King's Indian Defense: 9...a5, the Bayonet Attack, and the Data Gap of an Opening Product

Shanghai, 02:14. The temperature outside held at 31 degrees, August refusing to retreat even in the small hours. On the second monitor, an analysis file arrived. The title announced Round 2 of a team event in Samarkand: close calls, all top teams grabbing wins. I opened it.

Fourteen information points. Not one line mentioned Samarkand. No team. No pairing. No result. The entire body concerned something else entirely: a chess DVD teaching 9...a5 in the King's Indian Defense, Mar del Plata variation, when White plays the Bayonet Attack 9.b4.

Thirty-two years in the VTC commentary booth, I had seen every kind of error: wrong Elo, wrong date, wrong piece colour. But a document betraying itself in its first line — that was new. And in my trade, new is always worth more than pretty.

The board first, the headline later

The King's Indian begins 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7. Black surrenders the centre at move two, builds a fortress on the kingside, then counterattacks with ...e5 and ...c5. This is the system I call the martial art of the patient: White takes space, Black takes time, and by move thirty the two have traded places.

The Mar del Plata is the heart of that system. Black plays ...Nc6 and ...e5, White pushes d5, the centre locks, and both sides launch tank columns on opposite wings. White attacks the queenside, Black attacks the kingside, and whoever arrives first wins. By my own classification, the variation sits within ECO E97 to E99 — a personal filing choice, not something the source text states.

For decades, Black's main answer to the Bayonet 9.b4 was 9...Nh5. The knight jumps to the rim, menacing f4, forcing White into concrete replies. That is the classic forcing path: every move has consequences, every variation has an answer, and every answer was written into the books long ago. Long ago meaning very long ago — thousands of grandmaster games have travelled that corridor.

The product this document promotes takes a different road: 9...a5. A queenside pawn advances, threatens White's b4 pawn, and steers the game into a closed, strategically dense struggle rather than a forced sequence. That is the pitch — fewer concrete lines, less memorisation, ideal for a player who wants to think on the board rather than recite someone else's memory.

It is a clean marketing idea. But it stands on a foundation I need to inspect.

The number that is not there

A serious opening product in 2026 normally ships with at least four things: average ACPL across sample games, a win rate for the relevant side drawn from a database, an author's name with their own record in the line, and a few sample variations with engine evaluations.

Fourteen information points supply none of them.

No ACPL. No sample games. No author. No engine assessment. No win-draw-loss statistics. The only claims are that chances are equal and that 9...a5 leads to "sound positions with very few concrete lines".

I read that three times. Then a fourth time the next morning after four hours of sleep. On the fourth pass I understood: this is a positioning claim, not a technical one. It does not say 9...a5 is stronger than 9...Nh5. It says 9...a5 is easier to live with. Those are very different statements.

I once trusted feeling, until a number knocked at 3 a.m.

The number knocking this time was an absence. Not a wrong number — a missing one. And in analytical work, a silent gap is often the most important data in the whole file.

What 9...a5 actually does on the board

To be fair to the product, I must be clear: this is not a trivial move. 9...a5 is a known branch in professional circles, not an invention. It targets White's freshly advanced b4 pawn, and the logic behind it is coherent.

White has just played 9.b4 to expand on the queenside, drive the black knight off c4 and claim space. Black replies ...a5, striking directly at the pawn that just moved. If White pushes b5 to keep space, the pawn structure shifts and Black gains the a-file. If White trades on a5, the queenside opens in a way that favours Black's rooks more than White's. Black is not trying to prove 9.b4 wrong. Black is simply asking White a different question from the one White prepared for.

That is the oldest strategy in the second player's arsenal: change the question.

The character difference lives here. After 9...Nh5 the game detonates. After 9...a5 it contracts. Behind 9...Nh5, White must calculate precisely for ten moves in a row. Behind 9...a5, White can choose between several plans, and the cost of an imperfect move in the first three is far lower.

For a club player, that means survival. For a grandmaster, it means handing the opponent the stage.

That is why I say 9...a5 has real value — but the value sits at the user level, not the theory level.

What the data table shows

In the opening-tracking file I have kept for ten years, there is one indicator I always check before believing any opening pitch: the age of the claim that "computers cannot handle this position".

This document uses exactly that claim. It says 9...Nh5 "opens the position, forces the lines, and makes computer analysis important again", while 9...a5 "leads to sound positions with very few concrete lines" that "computers often do not know how to handle".

The framing sets up a human-versus-machine axis. The winner in that frame is the strategist; the loser is the tool-dependent player.

It reads pleasantly. But it is the language of the pre-AlphaZero era. Since neural networks replaced handcrafted evaluations, engines handle long-term strategic King's Indian positions far better than this sentence implies. A position without many forced moves no longer means the machine misplays it. The machine simply speaks a different language there, and that language still calculates every threat twenty moves ahead.

I understand why the old claim persists. It sells. It tells the buyer: purchase this and you become human at the board again. But when a friend sent me this file asking whether to study it, I answered with a different question: do you want to save memorisation time, or do you want to win?

Those goals can coexist. They do not coexist in every game.

Four columns I use to test an opening product

My trade, put briefly, is finding the numbers nobody bothers to read. Over the years I built a four-column table. It is not elegant, but it has never led me badly astray.

