FRITZ 20 and the Hard Position of Commercial Chess Software

In October 2026, in Bonn, Vladimir Kramnik lifted his knight off the board. He stared at the position for four minutes, set the piece down, and immediately realised what fifteen million television viewers had already seen: the machine across from him had a checkmate in one. It was game six, and a piece of software called Fritz had dethroned the world chess champion on the human's own stage.

Nearly two decades after that story, I read a product announcement from the very company that built that machine. The article mentions a product called FRITZ 20. I counted eleven declarative phrases in it: a revolution in training, a turning point, a personal coach, efficient, intelligent, individualised. I also counted the technical specifications: none. The performance metrics: none. The head-to-head comparisons: none.

Moscow 2026 – people remember the goals. I remember the empty space down the right corridor. Here, that empty space is wide enough that it deserves a long look.

From the touchline, I see the whole game. And in the match between a long-established chess software house and a market that has fundamentally changed, the most interesting thing is what the advert does not say.

---

Context: a system that ran for four decades

To understand why a seven-paragraph product announcement can be an important signal, it needs to be placed on the correct board of the chess software industry.

ChessBase was founded in 2026 in Hamburg, West Germany, by Frederic Friedel and colleagues. For more than thirty years the company held a near-monopoly in game databases and analysis tools for professional players. The ChessBase database became standard infrastructure: every serious coach, every national team player, every chess journalist once installed it on a personal computer.

The Fritz engine in its early phase was developed by Frans Morsch and Mathias Feist. Its name is tied to a series of iconic human-versus-machine matches. In 2026, in Bahrain, Deep Fritz drew 4-4 with world champion Vladimir Kramnik. In 2026, in New York, Garry Kasparov drew 2-2 with Deep Fritz. In 2026, Fritz beat Kramnik 4-2 in Bonn – including the famous game six, with a blunder recorded in every book on competitive psychology.

But here is the crucial point. The era of Fritz's dominance coincided with a period when engine strength was still a commercial advantage. When the strongest software in the world was something you had to buy, selling the strongest engine was a viable business model.

That model began to crack from within the open-source field.

Stockfish, initially developed by Tord Romstad, Marco Costalba and Joona Kiiski, gradually became the benchmark of engine strength while remaining entirely free. Leela Chess Zero arrived in 2026, built on neural networks and reinforcement learning, a community result of the AlphaZero revolution. In December 2026, DeepMind published its work on AlphaZero – a self-play algorithm that reached superhuman level in chess, shogi and Go. That event did not merely change how engines were built; it changed how people expected engines to be built.

In parallel, online platforms reshaped the entire experience. Chess.com became the largest chess venue in the world with tens of millions of accounts, integrating free engine analysis, tactical puzzles, courses and news. Lichess offered a fully open-source version at no cost, with Stockfish running in the browser.

The result is a harsh reality for any commercial software that sells analytical strength: a user can open a browser tab and receive analysis at grandmaster-superhuman level for free.

That was the system in operation until the FRITZ 20 advert appeared. The system does not lie, but it can only be heard when the data is thick enough. And here, the silence of the data is itself the data.

---

Core analysis: reading an advert like an opening

People watch the players run. I watch the whole block move. With a product announcement, the moving block is the information structure: what is brought into the light, what is held back, and the order in which they appear.

Eleven promises and one void

The article positions FRITZ 20 across three layers of language. The first is the product layer: this is a chess engine. The second is the elevation layer: it is not just an engine, it is a training revolution. The third is the audience layer: it is for ambitious players competing at tournament level, and also for those taking their very first steps.

These three layers do not stack naturally. A product serving both someone learning the rules and a professional grandmaster is rare in any industry. A beginner's instrument is not a concert artist's instrument. An apprentice carpenter's tools are not a master's tools. When a product claims to serve both ends of the ability spectrum, there are usually two possibilities: either it is genuinely an excellent multi-tier platform, or it has not yet found its own anchor point.

To distinguish between those two possibilities, you need data. And the data does not appear.

The article names no algorithm, no underlying engine version, no Elo strength figure, no opening database size, no price, no minimum system requirements, and above all, no release date. For a product carrying the number 20 – implying the twentieth version of a line with more than two decades of history – the absence of a time marker makes it impossible to place it on any comparative axis.

What is notable is not that the advert lacks specifications. What is notable is that the lack of specifications has structure. It is not a writer's random omission. It is a consistent editorial choice: every sentence can be verified by feeling, no sentence can be verified by number.

The three pillars of a campaign without coordinates

Looking more closely, the information structure rests on three pillars.

Pillar one: repositioning. The key phrase is personal chess trainer. This phrasing moves the product from the territory of engines into the territory of pedagogy. It is a strategically sound move, because the engine territory has been commoditised. When computational strength becomes a free commodity, added value must be created elsewhere – at the level of learning experience, of habit, of feedback.

