FRITZ 20 and the Training Revolution: When the Machine Becomes Vietnamese Chess's Coldest Coach

11 p.m. in a small training room in Hanoi, a 19-year-old player sits in front of two screens. On the left, his game from the national junior championship. On the right, a machine answering in numbers: +0.34. He stares at that number longer than he stares at his opponent in memory. I sit behind him, timing every position where he pauses, and write in my notebook something nobody in Vietnamese chess would have imagined fifteen years ago: the strictest, most patient and coldest coach of this generation is not a grandmaster. It is a piece of software.

That software has a new name. FRITZ 20, the latest release in the longest-running active line of chess engines, was published by ChessBase in November 2026. To part of the fan base, this is news for tech people. To me, it is a training milestone. Because in a sport where every move is recorded, what changes the game does not lie in the best move. It lies in the ability to explain why a move is best.

Context

Vietnamese chess has a beautiful paradox. At the elite level we have Le Quang Liem — once among the world's top 20 and world blitz champion in 2026 — alongside Nguyen Ngoc Truong Son, Pham Le Thao Nguyen and a generation of female grandmasters holding steady positions at the Olympiads. At grassroots level, the number of children learning chess has grown so fast that clubs run waiting lists. The gap between those two layers is not talent. It is method.

For years, chess coaching in Vietnam rested on three pillars: translated opening books, an experienced teacher, and whatever engine happened to be installed on an old computer. Those pillars are not wrong. They are missing one thing: data about the learner himself. A good coach can remember that a student struggles in rook endgames, but cannot remember exactly how much advantage that student lost on move 32 of a specific opening structure. The machine remembers. That is where everything starts.

Core

First, a clarification: a chess engine is not an answer book. It is an evaluation system. When the machine prints +0.34, it is not saying White wins. It is saying that with best play from both sides, White holds an advantage worth about a third of a pawn. The unit is the centipawn — one hundred centipawns to a pawn. An entire modern culture of chess analysis is built on that tiny unit.

FRITZ 20 inherits a serious history. Fritz appeared in 2026, developed by Frans Morsch and Mathias Feist, and became the name attached to milestones fans do not forget: in 2026 a Fritz version beat Garry Kasparov in a blitz event in Munich; in 2026 Deep Fritz drew 4-4 with Vladimir Kramnik in Bahrain; in 2026 Deep Fritz beat Kramnik 4-2 in Bonn. This is a line that has stood in elite playing halls longer than most professional players have been alive.

But what makes FRITZ 20 sits in how it organises training, not in its past. Three feature groups stand out.

First, neural-network evaluation. Modern engines no longer rely purely on classical search; they use networks trained on millions of games to sense a position. For a learner the consequence is concrete: balanced positions are judged better, and strange-but-correct moves — the kind the human eye skips — are flagged earlier.

Second, adjustable difficulty. This is the least discussed part and the most decisive for beginners. A supremely strong engine is useless in training, like putting a novice martial artist against a world champion: what does he learn? FRITZ 20 lets playing strength drop to a suitable level, but in a controlled way — tactical logic stays intact while search depth and accuracy are reduced. For a nine-year-old in Da Nang, that opponent is hard enough not to be boring and real enough not to teach bad habits.

Third, the training toolkit. Blunder checking in your own games, opening drills along a variation tree, endgame practice with a database, cloud analysis. This is where the software steps out of the referee's chair and takes the teacher's. It is also where it differs from an ordinary game-viewing site: it stores the user's error history over time.

I followed a group of young Vietnamese players using these tools over three months. How they read data differed sharply. The group that read metrics as an indictment — move 27 lost 1.2 pawns — improved slowly. The group that read metrics as a question — why did move 27 lose 1.2 pawns when I believed it was the best move — improved far faster. Same tool, two outcomes. People call that a shock; I call it unread data.

And here is something the coaching world rarely says out loud: the centipawn score does not replace emotion, it explains why our hand shakes when we touch a piece. A game is not decided on move 27. It is decided in the moment a player spends 40 seconds on a move the machine needs 0.02 seconds for.

Contrarian

There is an assumption the whole chess world takes for granted, and I am not sure it holds: that a stronger engine makes a better trainee.

The social consequence of digitised training deserves re-measuring. When every young player uses the same database, the same opening table, the same yardstick of move accuracy, individual style gets sanded smooth. Not because anyone forbids it, but because deviating from the best move is treated as wrong. Ten years ago a player could become famous for an odd style. Today, how long does an odd style survive before it is called objectively wrong?

This is what the data taught me, and it holds for chess and football alike: correlation is not causation. A player with 92 percent accuracy will not necessarily beat one with 87 percent. Chess is decided in positions no metric captures: time, nerve, and the capacity to tolerate ambiguity. The machine has no ambiguity. Humans live in it.

And age is the one variable that never lies. For a 45-year-old player, the greatest value of FRITZ 20 is not finding stronger moves. It is saving time. A young player can sit six hours reading variations. An older one must choose. A tool that pinpoints the real weakness across the last thirty games saves hundreds of hours — and at elite level, time is the scarcest resource, scarcer than talent.

Takeaway

The most striking thing about FRITZ 20 is not inside the software. It is in the question the software forces federations to answer: who is responsible for teaching young players to read data? In many countries that is a job title. In Vietnam it is usually a part-time coach who teaches in the evening and studies analysis after midnight.

Numbers are asceticism: you must give up comfort before you can see the truth. A machine can point at that truth in two seconds. Turning truth into a move takes a human patient enough to stay after the machine is switched off.

12:40 a.m. The 19-year-old turns off the right screen but leaves the left one glowing. He copies four positions into a notebook — with a pen, not a machine. I close my own notebook. That may be the best signal going into the next round.