FRITZ 20 and the Fight to Reclaim the Chess Engine's Role as Teacher

The desktop computer sits in the corner of the training room, its screen still covered in scratches left by decades of students' fingers. The copy of Fritz installed on it is fifteen years old, and it still runs, still calculates faster than anyone in the room. That afternoon, an eleven-year-old boy was struggling with a rook-versus-pawn endgame — the kind of position any serious tournament player must memorise like a multiplication table. The screen displayed the familiar evaluation: +2.34.

The boy turned to his coach and asked: "Teacher, it says I'm winning, but I don't see myself winning." The coach shrugged. He could not explain it either, because he had never managed to read a single line of that software's documentation — it was written in English, and it was written for an entirely different user: a player who already carries a system of symbols in his head, who has a concept of school and style, who can translate a +2.34 line into three concrete moves on the board.

That moment contains the entire problem of chess computing in a single sentence spoken by a child. The machine knows the answer. It simply does not know how to say it. And for thirty years, the whole chess software industry — a business worth hundreds of millions of dollars annually worldwide — has been circling exactly one problem: how to turn the machine into a teacher.

This week, ChessBase, Germany's long-established chess software company, offered a new answer. They introduced FRITZ 20, and in the product description they called it a "training revolution," a "personal chess trainer," something for "players taking their first steps as well as ambitious players and professionals."

I read that description four times. There is no release date. No technical specifications. Not a single comparison with Stockfish or Leela Chess Zero. No metric of training effectiveness. Only adjectives.

It reminded me of the boy with the rook endgame. He, too, once received a product with no instructions.

ChessBase was founded in 2026, in the Hamburg region, by Frederic Friedel and Matthias Wüllenweber. At the time, a digitised chess database was an almost insane idea: personal computers were weak, hard drives were measured in megabytes, and professional players still recorded games on carbon paper.

The Fritz engine appeared in the early 1990s, written by the Dutch programmer Frans Morsch. For nearly two decades, Fritz was the public face of chess computing — not because it was the strongest, but because it was the machine people put in front of human beings.

In 2026, a version of Deep Fritz drew 4-4 with Garry Kasparov in the "Man versus Machine" match in New York. Three years later, in November 2026, in Bonn, Deep Fritz defeated reigning world champion Vladimir Kramnik 4-2. It was the first time a world champion lost to a machine in a formally organised official match.

That memory is still warm. And I believe it is what ChessBase is actually selling, rather than a piece of software.

Because since 2026, the world has turned completely. Stockfish — open-source, free, maintained by a volunteer community scattered across the planet — overtook every commercial engine in raw strength roughly a decade ago. Leela Chess Zero, born in 2026, learned chess by playing itself through reinforcement learning and reached superhuman level within a few years, at zero cost to the end user.

On online platforms, Lichess offers free, unlimited engine analysis to anyone with an internet connection. Chess.com has hundreds of millions of registered accounts and, in December 2026, completed its acquisition of Play Magnus Group for around 80 million dollars, absorbing almost the entire ecosystem of elite chess tournaments, coaching and content.

In other words: what Fritz once monopolised — the ability to calculate the best move — has become tap water. Anyone can turn on the faucet.

So what does a software company sell in that world?

The answer lies in how ChessBase describes FRITZ 20. They do not talk about engine Elo. They do not talk about calculation speed. They talk about "training."

That is an understandable strategic move, and it is also a move that reveals position. When you can no longer win on the field of strength, you must move to another field — one the free competitors have not yet taken.

But this is where I want to pause, because it concerns more than one product.

In more than twenty years in the commentary booth at VTC, I covered almost every major final in Vietnamese and regional chess. My job was to observe what happened on the board and translate it into language. And I learned something no ranking list can measure: what determines a player's progress is not the quality of the move the machine points to, but the quality of the dialogue between the learner and that machine.

