FRITZ 20: When the Chess Machine Switches Roles from Opponent to Teacher

Bonn, November 2026. Vladimir Kramnik lifted his queen to h7. A legal move, perfectly within the rules, and seconds later the most famous move in modern chess history. The machine answered with mate in one. The world champion fell to exactly the position a ten-year-old at any local club would have spotted. I sat in front of a screen that night, nearly ten thousand kilometres from Bonn, writing the position into a notebook with a worn spine. One line went into that notebook: this is the last time a machine will make the entire chess world hold its breath. Nineteen years later, that machine no longer walks onto a stage with arbiters, soundproof glass and sponsorship contracts. It has gone back to the training room. The announcement introducing FRITZ 20 from ChessBase opens with three short lines: your personal chess trainer, your toughest opponent, your strongest ally. Three roles, one machine. That order deserves a pause. The word 'trainer' comes before the word 'opponent'. Where I live and write, this story carries a different layer of meaning. A sixteen-year-old female player in Can Tho, in the eleventh grade, with an old laptop and an internet connection that stutters in the evening, owns something her older counterparts twenty years ago never had: a training partner who never tires, never snaps, never asks to go home early. The problem is whether she knows what to make it say. ChessBase, based in Hamburg, brought Fritz to market in 2026, with the software developed by Frans Morsch and Mathias Feist. Four years later Garry Kasparov beat Fritz 4-1 in a match held in Cologne, and the chess world concluded that humans still held a safe distance. That distance narrowed season by season. In 2026, in Bahrain, Vladimir Kramnik drew Deep Fritz 4-4. In 2026, in New York, X3D Fritz drew Kasparov 2-2. Three years later, in Bonn, Kramnik lost 2-4, and that match closed the era in which humans could still treat computers as peers. After that, the machines stopped being news. Rybka, then Stockfish, then DeepMind's AlphaZero in 2026, then Leela Chess Zero. Computer chess turned into infrastructure, like electricity and running water. Nobody writes tributes to the birth of an engine any more, because engines are everywhere, free or nearly free. Vietnam travelled that road to its own rhythm. Le Quang Liem won the World Blitz Championship in 2026 in Khanty-Mansiysk. Nguyen Ngoc Truong Son became an international grandmaster at fourteen. In the women's game, Pham Le Thao Nguyen and Nguyen Thi Mai Hung held familiar places in the national team for years, while Hoang Thanh Trang, born in Hanoi and playing for Hungary, earned the international grandmaster title. What matters is how they studied. Most Vietnamese players of that generation grew up with copied, unlicensed software, patchy opening databases, and a coach who held the decisive role. They improved through translated books, through hand-copied games, through hours at the board with older clubmates. In other words, they improved through manual labour. A tool that calls itself 'a training revolution' has to be read through that history, not through its own advertising copy. The months of cancelled tournaments in 2026 taught me that a chessboard never dies, it only waits. In that silence, most young Vietnamese players moved online, and for the first time the serious playing hours of a child in a provincial town approached those of a child in Hanoi. The geographical gap in Vietnamese chess eroded from precisely that moment. When a machine introduces itself as a trainer, the question worth asking is not how strong it is, but what it teaches and whom it teaches. The first role, the trainer, runs on a feedback loop. The software remembers each learner's personal opening repertoire, detects recurring patterns of error, and turns them into exercises. The mechanism is not new in principle. What is new is the degree of automation, and the fact that it no longer requires a coach sitting beside you to operate. The second role, the opponent, runs on adjustable strength and assignable style. A young player can spar against a machine that defends solidly, then switch to one that attacks recklessly, then switch to one that grinds out endgames. Previously, to find three such opponents in a single week, a child in the provinces had to ask three different people with three different schedules. The third role, the ally, runs on data. A finished game of yours is placed beside millions of others and returns an answer with evidence attached. Those three roles belong to three different stages of a career, yet the software sells them as a single package. That is a commercial strength, and also a pedagogical blind spot. The marginal cost of an hour of training falls to almost nothing. A session with a highly qualified coach in Hanoi or Ho Chi Minh City costs money that many provincial families cannot sustain regularly. Software is bought once and used for years. For a middle-income family in the Mekong Delta, that difference decides whether their daughter trains seriously at all. A second barrier falls more quietly, and is discussed far less: the scarcity of sparring partners at your own level. In women's events, the pool of female players of the same age, the same strength, the same time zone and the same living conditions is thin. A fifteen-year-old girl who wants three serious games