FRITZ 20: When a Chess Engine Sits Down to Become a Coach

A small room on the third floor of a training centre on Nguyen Van Linh Street, Hai Chau District, Da Nang. A fifteen-year-old girl sits in front of a computer screen, her left hand resting loosely on the mouse, her right hand propping up her chin. On the digital board is a rook-versus-pawn endgame she lost two hours earlier in a training game against a girl her own age. No spectators. No coach. Just her and a program that is running.

I stopped at the doorway because of the way she asked her question. She did not ask "what is the best move". She asked: "Why did I not see this move back then?" Then she clicked back, replayed the line, and wrote a note in a notebook with a worn spine.

Thirty years ago, in a training room in Melbourne, I sat in exactly that posture. What lay in front of me was an opening book and a mechanical chess computer that clicked every time I pressed a move. The girl has FRITZ 20. The distance between the two generations lies in the fact that the machine has agreed to sit down and become a teacher.

To understand how a fifteen-year-old in Da Nang can sit and interrogate herself in silence like that, we have to go back more than three decades.

Fritz was born in 2026, programmed by Frans Morsch and brought to market by ChessBase, the largest chess software publisher in Europe, headquartered in Hamburg, Germany. Four years later, at the World Computer Chess Championship held in Hong Kong in 2026, a version of Fritz running on an Intel Pentium processor defeated mainframe machines that cost many times more. For the first time, a machine any household could buy beat systems that existed only in laboratories.

Then came the years I had to cover live. In 2026, Deep Fritz drew 4-4 with Vladimir Kramnik in the match billed as "Brains in Bahrain". In 2026, X3D Fritz drew 2-2 with Garry Kasparov in New York. In 2026, in Bonn, Deep Fritz beat Kramnik 4-2. After that night, the question "is there any door left for humans" became a shared obsession across the chess world.

The point few people mention lies elsewhere. Throughout those two decades, the machine could do only one thing: answer. It answered "which move is strongest", "who stands better in this position". It never asked anything back. It never told you: "You are weak in rook-and-pawn endgames, and you are weak there because three months ago you ignored a basic principle about king placement."

For women's chess, that gap is wider. In Vietnam, names such as Hoang Thanh Trang, Pham Le Thao Nguyen, Nguyen Thi Mai Hung, Vo Thi Kim Phung, Nguyen Thi Thanh An, Le Kieu Thien Kim and Bach Ngoc Thuy Duong have proven that talent is not in short supply. What is in short supply is infrastructure: the number of qualified coaches, the density of tournaments, the hours of one-on-one analysis. A male player in Hanoi may find five teachers. A female player of the same level often has one, and that teacher also doubles as manager, logistics officer and psychologist.

In 2026, when domestic tournaments froze, I spent six weeks filming athletes training in empty arenas. The empty corridors of 2026 taught me that the pitch never dies, it only knows how to wait. They also taught me something else: when the cheering stops, I can finally hear the true breathing of the game.

FRITZ 20 arrives at precisely the moment that gap has become most visible.

Three promises, and which one is credible

The publisher introduces FRITZ 20 with three adjectives: efficient, intelligent, individualised. This is the familiar language of every software upgrade of the past twenty years. But in chess, those three adjectives carry their own weight, because they touch exactly the three things a young Vietnamese player lacks.

Efficiency, in the language of the training room, does not mean faster. It means less waste. An average young player spends forty per cent of self-study time on things that produce no progress: copying opening lines already memorised, reviewing games they won, or worse, running an engine over games where they already know exactly where they went wrong. The real value of a training tool lies in showing the learner what to ignore, not in how many million moves per second it can calculate.

Intelligence is when the machine begins to understand a position rather than simply count material. The historic turning point came in 2026, when DeepMind's AlphaZero beat Stockfish by learning from the rules and self-play, without a vast opening database. Since then the Fritz line has also moved towards neural networks, most notably with Fat Fritz. For a trainee, the meaning is not about strength but about the style of evaluation: a traditional engine tells you the position is 0.32 pawns better. A neural engine tells you White should keep the pawn structure intact, trade rooks on the open file, and that the 0.32 advantage will only materialise twenty moves later. The difference between a number and a plan is the difference between a referee and a teacher.

Individualisation is the biggest promise, and the hardest to keep. In chess, personalised coaching was once the privilege of very few. A young player who wanted a personalised path had to pay a grandmaster coach, usually by the hour, and most Vietnamese families could not sustain that cost for years. When a program builds its own profile of the learner - strengths, weaknesses, types of positions where errors recur, rate of progress - that narrow door opens a little wider.

The line between an analysis tool and a teacher is not Elo

This is where I want to linger, because it is the point software marketing usually skips.

An engine rated 3,600 Elo and an engine rated 3,500 Elo make no difference to an amateur player. Both beat that player in every game. What makes a difference is the learning interface: whether the program can break content into exercises, whether it can track a learner's recurring errors, whether it dares to say "you are not ready for this section, come back later".

