Fritz 20 and the Training Room Revolution: How Vietnamese Chess Is Learning to Learn Again

The fourth-floor training room of a small chess club hidden in an alley off Nguyen Thi Minh Khai Street, Saigon. Eleven at night in September, rainy season, water drumming on a corrugated roof like applause inside an empty stadium. Table three was still lit. A fourteen-year-old boy sat in front of the screen, left hand on the mouse, right hand resting on the desk as if pieces were still there. His coach, a former national player now past sixty, stood behind him, arms folded, silent for twenty minutes straight.

The position was on move twenty-eight. The boy pushed a Knight to an empty square, a move that looked like nothing at all. Three seconds later, a line appeared in the lower-left corner of the screen, suggesting a different plan that the man had never considered in forty years of playing chess. He leaned down, read the line, read it again, then a third time. Said nothing. The boy said nothing either. Only the ceiling fan and the rain remained.

I sat at the last table with my recorder running, and what I captured was that silence. Not the move. The silence. The silence of a man realising that what he had taught for four decades was being taught back to him by a program — faster, cheaper, colder. And inside that same silence something else was happening: the child, who had never known a world without machines, was learning chess in a way my generation cannot even picture.

Two hours later I asked the boy why he chose that Knight move. He answered instantly: I thought of it myself. The machine just confirmed it. Then he turned back to the keyboard for the next game, as if the sentence meant nothing. To me it meant more than every analysis table I have read in three decades of writing.

I began covering chess in 2026, spending more than twenty years commentating on classic finals for Vietnamese radio and television during the era of Davis and Hendley before moving into sports journalism more broadly. Chess remained the root. And that root is being dug up from underneath.

Three eras of the Vietnamese training room

Vietnamese chess training has passed through three distinct eras. The paper era ran from the 1980s to the late 1990s. Learning chess meant copying: copying lines from books, from battered photocopies of old Informant volumes, from the notebooks of the player before you. Memory was capital, and the coach was the keeper of the vault.

The floppy-disk era began when the first computers reached the clubs. A single disk holding a few hundred games was treasure, passed hand to hand. That was when the name Fritz first appeared in training rooms, at first as a toy, then as a sparring partner that never tired.

The online era arrived later in Vietnam but arrived fast. Once connections stabilised and online playing sites spread, a training room stopped being a room. It became an IP address. A child in Can Tho, Nghe An or Dak Lak could play an International Master in Hanoi at midnight, and more importantly could train against a machine hundreds of times stronger than anything the local club could arrange.

Those three eras add up to roughly thirty-five years — a pace no other sport in Vietnam has gone through.

In 2026, when Deep Blue beat Garry Kasparov 3.5-2.5 in the New York rematch, I was in a small chess room. After the final game, the room went silent. Nobody argued about openings, nobody analysed the endgame. People only asked each other one question: so what exactly are we playing? That question has never been answered cleanly, and probably never will be.

In 2026, in Bonn, Deep Fritz beat Vladimir Kramnik 4-2 in an official match. The second game entered history for a reason the chess world rarely likes to mention: the world champion missed a mate in one and was mated immediately afterwards. I wrote a very long piece about that game, and I still believe what collapsed in it was not chess technique but a belief — the belief that a human can stay calm against an opponent that has never known fear.

Eleven years later, in December 2026, DeepMind published AlphaZero's results against Stockfish: 28 wins, 72 draws, no losses, in a format where the machine had been given almost no human chess knowledge. This time nobody asked what game we were playing. They asked something more practical: so what should we study?

In 2026, when Stockfish moved to a neural-network architecture, the game shifted again. Instead of being strong through deep calculation, engines became strong through evaluation. That sounds technical, but the consequence for a training room is concrete: if strength lies in evaluation, then a machine can teach humans far more than before, because evaluation is learnable, while calculating thirty moves ahead is not.

That is the ground on which a product like Fritz 20 appears, marketed not as an opponent but as a coach. The material describes it plainly: a toughest opponent, a strongest ally, a personal trainer, built to help players train more efficiently, more intelligently, more individually — for beginners taking their first serious steps and for professionals already competing at tournament level.