Pillar two: audience expansion. Targeting both beginners and professionals is often not an ambition to serve, but a consequence of not yet having identified the actual paying customer. In the chess market, beginners tend to use free tools; professionals tend to use databases and open-source engines combined with human coaching. The sustainable paying segment sits in the middle: ambitious club players, parents buying for children, serious but non-professional players.

Pillar three: naming the emotion. This is where words like revolution, turning point and breakthrough appear. They do not describe features; they create a mental state. In software marketing, this is a sign that the product is hard to sell on specifications.

The competitive board: who holds the centre

To assess a move, you must look at the whole board.

On the pure-strength square, Stockfish dominates. It is free, open-source, continuously updated by the community, and serves as the underlying engine on countless platforms. No commercial software can compete head-on on this square by selling computational strength.

On the machine-learning square, Leela Chess Zero represents the neural-network school. It is also free. It demonstrates that an AlphaZero-style approach can be reproduced by a community without the resources of a giant technology corporation.

On the integrated-platform square, Chess.com and Lichess hold the advantage. They do not sell engines; they sell ecosystems. Free analysis is the bait, and the real value lies in subscriptions, online tournaments, community and content.

On the structured-training square, platforms such as Chessable and Play Magnus (before consolidation) compete through course content, spaced repetition methods and interaction.

So where does ChessBase stand? On the professional-database square and the brand-history square. These are strong squares, but they are narrowing. Open databases, free game data sets and the convenience of online platforms continuously erode the exclusivity advantage.

On this board, positioning FRITZ 20 as a training revolution is a defensive move, not an attacking move. It concedes that the engine-strength battle is lost, and shifts to another battle – one whose rules are not yet clearly defined.

The problem with shifting to a battle without rules is that no one knows what winning means. If the product is a training tool, its measure must be learner progress. But progress is a hard metric to measure, slow to show, and dependent on hundreds of variables outside the software. This is precisely why adverts in this field avoid numbers and choose emotion.

Reading history to understand the present

One historical detail is worth remembering. The period when Fritz beat Kramnik in 2026 was the peak of commercial engine power. At the time, software beating a world champion was global news, because it proved a measurable claim: the machine is stronger than the human.

Two decades later, the question has changed. No one still argues that the machine is stronger. The argument has shifted to: how does the machine help the human improve, and who can sell the method of helping.

This is the point most observers miss. The shift from selling strength to selling method is not a decline; it is the maturation of a market. But that maturation imposes a far harsher requirement than selling engines: when you sell a method, you must prove the method works, and prove it with learner data, not with adjectives.

And that data does not appear in the FRITZ 20 announcement.

A lesson from being pelted with criticism

My first article got pelted with criticism. Data never takes offence.

In 2026, when I wrote an analysis of a Bundesliga club's gegenpressing model, the online community reacted fiercely. They called me a conservative. I did not argue. I spent six weeks rewatching footage, noting more than four hundred failed pressing situations, and published a correction with specific data.

That lesson applies intact to this case. A claim about a product cannot stand if the person making the claim refuses to be verified. And the fastest way for a claim to avoid verification is to keep it at the emotional level.

What would change my assessment

If the FRITZ 20 announcement came with any of the following data, my assessment would change immediately.

First, an ACPL comparison – average centipawn loss – between FRITZ 20 users and a control group after a defined training period. This is the standard measure for move quality, and it has long existed in engine analysis.

Second, the move-matching rate against optimal engine suggestions, measured across training phases.

Third, a description of the personalisation algorithm: on what criteria the system selects exercises, how it updates difficulty, and by what mechanism it addresses learner weaknesses.

Fourth, a release date and a price. The two minimum pieces of information for consumer decisions.

None of these four appeared in the original article.

The structure of an unanchored belief

There is a principle in chess analysis: when a position has too many equal options, people tend not to find the best move, but the psychologically safest. In marketing, the same phenomenon occurs: when a product has no measurable advantage, people sell reassurance.

FRITZ 20 is sold as a promise about a journey: you will get better, efficiently, intelligently, individually. This is a promise anyone wants to believe, and anyone finds hard to verify immediately.

But precisely because it is hard to verify immediately, the promise becomes fragile against time. A user buys, uses it for three weeks, sees no clear change, and concludes they paid for a feeling.

The difference between a good training product and a well-advertised training product is this: a good product makes the user not need to believe in it, because they see progress before they get a chance to believe.

A view from the market structure

From the touchline, I see the whole game.

There is a recurring pattern in the software industry at large. When foundational technology becomes a commodity, companies that once lived off foundational technology must move up to the application layer. IBM once sold computers; now it sells cloud services. Nokia once sold hardware; now it sells network infrastructure. Adobe once sold boxed software; now it sells subscriptions.

