A Samarkand Headline, a Mar del Plata Body: The 9...a5 DVD and a Forensic Audit of a Feed Error

Last week I turned a page in the ledger and stopped. The headline said Round 10 in Samarkand, women's section, Kazakhstan closing the gap. The body said something else entirely: a video course on the King's Indian Defence, an answer to the Bayonet Attack with 9...a5, and underneath it, in small type, “Product Introduction.” I read all eleven information points from first to last. Not one female player is named. No Samarkand, no Kazakhstan, no round-10 result, no standings, no ratings. Every one of the eleven points is a DVD advertising itself in its own words. The gap between the headline and the body is the single most important fact here; the question is not about a product's quality but about a document's integrity. In the 1990s, building xG and PPDA columns on a Mumbai football desk, I learned that a document careless about itself is careless about everything else. Digitizing 1,240 games from 41 national championships between 2026 and 2026 hardened that lesson.

You have to put the board in front of you first. 1.d4 Nf6 2.c4 g6, then ...Bg7, ...d6, ...O-O—in the King's Indian, Black deliberately hands White a big centre and plans to break it later. In the Classical branch Black plays ...e5 and shuts the centre; in the Mar del Plata (7...Nc6) the knight goes to e7 and ...f5 launches a kingside storm. White then breaks through on the queenside: 9.b4, the Bayonet Attack, followed by c5 and b5. At that point Black has two roads. The familiar main road is 9...Nh5, where the lines are forcing, memory-heavy and engine-saturated. The second road is 9...a5—old, known, and neglected by the theory market, where Black stops White's b5 expansion before it starts and keeps the structure intact.

The video-course market has to be read against that background. Club players and sub-2026 juniors are suffocating under the engine-era theory explosion: a new line every month, a new “forced” branch every week, and finite memorization capacity. That pressure created demand for “less memory, more understanding” courses. The delivery channel has shifted too—physical DVDs are effectively finished, with theory now carried by interactive courses, streamed lectures and streamers' live explanations. In my own notebook there is a column I call “platform-adjusted,” where I record how the medium changes the meaning of the data. It became mandatory after 2026, when I spent six weeks refusing to count online games at all.

Now to the document itself. The DVD claims that 9...a5 is a workable antidote to the Bayonet Attack; that the lines contain few concrete, forcing sequences, so the memory load is lighter; and that the emphasis falls on strategy rather than tactics. It adds that computers do not handle these positions well, so an unprepared White can quickly get into trouble.

Beside those claims I set out my ledger columns. Column one: engine match rate? Absent. Column two: average centipawn loss? Absent. Column three: database win-draw-loss rate for the line? Absent. Column four: how many reference games, and by whom? Not a single name. Four empty columns, and yet a verdict is already written—“objectively, the chances are equal.” That is the difference between a valuation built from sentences and one built from columns; the spreadsheet does not argue, it waits for the next correction.

“Computers cannot handle these positions” carries the imprint of a specific period in chess technology. Before the neural-network engine era of roughly 2026 to 2026, engines were comparatively weak in closed, slow-burning structures like the Mar del Plata; in course marketing of that time the sentence was legitimate. Today's engines evaluate the same structures with far greater reliability. The claim is not false so much as dated—its parentage suggests this course is a re-release pulled from a catalogue's long tail, not fresh research.

The second fracture runs deeper. Every claim that establishes the course's value—antidote, danger for White, equal chances—comes from the course's own mouth. A product that is its own witness is not an independent witness. Covering the 2026 Batumi Olympiad game by game, logging all 44 of Bangladesh's men's games, I learned to place a witness beside every claim: cross-reference the tape, the box score, and the silence between them. Here there is no tape, no box score, only a sales brochure.

Where the headline fracture came from can be inferred. Catalogue pages and feed templates bleed into each other; it is an old failure mode. A live tournament report's headline stays in place, the body is replaced by a product listing, and nobody updates the headline. If the headline is genuine, then a separate report on Round 10 of a women's event in Samarkand exists somewhere and has been lost from this record. If the headline is spurious, this document is pure advertising with zero news value. Either way, the action is the same: verify before publishing.