A name that has lived thirty-four years

Fritz was born in 2026, developed by Frans Morsch and Mathias Feist and published by the German chess publisher ChessBase. A year later there were reports that a version running on a 486 beat Garry Kasparov in a blitz game in Cologne — the first time a commercial chess machine had beaten the world champion at a fast time control, even in an unofficial single game.

In 2026, Fritz won the World Computer Chess Championship in Hong Kong, becoming the first microcomputer program to take a title that had belonged to expensive mainframes. That is the most important milestone in the brand's history: proof that serious chess strength could fit in a machine an ordinary family could buy.

Thirty-four years later, many stronger engines have come and gone. Stockfish survives, but Stockfish is not a product for learners. The strongest engines in the world are not built for ordinary players, because they answer only one question: which move is strongest. Humans need more than that.

Here is the most beautiful paradox in chess: the strongest machines are not the most loved ones. A program rated 3,500 can still fail to teach a ten-year-old how to think about a good pawn structure. Fritz survived not through strength but by accepting the role of teacher while every other engine wanted to be champion.

With Fritz 20, the emphasis is on adaptive training — a sparring partner that adjusts to the learner's level and style instead of crushing them with perfect moves — alongside the familiar toolkit of a modern chess program: a trainable opening repertoire, blunder detection, game analysis, and the ability to turn a played game into a structured lesson.

On paper these are standard features. The difference is philosophical, and it sits in the word individual.

What actually changed in the Vietnamese training room

Based on my experience watching chess matches and training sessions in Vietnam over three decades, the biggest change is not engine strength but the economics of learning chess. Fifteen years ago, a student needed three things: a good coach, a group of peers at the same level, and enough books. All three were distributed brutally unequally. Hanoi and Saigon held nearly all the strong coaches. A provincial player who wanted a decent teacher had to wait, move, or accept patchy instruction by phone.

Today the hardest of those three has become the easiest. A machine at International Master level, adjustable at will, runs on a mid-range laptop and costs about as much as a few coaching sessions. It is a flattening of geography that nobody announced and nobody inaugurated, happening quietly in thousands of living rooms.

But there is a paradox: the easiest opponent to find is also the easiest to abuse. I have watched sessions where a student played the machine for three hours and could not recall a single move afterwards. I have also watched sessions where a student played for forty minutes and then spent an hour copying by hand what had just been learned. The second student improved faster despite training less. The difference was never the machine. It was the user.

This is where training software genuinely matters. An engine that only outputs the strongest move is useless to a learner, because it turns study into lookup. An engine that sets positions, suggests, asks questions, detects recurring weaknesses, and refuses to hand over the answer until the student has thought long enough — that is a useful engine.

The modern training loop I observe in the fastest-improving students has four clear steps. First, play a game at a controlled time limit, ideally against a human. Second, analyse it with your own eyes and note every moment you hesitated, correct or not. Third, compare with the engine and find the three biggest discrepancies, not ten. Fourth, turn those three moments into three small exercises for the following week.

Step two is the most important, and it is precisely the step the machine cannot do for you. That is where chess remains a human sport.

An anchor fact

In 2026, in Khanty-Mansiysk, Le Quang Liem won the World Blitz Championship — one of the greatest titles a Vietnamese player has ever taken. By then he was already one of Asia's leading players, a multiple Aeroflot Open winner in Moscow who had broken into the world's elite.

The memorable part of Le Quang Liem's story is not the title but how he trained. He belongs to the first Vietnamese generation raised alongside analysis software, yet he kept a distinctly personal style — famous for blitz accuracy in complex positions and rare composure in tense ones. If machines could manufacture style, a whole generation would look identical. They do not.

This is what I want to tell parents buying software for their children: the machine does not create style. It creates capability. Style still has to come from the person, and it usually comes from deeply untechnical places — a defeat that stings for years, a demanding coach, or a mistake a child decides not to correct.

Vietnam's next generation — Nguyen Ngoc Truong Son, Le Tuan Minh and Nguyen Anh Khoi among the men, Pham Le Thao Nguyen among the women — all grew up in an environment where the computer is inseparable from study. None of them can imagine preparing for an international event without a database and an engine. The question is no longer whether to use machines. The question is how to use them without losing yourself.

Blind spot one: engines cannot teach nerve

An engine can point out the correct move in a hard position. It cannot teach a player how to sit at the board in the ninth round of a ten-day tournament, legs aching, head heavy, when the only correct move demands a calm the player is not sure they still have.