An engine tells you the move you just played was wrong. It says so quickly, accurately, and with total indifference. It does not say why you thought that move was right. It does not know what you saw three moves earlier. It does not know you have just come through four hours of play and are exhausted. It does not know what you fear.

Progress in chess comes from diagnosing where the error sits in the chain of reasoning — not from knowing which numbered move the error sat on. A seventh-grader in Hanoi plays a bad middlegame move not because he does not know the right move, but because he has not yet formed a concept of pawn structure. Stockfish will show him the right move. But for him to understand why that structure matters, he needs a person — or a piece of software that knows how to ask a question.

This is precisely the gap FRITZ 20 claims to fill. And it is precisely where every marketing claim must face the harshest scrutiny, because in the history of chess software, nobody has ever proven with data that their product makes people improve faster.

I have looked. I have looked for years. There is no independent study comparing the improvement rates of two groups of students using two different training programs over six months. No ACPL — average centipawn loss — metric is tracked as an output measure by any training product on the market. No engine match rate is used as an evaluation scale for coaching effectiveness.

This industry sells progress but has never measured it.

In the FRITZ 20 description there is one detail that deserves a slower read than the rest. The product is said to serve "players taking their first steps" and simultaneously "ambitious players and professionals."

These two user groups do not have the same needs. A beginner needs to be taught how to see the board, how to count, how to recognise threats, how to accept that he will lose many times. A professional needs deep opening databases running thirty moves deep, multi-variation analysis tools, a module managing an entire personal game archive, the ability to export reports for a coaching staff.

A product that serves both groups is usually a product that serves neither completely. And in today's chess market, the most successful names all pick a side: Stockfish picks pure strength; Chessable picks structured memorisation; Lichess picks free and open; Chess.com picks the integrated experience.

FRITZ 20 standing in the middle is not necessarily a mistake. It is the sign of a brand trying to hold on to a user base it once dominated — people who bought Fritz 12, Fritz 13, Fritz 14 over twenty years, and who now need a reason to buy again.

And here is what I take to be the core: most people who buy FRITZ 20 will not be buying it. They will be buying the memory that Fritz was once the machine that beat Kramnik.

I have spent many afternoons in chess clubs in Hanoi, Ho Chi Minh City, Da Nang, and in school chess classes in the provinces. I have a habit of visiting youth training sessions every Wednesday, noting gesture, gaze and breathing rather than fixating on results. That habit began on an afternoon in 2026, when I skipped a newsroom meeting to watch a young student practise free kicks. I kept the same method when I returned to the chessboard.

In those places, I have never seen a club pay licence fees for training software. I have seen computers running old programs. I have seen installations of uncertain origin. I have seen clubs with fifteen students and one projector. I have seen coaches using their own phones to look things up because the room's computer is offline.

In many places, the strongest training tool in the room is not the paid software. It is Lichess, open in an incognito tab on the coach's phone.

That is the market reality any ChessBase product strategy must confront, and the FRITZ 20 announcement does not mention it.

So what actually helps a young Vietnamese player improve?

From my observation, I rank it thus: first, hours of quality chess with someone checking your work; second, memorising basic tactical patterns to the point of reflex; third, playing long games against slightly stronger opponents; fourth, analysing your own games; and only last, the strength of the analysis tool.

Software sits at the bottom of that list, yet it absorbs almost the entire budget families are willing to spend.

That is the paradox I believe FRITZ 20 — and any future product — will have to solve. Not by being stronger than Stockfish. But by doing what Stockfish was never designed to do: making the learner come back to the board the next day.

If I had to bet on a single metric for the future of chess training software, I would not bet on engine strength. I would bet on the thirty-day return rate.

That is a metric nobody in the industry publishes. And that silence is itself a form of data.

At this point I want to talk about my own profession's mistake.

When chess software is released, the reflex of commentators is to compare Elo, search depth, nodes per second. We measure the machine by exactly the standards the machine understands. That makes us part of the problem.