I think back to the 2026 audit. After digitizing 1,240 games from 41 national championships, one column stood up on its own: average centipawn loss. In 2026 Niaz Murshed's games averaged 28 ACPL; the 2026 title field averaged 47. I sent that nineteen-point gap to three Dhaka sports desks; all three passed. I still keep the rejected pitch on the first page of the ledger, because an unpublished finding still has to be correct. Had this DVD document landed on my desk, the first thing I would have done is replace that nineteen-point gap with four zeros.

At Batumi in 2026, filtering 44 games for positions the engine rated +1.00 or better, I counted 13. Bangladesh converted 3—a conversion rate of 23 percent. Something else matched: nine of the team's eleven losses arrived after move 35, in time trouble. Since then I have refused to file an Olympiad piece without a move-number distribution behind it. Today a claim sits in front of me with no move numbers, no engine match, no reference games—all three missing—while the product insists White will quickly be in danger.

In 2026 the boards were packed away and the arenas empty; Leonine Chess Club moved online. For six weeks I refused to treat online games as data—mouse slips, no touch-move rule, unverifiable settings. Then I built a time-per-move model on 412 online games and found the error rate roughly 2.3 times that of over-the-board rapid play. That is why every dataset I keep now carries a platform-adjusted column. In this DVD's case the platform itself is the old one: the product is not standing at the market's current address, it is standing at a former one.

A real signal remains, and it cannot be waved away. At club level, demand for “less memory, more understanding” opening solutions is durable. The reason is simple: an amateur cannot memorize a fresh theory branch every week, but a strategy, once understood, pays for years. 9...a5 is exactly that kind of solution—legitimate, playable, and unexpected at club level. In my reading, that demand is a rational response to the engine-era memory race.

This is where the counter-turn arrives, and it goes against my own first reading. My instinct was to blame the product for the headline-body fracture. The arithmetic says the fault is the feed's. A product will advertise; labelling advertising as advertising is the feed's job. When the border between news and advertising dissolves, the loss is not the product's—it is the reader's trust. The second counter-turn: 9...a5 is not a bad move, and the fact that someone chooses it does not prove that a country or a generation is falling behind. The nineteen-point gap across 41 championships was an institutional ledger, not a moral failure of players. By the same logic, a DVD written in older engine language does not establish that chess education has declined. Confusing correlation with causation is the most common error in my profession, and I am not willing to commit it.

The third counter-turn is the most uncomfortable. These claims are cheap to verify: one engine, one game database, two hours of work. Almost nobody in the theory market does those two hours, because the language of advertising and the language of analysis have merged. Verifying a number costs me an hour; verifying a sentence can take ten years, because a sentence has no columns.

The medium is part of the arithmetic too. The physical DVD is an exhausted market. If this document is a re-release pulled from a catalogue's long tail, it signals monetization of old stock rather than fresh investment—commercially reasonable, analytically equivalent to trusting an old-dated witness.

Who benefits can also be traced inside the line itself. 9...a5 makes White's counter-preparation easier: if Black always plays ...a5, White's plan becomes clearer. The more popular the line becomes, the faster it will be defused, at least against a prepared White. That constant has held for three decades: a “secret weapon” that gets published has a shelf life equal to one news cycle.

So I will watch four things. First, where current engines place 9...a5, and whether the course's “equal chances” claim survives. Second, feed integrity—how often headline and body walk separate paths; if it recurs, confidence in the whole corpus drops. Third, whether the King's Indian, and specifically the Mar del Plata, is rising or falling at elite level. Fourth, the pace of migration from physical DVD to platform courses, and whether club players actually encounter the Bayonet Attack at all.

By two decades of habit I set the tape, the box score and the silence side by side. This document contains one of the three—silence. But this silence is not testimony; it is a cover placed over the claims.

One question still hangs, and it will not sit in any column. If that Round-10 report from the women's section in Samarkand genuinely exists somewhere—a player, a position, a result—today's discussion will not fill its place. I have left the ledger page open. I digitized the past so the present could not lie to me. The feed is now waiting for its next correction, and somewhere at a club board, White is still preparing to push b5.