This sounds obvious, but the consequences are large. If what decides results at the top is decision-making under pressure, then most engine training hours are spent rehearsing a different skill: finding the best move when nothing threatens, the clock is not running, there is no audience, and there is nothing to lose.

Based on my experience watching matches, games decided by pure technical difference are a minority. Most are decided by who stays cold longer. At that, the machine is a poor teacher, because it has never been afraid.

This is also why I believe Vietnamese youth tournaments — however short of money, arbiters or venues — remain irreplaceable. A child who plays three hundred online games a year is still not equal to a child who plays thirty over the board, because those thirty teach something no software has: the feeling of being accountable in front of another person.

Blind spot two: machine eyes and human eyes

When an engine analyses, it sees the position in percentages. Plus three, minus six, dead equal. Human players, however hard they try, cannot see positions that way. Humans see them through experience, fear, and the memory of a similar loss ten years ago.

I have watched a generation of students read evaluation numbers very quickly but fail to explain why a position favours one side. They know the answer without knowing the reason. That is the most dangerous kind of knowledge, because it looks like knowledge but is only a lookup habit.

The strongest students I have met in the past decade share one habit: they switch the machine off before analysing. They play the game, close the screen, walk, drink water, and only then look at what the engine says. That delay looks pointless in an age of instant everything. It is the exact distance between a learner and a searcher.

A good engine is one that knows when to stay silent, when to wait, when to pose a question instead of delivering an answer — a design instinct opposite to that of most software developers, who are trained to answer as fast as possible.

Blind spot three: opening homogeneity

A classic fear in chess is that machines will flatten every style into copies of one opening book. Looking at elite play over the past decade, that fear is partly right. Optimal plans became common property, and optimality was distributed to everyone at once.

Look closer, though, and something more interesting happens. Once opening knowledge becomes common property, the advantage no longer lies in knowledge. It moves to choosing which opening to sidestep, when to leave the book, which opponent to steer into it. Creativity does not vanish; it migrates from inventing moves to inventing the strategy of choosing moves.

In Vietnam this matters especially, because Vietnamese players have never been strongest in heavy opening theory. The traditional strength lies in calculation, in nerve in messy positions, in playing well once the book ends. If engines flatten the early game, the middle game is where a Vietnamese player can still reclaim the edge. Data patterns show this clearly: in events where Vietnamese players outperform expectations, most wins come after move twenty.

For young coaches that is a useful instruction: do not spend seventy percent of your time on openings simply because engines analyse openings well. Spend it on the positions machines cannot teach.

Blind spot four: the measurement trap

A trap I see often in Vietnamese families who begin serious chess study: once the computer is present, everything becomes measurable. Games, minutes, rating, errors, weekly progress charts. And because it is measurable, it is judged constantly.

The result is a new kind of pressure. A child trains three hours a day but dares not admit they love an outdated opening, because the chart says it is five percent worse. A child studies to optimise a metric instead of to understand a game. I have seen twelve-year-olds play brilliantly while no longer remembering why they started.

The most romantic mistake of mine was believing the ball could write poetry. With chess I nearly believed the same thing: that the pieces could write poetry, and that a machine could help us write the correct poem. Both beliefs fail in the same way. The machine does not write poetry. It only tells us where we wrote badly. The rewriting is still ours.

Style does not disappear. It relocates

It would be too simple to conclude that machines killed style. Style only moves. Thirty years ago a player's style lived on move fifteen, in a strange opening, in a plan nobody taught. Today move fifteen has been analysed to exhaustion, so style must find another home. It lives in choosing which tournaments to enter, in managing ninety minutes, in declining a draw offer, in picking opponents to beat and opponents to avoid.

In other words, as machines make the move less personal, the human part of chess is pushed to a higher floor: strategy, psychology, tournament planning. Players who realise this early will hold a large advantage over the next decade. Players who still believe winning means memorising more lines will fall behind fast, because memorised lines have become a commodity anyone can buy for pocket change.

There is also a reverse trend worth noting: when everyone has the same analysis power, endurance becomes a competitive advantage. In a world where anyone can find the best move on a machine, the winner is the one who finds the best move in their own head, in twenty seconds, opposite someone trying to beat them.