ChessBase stands at exactly that crossroads. The engine is a foundational layer that has been commoditised. The database is a foundational layer being commoditised. The application layer – personalised training – is where marginal value still exists.

But the application layer demands a different competence: the ability to design a learning experience, not the ability to compute. This is a difficult shift, and it explains why the advert reads like an attempt to name a product before the product has named itself.

Numbers speak, but the numbers are absent

In chess analysis, a distinction is drawn between static and dynamic positions. Static is the structural position: material, pawn structure quality, king safety. Dynamic is potential: open lines, attacking chances, tempo.

A product advert has these two faces as well. The static face is what is presented: adjectives, promises, positioning. The dynamic face is market potential: whether the product can survive and grow in a competitive environment.

The static face of the FRITZ 20 advert is thick. Its dynamic face is thin. And this is worrying for a product line that has existed for two decades.

A long-lived product line has a special advantage: historical data. It has loyal users. It has a community. It can demonstrate evolution across versions. A twentieth-version announcement that fails to use that heritage is an announcement squandering its own greatest asset.

This leads to a hypothesis. Perhaps the silence on specifications is not concealment, but a real void. Perhaps the twentieth version does not bring a technical leap large enough to boast about, so the communications campaign must speak in another language.

A comparison with how rivals present themselves

A simple comparison for orientation.

Stockfish publishes its source code, publishes match results at engine championships such as TCEC, publishes its development progress. Users can install it, run it, measure it.

Leela Chess Zero publishes training batches, publishes self-play game counts, publishes strength progression over time. The transparency here is extreme.

Chess.com publishes user metrics, publishes daily game volumes, publishes new features with each release cycle. It is a platform, and it speaks in numbers.

A commercial engine has a legitimate reason not to open its source. But it has no legitimate reason not to publish any metric related to performance and features. Secrecy about algorithms and secrecy about user value are two entirely different categories.

When a product keeps its algorithm secret, that is protecting intellectual property. When a product keeps its price, release date and features secret as well, that is the sign of an unfinished positioning.

A slow but correct move

Fairness toward ChessBase is required. The shift from selling an engine to selling a training experience is the right direction. The market has gone there, and arriving late is still better than not arriving.

Patience is not stillness. Patience is waiting for the opponent's pressing rhythm. In this case, the opponent is not a specific company, but the change in consumer expectation itself. Today's consumers expect transparency before they pay. They expect data before they believe.

A communications campaign that does not meet that expectation will lose exactly the customers it needs most: those serious enough to buy a product and careful enough to skip it when evidence is missing.

---

Contrarian angle: what the ambiguity conceals

An analysis that stops at criticising the lack of data has not gone the whole distance. Because if we look only at the advert, we conclude ChessBase is weak. But if we look at the whole board, there is a reverse reading, and that reading is more interesting.

Ambiguity may be a psychologically optimal choice

Imagine the situation: a commercial chess software house, in 2026, launches a new engine. Its engine strength almost certainly does not exceed Stockfish – that is nearly impossible, since Stockfish is the accumulated result of hundreds of volunteer developers worldwide. If the company published an Elo comparison with Stockfish, the table would show a gap, and that gap would itself announce the obsolescence of the positioning being sold.

Conversely, if the company published training metrics, those metrics require a long time to prove, require controlled studies, require a sufficiently large learner sample. This is an extremely expensive class of evidence, and the chess market is not large enough to invest in it easily.

In that situation, a communications campaign built on value language rather than specification language is not foolish. It is commercially rational. It bets on a characteristic of chess consumers: most buy on inspiration, not on analysis.

But betting on inspiration carries a price. It creates the most easily abandoned customer base possible: people who buy because they believe, not because they see.

Where the blind spot lies

The blind spot is not in the strategy. The blind spot is in the measurement model.

When a company shifts from selling a measurable product to selling a hard-to-measure one, it usually does not build a new measurement system. It only changes the language. Communications change first, the product changes later, and measurement capability does not change at all.

The consequence is that for years the company believes it has transformed. It has renamed the product, changed the packaging, changed the message. But it has not yet done the only thing that matters: build a bridge between the product and learner outcomes.

In chess, this is especially dangerous because chess learners find it very hard to measure their own progress. Elo moves slowly, fluctuates widely, and depends on opponents. A club player needs months to see a small change. A beginner needs weeks just to become familiar with tactical patterns. During that time, the feeling of progress is easily confused with the initial inspiration of a new product.

And here is the collapse point: inspiration fades after a few weeks, while real progress only begins to show after a few months. If a product has no mechanism to retain users across that lag, it loses them exactly at the decisive stage.

A paradox of free tools

There is a rarely mentioned paradox when discussing Stockfish and Lichess. Free does not mean easy. The strongest analysis tool in the world does not automatically turn a user into a better player. It produces raw data, and interpreting raw data requires knowledge the user does not yet have.