I have sat beside enough young athletes to see a pattern. Given a raw engine, they tend to fall into a passive state: switch on, run analysis, nod, switch off, forget everything within a week. Given a tool with a structured exercise plan, retention rises markedly, because knowledge is attached to a specific task that must be completed that day.

Tactics are like people: you have to know who you are before you want to become someone else. A machine cannot tell a fifteen-year-old girl whether she is an attacking or defensive player, whether she prefers closed or open structures. It can only lay out the evidence and let her read it. A good tool is one that generates questions, not one that supplies answers.

At beginner level, the greatest value is patience. Beginners need to be treated as beginners: if they are still making mistakes in basic captures, do not present them with complex endgames. At tournament level, the greatest value is precision in opening preparation and handling of middlegame positions. At professional level, the only thing left is saving time - and anyone who has prepared for a long tournament knows that saving twenty hours across two weeks can change an entire career.

The Vietnamese story: half the infrastructure gap is being flattened

What caught my attention about FRITZ 20 is not the software itself, but its consequences for a chess nation like Vietnam.

A young player in Hanoi, Da Nang or Can Tho can buy a software licence for roughly the price of a few tutoring sessions, install it on the laptop already used for online classes, and hold in their hands the same tool a grandmaster in Europe is using. Technically, the gap is close to zero.

But in terms of opportunity, the gap remains. No software creates an international tournament on home soil. No software pays for plane tickets, hotel rooms or entry fees. A machine can teach you how to play a queen-versus-rook endgame, but it cannot teach you how to stay calm when the clock shows thirty seconds and your opponent is a real person sitting two metres away.

Drawing on my experience of watching matches over many years, I have observed that Vietnamese female players often progress quickly in the early phase - the phase decided by technique and memory - and then plateau in the phase that requires international exposure. The cause is not computational ability. It is the number of elite games a person is allowed to play each year. A young player in Europe may play thirty open tournaments in a season. A young player in Vietnam may get five.

A good training tool cannot replace the five missing tournaments. It only ensures that when opportunity arrives, that player does not squander it through elementary errors that should have been fixed two years earlier.

What a typical training session looks like

Let us return to the room in Da Nang.

Under the traditional model, the girl in my story would wait until next week's lesson, bring her notebook, read the game aloud to her coach, and receive advice. That cycle takes seven days. Within those seven days, the memory of losing has cooled, and the lesson is compressed into a note without context.

Under a model supported by a tool, that cycle shrinks to two hours. The feeling of losing is still raw. The girl retraces every move herself, discovers on her own that on move thirty-two she pushed a pawn on the queenside instead of bringing her king across, and from that point the whole endgame changes direction. When the coach arrives the following week, the conversation is no longer a one-way lecture. It becomes a dialogue between two people who have examined the same data.

Pedagogically, this is a bigger change than any Elo number.

What the software cannot do

I want to use this section to speak plainly, because I have seen far too many cases where someone bought the software and nothing changed.

Buying FRITZ 20 does not buy Elo. Software is a tool, and a tool only has value when someone uses it with discipline. A good piano does not create an artist. In chess, that discipline means sitting down with your own games in your clearest state of mind, accepting that you were wrong, and writing down why.

The second risk is what I call the test-tube player. These are people who can analyse a complex position at a screen with high accuracy, but who, facing a real opponent with a running clock and no undo button, cannot find a move. The cause is simple: analysis is a skill without time limits, while competition is a skill with time limits. The two do not automatically convert into each other.

The third risk is subtler. When every young player in the world trains with machines from the same lineage, playing styles tend to converge. The odd moves, the ideas stamped with individual character, the handling of positions the machine rates as slightly inferior but which irritate human opponents - all risk disappearing. Women's chess has historically been rich in personality: Judit Polgar attacking like a storm, Alexandra Kosteniuk cold in complex positions, Vietnamese players resilient in defence. If an entire generation learns from a single textbook, that diversity thins out.

But I do not want to end this section on worry. There is one thing the machine has achieved, and it is genuinely liberating: it has removed the gatekeeper. Previously, whether a young girl gained access to elite chess knowledge depended on whether someone was willing to teach her. Now the prerequisite reduces to a computer and curiosity. That is a structural change, and it cannot be reversed.

What is changing

The girl in Da Nang will not become a grandmaster simply because FRITZ 20 is installed on her machine. That road is still long; it still requires tournaments she has not been invited to, opponents she has not met, and years during which her family must weigh studying against playing.

But one thing has changed compared with thirty years ago, when I sat in that room in Melbourne with an opening book and a mechanical machine. Back then, I had to wait for someone to come and teach me. The girl now, at eleven at night, after the whole house has gone to sleep, can open her laptop and ask herself the hardest question.

When a generation learns how to ask its own questions, all it needs next is the opportunity to answer them.