Collective memory and the opening book that belongs to everyone

Vietnamese chess's collective memory has changed completely. Thirty years ago, when I asked players which games they remembered best, the answers were games they had played or watched in person. Today that memory is formed from identical databases. A whole generation remembers the same game, the same line, the same blunder.

There is a beautiful side: a generation sharing a knowledge base means far better chess arguments. A fourteen-year-old in Nghe An can debate an opening with a coach in Saigon using the same data vocabulary — impossible thirty years ago.

There is also a worrying side. Shared memory easily becomes blurred memory. When everyone recalls the same game through a chart, nobody remembers it as an event: who sat at which board, what the arbiter said, the sound of a chair dragged across the floor, a hand shaking as a piece was placed. Data remembers everything; people remember less.

An empty stadium still keeps the applause of the dead. I wrote that line in a 2026 feature about football without crowds, and I have found it true of chess in a different way. Tournament halls during the pandemic were so quiet you could hear a player breathing at the next board. Those halls had no spectators, yet still held the applause of a generation that had passed. Machines cannot record those sounds. Someone has to sit there and write them down.

Fritz and the lesson of endurance

Back to Fritz 20. What interests me is not the twentieth edition of a program but the fact that it still exists to have a twentieth edition. In the chess software industry, thirty-four years is enormous. Most programs that dominated the 1990s have vanished or shrunk into personal projects. Fritz survived by accepting a lower technical position and a higher practical value. It is no longer the strongest engine. It is the most usable one for someone trying to improve.

There is a lesson here for Vietnamese sports products, whether software, platforms or tournaments. The race to be strongest is short. The race to be most useful is long. And in chess, as in every sport, the end user is not the best player. The end user is the child still trying.

For a child in Vietnam, a personal coach inside a laptop is both a miracle and a hazard. A miracle, because it teaches things that once required money and geography to buy. A hazard, because it can turn chess study into serving a machine instead of making the machine serve you. The difference does not live in the software. It lives in whether adults teach children how to ask questions.

What I fear most

What I fear most is not machines becoming stronger than humans. That happened long ago and humanity has coped reasonably well. What I fear is a generation of young players who learn how to play chess without ever learning how to love it. Children who play very well, hit norms very fast, but have no single game they remember forever because it was beautiful. Children who can point out twenty mistakes in their own game but cannot name a game that made them happy. Children taught everything by a machine except why they should play at all.

When a seventeen-year-old prodigy still does not know they are about to become a legend, they play for a reason that is private, naive and usually quite silly. That silly reason is something no software can manufacture. If we use machines to erase it, we will produce a generation of accurate and empty players.

I believe this is the line the Vietnamese chess training room now stands on. On one side, genuine progress: children in provincial towns gaining access to tools that twenty years ago only national team members had. On the other, soulless optimisation, where studying chess becomes a weekly scoreboard update. There is no technical answer to that line. It is a question of education, not technology.

Back in the training room

The fourteen-year-old was still there, playing his fifth game of the night. His coach had finally sat down, no longer standing with folded arms. He was taking notes in a paper notebook, in ballpoint pen, the way he had for thirty years.

I asked what he thought of the machine running on the screen. He answered briefly: It teaches better than I do. Then, after a pause: But it does not know who this child is.

I think that is the best summary of the whole story of the twentieth engine. Everything machines can do for Vietnamese chess, they are already doing well: flattening geography, cutting costs, accelerating learning, providing an opponent that never tires. Everything machines cannot do remains on the human side: knowing who a student is, knowing when to stay silent, knowing why a game is worth remembering.

The twentieth engine will not produce a world champion for Vietnam. It can only help some child in a provincial town prepare better, and get closer to a threshold that was out of reach twenty years ago. The rest is the old story: one person at a board, one person behind them, and a question that exists in no database.

After three decades at the keyboard, I have realised I never wrote about sport. I wrote about people trying to become themselves inside a game whose rules are too clear to lie about. Chess gives them sixty-four squares to do it. Fritz 20, or whatever engine comes next, is only another pen in the same room. The question remains who holds it, and who they are writing for.

That night, as I left the alley off Nguyen Thi Minh Khai, the rain had stopped. The fourth-floor training room was still lit. And I heard, very clearly, the sound of a wooden piece being set down on a board — a sound no machine can make, and no machine can record on my behalf.