Because those standards lost their value long ago. Once Stockfish surpassed every commercial engine, one engine being a few dozen Elo stronger than another means nothing to anyone outside the people writing engines. To an eleven-year-old in Hanoi, the gap between the world's strongest engine and the world's tenth strongest is zero. Both are a thousand Elo above him.

The blind spot of collective memory lies here: we still remember Fritz as a winning machine, and we still judge Fritz by the standard of winning. But in a world where no machine beats the free machine, that standard has become the standard of a contest nobody watches anymore.

The right question is not how strong FRITZ 20 is. The right question is how long it keeps a learner in his seat.

And no promotional document answers that, however many pages it runs to.

There is another dimension I consider more important, and it is usually skipped in every debate about chess software: the capacity to endure boredom.

I have sat beside young players through four-hour endgame sessions, repeating the same position until it became reflex. There is nothing exciting in it. There is no golden moment. Only the clock, the sound of a piece set down, and a teacher patient almost to the point of indifference.

Progress in chess is built out of hours like those. Thousands of hours. And training software, in the end, can do only one valuable thing: make those hours a little less lonely.

That is a design problem, not a strength problem.

I wonder whether ChessBase sees it, or whether they are still trying to sell the world a championship machine while the world needs a training partner.

There is a line I often write in commentary, and I find it truer here than anywhere: An empty stadium still holds the applause of the dead.

Fritz once had an auditorium. It had Bonn, it had New York, it had Kramnik sitting across the board, his face pale after move thirty. Those sounds still live in the memory of a generation of chess players. But memory is not a product. It is an ingredient, and ingredients alone do not cook a meal.

I have seen this in another field. In the early 2010s I followed the rise of esports. Its first commentators were mocked as noisy, unprofessional, lacking standards. But they were the ones who created a language for a new sport. The clatter of esports keyboards is also a heartbeat of the modern arena.

Chess computing stands at exactly that moment, and it has not yet found its language.

There is one more thing I want to say, and it belongs to the other side of the story.

For years I told myself chess was a static sport. That there were no injuries, no fitness, no breathless final minutes. I was wrong. I have sat close enough to see a young player's hand tremble as he touched his king on move forty. I have seen people leave the board with red eyes, not from crying but from twelve consecutive days of concentration.

And I have seen what I believe matters most in this trade: a seventeen-year-old boy nobody knows, writing his games into a notebook, unaware that in two years he will stand before the chance of a lifetime. When a seventeen-year-old prodigy does not yet know he is about to become a legend, the only thing he needs is a patient person beside him. Not an engine evaluating +2.34.

That is why I did not write this piece as a product review.

So if you ask me whether to buy FRITZ 20, I will not answer yes or no. I will ask you back: how many hours a week do you intend to use it, and what will you do in those hours?

If the answer is "analyse my games after each round," then Lichess does that for free, right now, on the phone in your hand.

If the answer is "I want a study path," the next question is whether that path is available in Vietnamese, whether it fits the level of Vietnamese students, whether it accounts for the fact that they train twice a week.

If the answer is "I want to support a brand I grew up with," that is an entirely legitimate reason — and the most honest reason I have ever heard from people who buy chess software.

There is nothing wrong with buying a memory. You just need to know what you are buying.

After three decades of writing, I have realised I never wrote about sport. I wrote about people trying to become better than themselves, and about the tools they use to do it.

Chess computing is the best story of them all, because it is the story of a tool that became too good. The machine surpassed humans long ago. It no longer has anything to prove. But precisely because of that, it must now learn a far harder skill: teaching from the beginning to those who come after, with a patience it has never had.

FRITZ 20 may be the first step on that journey. Or it may be a final chapter written in adjectives. We will know in thirty days, when the people who bought it either come back or do not.

Meanwhile the boy with the rook-versus-pawn endgame is still sitting in the corner of the training room. The +2.34 is still on the screen. And if there is any software in the world that makes him understand why he is winning, I will be the first to write about it — not as a product, but as a moment.