FRITZ 20 and the Repositioning of the Chess Engine in the Age of Free Analysis

On a Friday evening at a small chess club on Nguyen Thien Thuat Street in Nha Trang, a fourteen-year-old student opened his laptop and asked me the question I have heard at least twenty times in the past three years: "Teacher, if I install good software, will I play better?" He had installed FRITZ 20 three weeks earlier. He had opened it seven times. Then he went back to Lichess, where he solves puzzles every night before bed and has never paid a single dong.

I did not answer right away. I pulled out my 2026 analysis notebook — the one I used while working as a sports science researcher for Sanna Khanh Hoa BVN — and showed him a page covered in arrows. For six straight weeks I had dissected a single movement by Quang Hai: every time he dropped five metres deeper, the opposing back line stretched by 4.2 metres. He looked at the page, looked at the FRITZ 20 screen, and asked: "So can the software draw this for me?"

That is the right question. And it is also the question that the FRITZ 20 product announcement, published by ChessBase itself, does not answer.

What matters is not where the ball stands, but where the ball is about to fall — that is the true spatial piece. In chess, the principle translates into something simpler: what decides a game is the square a piece is about to occupy, not the square it currently sits on. And what decides the value of a training tool is the data it produces, not the adjectives it uses to describe itself.

A thirty-five-year-old engine at a crossroads

ChessBase is Germany's long-established chess software publisher, headquartered in Hamburg. The Fritz line began in 2026, developed by Frans Morsch and Mathias Feist. The name Fritz entered chess history through two landmarks: the 2-2 draw of X3D Fritz against Garry Kasparov in New York in 2026, and Deep Fritz's 4-2 victory over Vladimir Kramnik in Bonn in 2026. For Vietnamese players born in the late 1980s and early 1990s, Fritz was the first software they installed on a desktop computer — the thing that beeped and flashed "Mate in 5" when they were just learning the game.

Thirty-five years later, the board looks completely different. Stockfish is a free, open-source engine that has consistently topped engine strength rankings. Leela Chess Zero brought neural networks into analysis and produces evaluations that feel more human. Lichess offers a full analysis toolkit for free, with no game limit. Chess.com packages an engine into a structured learning experience with monthly subscriptions. The "computational strength" component of chess has been commoditized to the point where it is free, universal, and no longer a competitive advantage for anyone.

The FRITZ 20 announcement recognises that reality but chooses a different response. It does not sell engine strength. It sells something bigger: a "training revolution," a "personal chess trainer," a training solution that is "efficient, intelligent, individualised" — for ambitious players, for those already competing at tournament level, for professionals, and also for those taking their first steps.

One thing must be said plainly about the source. This is material published by the manufacturer itself, with no third-party verification, no publication date, no technical specifications, no price, and no specific feature list. In my line of work, such a source is filed under 'cross-reference before citing.' I cross-referenced it, and what follows is what I found — along with what I did not find.

Measure first, conclude later: four unfilled spatial pieces

The first piece is segmentation. A product that claims to serve both those taking their very first steps and grandmasters preparing for top-level tournaments is usually describing two different products in one box. A beginner's needs are pattern recognition, repetitive tactical exercises, and the feeling of winning. A tournament-level player's needs are deep opening trees, specific opponent preparation, and endgame accuracy down to the single move. These two ends demand two different interface architectures and two different pedagogical philosophies. No tool serves both equally well without weakening one of them.

The second piece is measurement. In sports science, we do not say 'this player plays better.' We say this player has an 87 percent pass completion rate, covers 11.2 kilometres, and makes 6.4 entries into the final quarter of the pitch per match. Chess has an equivalent index system, and the international analysis community uses it daily: ACPL — average centipawn loss per move, the rate at which moves match the engine's top recommendation, calculation depth by game phase. The FRITZ 20 announcement offers none of these. For a product that calls itself a training revolution, the absence of a results-measurement index is the largest gap of all.

The third piece is economic structure. ChessBase built its brand on a boxed-software model: pay once, install on your machine, use for years. That model is under pressure from two directions — from the free side, where Lichess provides unlimited engine analysis, and from the subscription side, where Chess.com turns chess learning into recurring revenue. Within that structure, moving Fritz from 'strong engine' to 'integrated training platform' is a strategically rational move. ChessBase's pressing problem is not engine strength; it is how they redraw the market map — where the biggest rival is no longer Stockfish, but the user's habit of getting things for free.

The fourth piece is the Vietnamese user. A Fritz licence typically runs between 80 and 100 euros, roughly 2.2 million to 2.7 million dong at current exchange rates. For a middle-school student in Nha Trang learning chess at a local club, that equals three months of tuition. For a provincial coach teaching twenty students, it is a decision weighed between software and a professional training course. The paradox is that the group most in need of a tool with pedagogical structure — young coaches in the provinces — is the group most sensitive to price.