This is the gap that commercial software can fill. But to fill it, the software must do what a free engine does not: turn analysis into lessons, turn correct moves into habits, turn habits into reflexes.

That is a pedagogical problem, not a technical one. And a pedagogical problem has its own measure: does the learner improve faster or slower than a control group. There is only one way to prove it, and that way is publishing data.

Whoever solves the intermediate problem between the raw data of a free engine and the learning needs of the ordinary player does not need to call itself a revolution. The market will do the naming.

The dark side of targeting everyone

Another argument deserves the table.

When a product targets both beginners and professionals, the design burden grows exponentially. An interface for beginners must be simple, intuitive, hiding complexity. An interface for professionals must be detailed, fully customisable, hiding nothing. These two requirements directly conflict.

The common solution is tiered display modes. But modes do not solve the core problem: content. Beginners need lessons on basic opening principles. Professionals need databases of millions of games with sophisticated filters. These are two different products, not two modes of the same product.

When a company dares not choose a customer segment, it is usually because it fears losing anyone. But in a saturated market, not choosing is the most dangerous choice. Beginners will leave because the product is too complex. Professionals will leave because the product is not deep enough. And the people in the middle – the most sustainable segment – will never find themselves in the message.

What to learn from the silence itself

Here, I want to return to an old principle of mine. The system does not lie, but it can only be heard when the data is thick enough.

In the FRITZ 20 case, the system is the advert. And the system is saying something very clearly: this is a product with an unfinished positioning. It is a move in a transition phase, not a decisive blow.

This is not bad. Many successful products begin with transitional moves. But it imposes a requirement on the observer: do not judge a product by its claims, judge it by the space around those claims.

And for consumers, it imposes a simpler requirement: wait. Wait for independent reviews. Wait for the feature list. Wait for real user data. There is no urgency in buying a tool you will use for years.

A note on the wider context

Zooming out, the FRITZ 20 case is a small manifestation of a larger phenomenon: the commoditisation of artificial intelligence.

When AlphaZero appeared in 2026, the chess world marvelled at a self-learning algorithm reaching superhuman level. Eight years later, open-source communities have reproduced and improved similar techniques. What was once a scientific breakthrough has become foundational tooling.

This commoditisation is not happening only in chess. It is happening in machine translation, image recognition, speech synthesis, and will continue to expand. Business models based on selling access to an AI model become harder to sustain when equivalent models are released for free.

In that world, value shifts from model to application, from algorithm to experience, from strength to convenience. And in that world, the winner is not the one with the strongest technology, but the one who best understands what users need from that technology.

Fritz was once the strongest technology. FRITZ 20 is trying to become the best understander of users. These are two different games, and switching games mid-match is never easy.

---

What I got wrong

I owe readers this section, following a habit I have kept since 2026.

In 2026, before the World Cup in Qatar, I wrote six articles predicting Argentina would be eliminated in the quarter-finals because their defence was too thin. I was wrong. I watched the final live, saw Lionel Scaloni adjust the team's spacing after going 2-0 down to keep the match's rhythm, and the following week I published a self-critique of nearly five thousand words. What I had underestimated was the depth of the bench and the coach's ability to read the game.

On this article, I may also be wrong in three ways.

First, I may underestimate ChessBase's communications strategy. Perhaps the company deliberately holds back specifications to reserve information for other channels, launches and sponsored independent reviews. If so, this advert is only the first piece of a larger campaign, and my assessment will need adjustment when other pieces appear.

Second, I may underestimate the product's real technical strength. Absence of evidence is not evidence of absence. A lack of specifications in communications does not prove a lack of specifications in the product.

Third, I may underestimate brand strength. Fritz is a name several generations of players grew up with. With a loyal customer base, a brand can sell even without specifications. Loyalty is a form of capital that is not measured in Elo.

I leave these three possibilities open, and will return to verify when the data is thick enough.

---

Takeaway: which move to watch

I am sixty-nine. I still learn from the young ones. Football does not retire, and neither does chess.

If I had to set one question to track over the next twelve months, it would be: will ChessBase publish a learner-progress measure for FRITZ 20.

That is the decisive question. If the answer is yes, this product line enters a new phase, competing on results rather than claims. If the answer is no, the product will continue to exist in a grey zone – enough to sell to believers, not enough to convince sceptics.

For chess players, my recommendation is simple. Before buying any training tool, ask yourself: how will I measure my progress over the next three months. If you cannot answer that question, no tool will answer it for you. If you can answer it, you will know what tool you need.

As for the market, I hold an old belief. The strongest technology always becomes free. The remaining value belongs to whoever turns free technology into personal results. That game has only just begun, and it will not end with a software version.

It will end with a generation of players who understand themselves better.