Column one is author identity. Not reputation. Identity is measurable — titled or not, with what record in the very line being taught. If someone teaches 9...a5, I want to know how many games they played with it, against whom, and with what result. This document answers none of that.

Column two is novelty. 9...a5 is not a new move. It is a named branch that White has studied. The value of a product on an old branch lies in systematisation, not discovery. The document does not claim novelty — a point of fairness worth recording.

Column three is empirical evidence. How many elite games entered the branch, and how did Black survive. Not one game is cited.

Column four, the most important and most often skipped, is durability after the counter. An easy-to-learn move is only worth something if it still stands after the opponent prepares against it. For 9...a5, that question has no answer in this document.

Four columns, none filled. That is not a verdict. It is a blank table waiting for the seller to write in it.

The real twist is elsewhere

surprise

The document's real twist is not on the board. It is in the file I opened at 02:14.

Fourteen information points about a DVD were attached to a headline about Round 2 of a team event in Samarkand. No bridge joins the two halves. For someone who reads copy for a living, this is no small error. It is a signal about the entire pipeline behind it: an editorial process that stitched two sources without checking, or an aggregation system that crossed its data fields.

That confusion has concrete consequences. Anyone who reads this and cites it carries two distortions forward: one about a Samarkand event that does not exist in the body, and one about an opening product filed under an unrelated label. Wrong data does less harm than wrong data that is believed.

If the headline is right, there really was a team event in Samarkand — Uzbekistan has a strong chess culture — and it deserves its own primary source rather than a DVD tail grafted onto it like a foreign object.

If the body is right, there is an opening product looking for buyers, and it deserves assessment on its own numbers, not on a headline borrowed from somewhere else.

Neither half is proven. And that is precisely what I want to tell the reader.

When the theory market forgets itself

For years I have tracked the football transfer market alongside the chess theory market, and both share a disease. Both sell the story before the product.

The transfer market does not buy the past. It buys what the data has forgiven.

That line of mine has a translation into opening theory. A DVD does not sell moves. It sells the belief that those moves have already been forgiven by analysis. Has 9...a5 been forgiven? No one has supplied the evidence.

There is a lovely paradox here. The more closed an opening branch, the fewer forced moves it has, the harder it is to market with numbers, because the data for a closed position is poorer and spread across fewer games. That is why products of this kind shift into emotional language: easy to learn, light on the mind, free to think. Those words cannot be measured, so they cannot be refuted.

I do not blame the seller. I record a pattern.

Revival

revival

One thing must be said, or this piece will be unfair to the King's Indian.

The Mar del Plata is not dead. It has merely left the main stage of elite chess and moved into the role of a surprise weapon. In the engine era, its share of classical games has fallen sharply, because White prepares better and Black must accept high theoretical risk. But a system does not vanish when it stops being popular. It waits.

I have watched this many times across thirty-two years in the booth. A variation is abandoned for a decade, then a young grandmaster brings it to an open event, wins two games, and suddenly it is in every preparation file. The death of a chess variation is almost always a sleep, not a funeral.

If 9...a5 has a future, it will not come from a DVD. It will come from a specific game, at a specific event, between two named players. A move revives only when someone dares to play it and owns the result.

The contrarian angle

Most pieces about opening products end by asking whether to buy. I want the reverse question: for whom is it useful?

For a 2400-plus player, almost certainly insufficient. At that level, opponents prepare with databases and engines, and a known secondary branch will be neutralised within one theory cycle — a few months. A buyer at this level pays for something they could have found free.

For a club player rated 1600 to 2026, the answer changes. At that level the issue is not the strongest theory but a position you can play without self-destructing in the first thirty moves. If 9...a5 gives a Black player a structure they understand, a plan they remember, and a psychology they can hold, it has real value regardless of what an engine says about move nine.

For a junior coach, the value lies elsewhere again. This product, despite its missing data, is a lesson in how opening ideas are sold. It teaches a student something variation tables cannot: telling a promise apart from evidence.

The product's greatest risk is not that it is weak. The greatest risk is that a buyer enters a game with the wrong expectation, gets hit by a prepared opponent, and concludes the King's Indian is a bad system. That mistake does not belong to the system. It belongs to the seller who promised too much.

Endgame

I light candles for data. But I always let the flame of feeling illuminate the question.

That night, after closing the file, I opened an old game I had once commentated more than a decade ago, a Mar del Plata from a domestic league. Black lost in forty-one moves, and I remember saying in the booth that he chose the wrong branch. Now I am not so sure. Perhaps he simply chose a branch with no data.

For a chess game, that is always a gamble. For a writer, it is a lesson in checking the source before checking the move.

From today I will track a few more signals. Whether 9...a5 appears in a classical game at 2700 level — if it does, the DVD's premise is confirmed; if not, it is only an invitation to study. Whether the Bayonet Attack remains fashionable. And whether anyone can locate the original Samarkand report, to restore to the headline the truth it deserves.

Three signals. No more needed.

At the deepest level, all of us play the same game against time. Some moves are forgotten, but data never forgets them. The only thing I keep reminding myself is not to confuse a move that has not been recorded with a move that never had value.

The empty square on the board is often the most beautiful place to put a piece.