From these four pieces I draw one verifiable judgment: if FRITZ 20 truly is a training revolution, it must prove itself through training data, not through adjectives. A measurable revolution would say: after eight weeks, by how much did users' ACPL fall, by how much did endgame win rates rise, by how many seconds did tactical pattern recognition time shorten. Without those indices, what we have is a handsome software box and a very skilfully written marketing story.

I remember the 2026 livestream, when the V.League was postponed indefinitely and I surveyed twelve clubs on the effects of playing without crowds. Home teams' pass completion dropped 7 percent, and three hundred people joined the session where I shared the results. What I learned was not in the data table: people want tools that help them understand, not tools that overwhelm them. A chess software is the same. Tactics only become complete when they are told in a language the players dare to believe — and no fourteen-year-old believes an interface that only praises or scolds.

If I had to grade a training tool

I have spent eighteen years watching the sports industry and recording how teams build their training processes. From that experience, I grade a training tool on five criteria, all of which can be checked with data.

Criterion one is the ability to generate data over time. A good tool stores a user's game history and shows the trajectory of core indices. Criterion two is a clear pedagogical structure: lessons divided by phase, each phase with measurable objectives. Criterion three is the ability to adapt to individual weaknesses instead of dumping thousands of identical exercises on everyone. Criterion four is integration with real competitive data — the user's games on platforms, not just practice games inside the software. Criterion five is cost per actual hour of use, a division most buyers forget to perform.

Applying these five criteria to the FRITZ 20 announcement, I cannot score a single one. The document does not provide enough information to score. In research, this case is recorded as 'insufficient data to conclude,' which is entirely different from 'poor product.' I want to be clear about that, because I have no intention of convicting a product I have never installed and tested. What I criticise is the way the story is told.

How to read a product announcement

When a manufacturer is both the source and the seller, the reader needs a filter. My filter consists of four questions.

Does the announcement contain any specific figures, or only adjectives? If only adjectives, you are reading advertising, and the next step is to find an independent source. Does it name competitors and compare directly? If not, the product likely has no advantage on any comparable criterion. Does it state a publication date and version number? If not, you cannot know whether you are reading this year's document or an old one. And finally, does it define its target user through a single portrait? If it targets everyone, it usually targets no one precisely.

These four questions require no chess knowledge to answer. They only require one habit: reading more slowly than the writer wrote.

The counterintuitive angle: the blind spot belongs to the user

There is a reverse reading worth considering. ChessBase's decision not to include technical specifications may be deliberate, and it reflects a genuine market reality: people buying chess software today rarely choose on engine strength. They already have Stockfish. They choose on experience, on interface, on the feeling of being guided. From that angle, 'training revolution' accurately describes the function of a boxed product, even if the term is overused.

But the blind spot lies elsewhere, and it belongs to the user more than the manufacturer. Across eighteen years of observation, I keep seeing the same pattern: players buy tools to feel they have improved, then leave the tools sitting idle on their hard drives. Software does not create discipline. In chess, the gap between a 1600 player and a 2026 player is not decided by how strong the engine is, but by the number of hours of structured analysis and the number of times one endures looking back at one's own mistakes. The best tool can only shorten that road; it cannot walk it for you.

The second blind spot lies with the Vietnamese chess community. We have Le Quang Liem, Nguyen Ngoc Truong Son, Pham Le Thao Nguyen — names that have proven Vietnamese players compete at world level. But the training infrastructure behind them remains thin: no standardised game archive system, no long-term progress-tracking data for junior cohorts, and too few people doing the work of recording, measuring, and cross-referencing. Buying more software will not fill that gap. Building a data-driven training process will.

The transfer market is where economic figures are dressed up as on-pitch dreams — and the training software market is the same, except that here the costume is pedagogical terminology. In both markets, the alert buyer is the one who can ask a single question: which of my indices does this product change, over what period, and how do I measure it?

What I will do starting next week

If you are considering FRITZ 20, I suggest a simple test. Pick a single index — your average centipawn loss over your last twenty games, for instance — measure it before installing, use the software on a fixed schedule for eight weeks, then measure again. If the index does not move, you have saved money and gained a lesson worth more than the software. If it does move, I would very much like to read the result, because it would be the first independent data point the Vietnamese chess community has about a product calling itself a revolution.

That Friday evening, I did not tell my student to buy or delete FRITZ 20. I told him to write his index in a notebook and solve ten tactical puzzles every night, on any platform. He agreed. The next morning he messaged me one line: "Teacher, I measured it." I think that is the real training revolution — and it does not come inside any software box.