a day usually has to rely on the boys at her club, who keep different schedules, carry different competitive psychology, and are treated differently. A machine does not distinguish the gender of the person sitting in front of the screen. It reads moves. For women's chess, this change carries more weight than any promise about artificial intelligence. The third change is psychological. Children learning chess at provincial clubs often grow up with the sense that correct knowledge resides in someone above them, and that they must wait to be told. Software inverts that relationship. It does not issue a verdict backed by authority; it offers a sequence of moves with a score attached, and the learner has the right to object. A child who dares to say 'the machine is wrong' and then checks it is a child learning to think independently. In training sessions I have observed with the national women's team and at several clubs in the south, a fairly effective routine has taken shape, and it runs almost opposite to a beginner's instinct. After finishing a game, the learner switches the machine off, writes down the points of doubt, proposes alternatives, and only then turns on the analysis. When your prediction is set against the machine's output, the discrepancy becomes the lesson. Switch the machine on first and the learner receives only an answer, and the brain processes an answer very differently from a question. Based on my experience following matches and analysis sessions over many years, most of the software's value lies after the game has ended, not in the speed of its calculation. A machine ten times faster helps a player understand nothing extra if that player does not know what they want to ask. At the organisational level there is a problem the software never touches. The big chess centres in Hanoi and Ho Chi Minh City recruit hundreds of children each year, place them in classes, enter them in junior events, and keep those with the best results. The number who genuinely reach international competition regularly is a small fraction of the original intake. That model resembles how large football academies operate: hoard the talent first, filter afterwards. A machine in the backpack of a child in a rural district cannot fix that structure, but it gives her a second path, one that does not depend on whether she survives the shortlist. In chess I rarely watch the screen; I watch how people stay calm while they are losing. Software cannot teach that skill, but it creates many opportunities to fail in silence, and that is a form of training earlier generations never had. Here is what few people want to hear: better tools usually widen the gap rather than close it. Software operates as a multiplier, not an addend. A multiplier applied to a number close to zero still yields close to zero. Two fifteen-year-olds buy FRITZ 20 in the same week, own similar laptops, and have ten free hours a week. Two years later, one may have advanced three times as far as the other. The difference is not the software. It is whether the first player has someone who knows how to teach her to ask better questions — a coach, an older teammate, or simply a demanding friend willing to sit and analyse with her. The second risk is underestimated: individualisation can slide into isolation. Chess is a game in which most training happens inside one head. A machine turns that time entirely private, and that sounds efficient until you notice that children learn fastest when they are beaten by a peer, and then have to sit down together to understand why they lost. Software does not create that gentle sting of embarrassment. Nor does it create shared joy. The third risk concerns the very idea of efficiency. A tournament game lasts five to six hours. At move forty, what decides matters is often no longer opening knowledge but the capacity to endure ambiguity: to sit with a position of unclear value, with no answer available, and still make a decision. Software is designed to eliminate ambiguity. That is its benefit in the training room and its limitation away from the board. One further consequence is rarely discussed: engines flatten style. When every young player grows up on the same database, the same evaluation scale, the same standard of the 'correct' move, games become more alike in the first twenty moves. The difference is pushed backwards, into stamina, into psychology, into the ability to handle a position that has never appeared in any database. Strategy resembles people: you must know who you are before you try to become someone else. A player who uses FRITZ 20 to become a copy of the machine will lose to a player who uses it to understand herself more clearly. When the cheering stops, I finally hear the true breathing of the game. And in the training room, when the ceiling fan drowns out the sound of pieces on the board, I hear something more important than any software update: the patience of a person willing to sit with her own mistakes. What I await next season is not another version with a different name. I await the image of a fifteen-year-old girl in a provincial town, switching off the machine, closing her notebook, and walking out of the room with a new question in her head — a question she invented herself, not one the machine handed to her. The tool is already on the table. People who know how to teach its use are still what is missing, and that is where the next decade of investment belongs.