be acceptable. For an academic claim, it is not enough.

The engine-does-not-understand argument and its erosion over time

The central argument of this product is that the complex strategic positions arising after 9...a5 are difficult for computers.

I have heard this argument many times over twenty years. It used to be true.

It was true at the time of engines based on pure search. A position with a locked centre, two separated wings, no direct tactical blow, and no clear target for the next thirty moves was the kind of position where depth-first search struggled.

But the current generation of engines evaluates King's Indian positions reliably. Neural networks have learned that a locked central structure with two separated wings has its own value, independent of whether a direct tactical blow exists.

That means the engine-does-not-understand argument is losing weight every year. A product built on that argument will lose value at the same rate.

This is a strategic problem for anyone producing opening content. You can sell a move, but you cannot sell an argument that time is eroding.

Format and the shift of the industry

There is a detail that is rarely noticed in this whole story. The product is presented in DVD format.

I do not say that to criticise. I say it to point out a structural problem.

A DVD cannot be updated. When a strong player finds a refutation for a branch in the product, that branch stays on the disc. When an engine updates and re-evaluates a structure, that evaluation cannot be inserted into the disc. A static product competing against a discipline whose theory changes weekly is an unequal race.

Subscription platforms and interactive databases have changed how opening knowledge is transmitted. A course can be corrected within a day. A DVD cannot.

This does not make 9...a5 wrong. It makes the distribution channel a bigger weakness than the content itself.

A misplaced fragment of a title

In the material I received, one detail made me stop. The title referred to the European Schools Chess Championship, with a location in Armenia and a cultural festival. The rest of the content was about an opening product.

The two subjects are unrelated.

I raise this for a professional reason. When a document has a title and content that do not match, every conclusion drawn from it must be downgraded by one level of confidence. That is a principle I learned after years of working with match data.

If the title is correct, then we are talking about a junior event, about the development pipeline, about a school chess system. That is an entirely different story, and a far more important one than a DVD.

If the content is correct, then we are talking about an opening product for a very narrow readership.

I do not have enough data to decide. And when there is not enough data, the most honest thing is to say that I do not know.

Vietnamese chess and the King's Indian problem

In Vietnam, the King's Indian Defence has never been the first choice of the leading players. Le Quang Liem spent a long period playing open systems, where he controls tempo through simple pawn structures. Nguyen Ngoc Truong Son is known for tight games, where he prioritises structural safety over complication.

That does not mean Vietnamese players do not meet the King's Indian. On the contrary, in junior and domestic events the system appears frequently because it allows Black to seize the initiative without learning too much open theory.

And this is where 9...a5 becomes notable in a Vietnamese context. A move that allows Black to escape forced lines, keep a stable structure, and play by plan rather than by memory fits an environment where opening preparation time is limited.

But it also raises a training question. If young players learn to avoid complex lines before understanding why those lines are complex, they are trading a skill for comfort. That comfort can be paid for at a higher level, when the opponent forces them to calculate precisely.

I have observed this across many sports. Young people learn to avoid weaknesses before they learn to fix them. It is a reasonable shortcut in short-term results, and a debt in long-term development.

The blind spot of pragmatism

There is a paradox in this whole story that I want to point out.

The product is sold as a pragmatic solution. It promises to help Black escape forced lines, save study time, and focus on understanding plans. That is an appealing and honest promise in intention.

But pragmatism in chess has a limit. When you choose a move to avoid calculation, you do not remove the need for calculation. You move it to another phase of the game, where you have less time and less information.

In 9...Nh5, the Black player must calculate on move fifteen, when the clock is full. In 9...a5, the Black player may not have to calculate on move fifteen, but will have to calculate on move thirty-five, when the clock is short and the consequences are larger.

This is the blind spot of every pragmatic opening product. It does not reduce the workload. It reallocates the workload.

And for some players, that reallocation is right. For others, it only creates a sense of safety disproportionate to reality.

What I got wrong

In 2026, I wrote six articles before a major tournament and predicted wrongly about a team because I underestimated the depth of the bench and the coach's ability to read a match.

I mention it here because it relates directly to how I view 9...a5.

In both cases, my mistake had the same shape. I judged a system by what I saw on the surface, and ignored the operator's capacity to adjust. A move does not play itself. A plan does not execute itself. The value of both depends on the person sitting behind the board.

With 9...a5, that means the value of the move does not lie in the move itself. It lies in the Black player who understands the structure the move creates. If that player understands, the move is stronger than the engine's evaluation. If not, it is weaker than any evaluation.

Patience is not stillness

Patience is not stillness. Patience is waiting for the right tempo of the opponent's pressing.

In chess, that tempo is the moment the opponent commits to a structure. After 9.b4, White has committed. Black has not. 9...a5 is a way of extending the period in which Black still has a choice.

That is why I find the move interesting, regardless of whether the product's technical claims hold. It represents a school of thought: extend the options rather than seize an advantage immediately.

This school wins at club level, where opponents are often not patient enough to punish the delay. It wins less often at the top level, where every tempo is counted.

People watch players run. I watch the whole block move.

In chess, the block is the pawn structure. People watch 9...a5 and see a pawn move two squares. I watch 9...a5 and see the whole queenside structure change: the b4 pawn loses its footing, the a-file opens, the c4 pawn loses a potential support point, and the g7 bishop gains a diagonal.

One move on the edge of the board changes the value of five pieces in the centre. That is the kind of movement I care about.

In many games I reviewed, I noticed Black did not lose because of choosing the wrong ninth move. He lost because after choosing the right ninth move, he played the next twenty moves as if the structure had not changed.

The system does not lie. But it can only be heard when the data is thick enough. And in this case, the data is not thick.

A note on honesty with data

I learned a lesson in 2026. I wrote a long analysis of a club's pressing system and was fiercely criticised. Instead of arguing, I spent six weeks reviewing all the footage and recounting every situation.

The result did not fully support me, and did not fully refute me. It was more precise than both. I published a correction with specific figures, and since then I always write in the form of hypothesis, verification, conclusion.

I apply the same principle to 9...a5.

Hypothesis: 9...a5 gives Black a stable structure with fewer forced lines than 9...Nh5.

Verification: this requires a database of games at 2500 and above, with win rate, draw rate, and average centipawn loss for both branches.

Conclusion: not yet possible, because the data does not exist.

That is the most honest conclusion I can offer, and I accept it.

My first article was stoned. Data never takes offence.

I am sixty-nine. I still learn from the young people sitting in the back row of the analysis room, the ones who take four minutes on a move my generation took twenty-seven minutes for.

Perhaps they are right. Perhaps reducing forced theory is a genuine advance, and people like me are clinging to an outdated standard. Perhaps they are trading a skill for comfort, and the price will arrive at a stage they cannot yet see.

I do not know. And I think saying I do not know is part of this profession.

Takeaway: what to verify next season

9...a5 in the King's Indian Defence is a valid move, with clear logic, and a group of players it suits. It is not a theoretical revolution, and it has not been shown to be better than the main move 9...Nh5.

Four signals I will track next season.

First, the actual win rate of 9...a5 at 2600 and above. If it is positive, the move will shift from the edge to the centre of theory.

Second, its frequency in major events. A move only exists when someone plays it where there is an audience.

Third, the number of games in which White chooses 10.Ba3 or 10.b5 and gains a clear advantage. That is the most direct measure of the move's defensive value.

Fourth, and most important to me, the question of skill. Whether a generation that learns openings through low-forcing moves retains the ability to calculate precisely when it reaches the endgame.

I will record every game in my twenty-seven-column spreadsheet. If after twelve months the data shows I am wrong, I will write a correction. I have done it before, and I will do it again.

Seven months without football, seven months of asking why. I learned that a gap is not the enemy of analysis. It is the condition that makes analysis more honest.

And in the data gap around 9...a5, I choose to keep the question rather than answer it myself.

Moscow 2026, people remember the goals. I remember the empty space on the right flank. In chess, that empty space is the b4 pawn that has just lost its footing.