FRITZ 20 and the Chess Training Turning Point: When the Engine Leaves the Analysis Board for the Coach's Chair

2:47 in the morning. The screen is still on in a small apartment in Hanoi. On the board sits a seven-piece endgame: the white king on g6, the black king on e7, a black rook on a1, three white pawns on f4, g5 and h4. A fourteen-year-old boy has been sitting at the keyboard for two hours, rewinding the same recording again and again. He knows he lost this game. He just does not know at which move he lost it. Then he switches on the analysis engine. Three seconds later, a blue line draws out a move he had never considered. The move is not pretty. It looks almost foolish: the black king voluntarily retreats into the corner, surrendering the entire centre to the opponent. The evaluation on screen swings from -0.3 to -2.1 in a single move. What silences the boy is not the move itself. It is the question that comes with it: if that move exists, why did his mind never once sweep across it during those two hours? This is the question every chess engine in history can answer but cannot resolve. The engine tells you the right move. It does not teach you how to find the right move yourself. That gap is exactly what FRITZ 20, the newest release from ChessBase, claims to close with a different positioning altogether: not the strongest computer, but the personal coach of each individual player. CONTEXT: FROM SCORING MACHINE TO TEACHING MACHINE To understand why a chess product deserves serious analysis, you have to go back to its own starting point. FRITZ was born in the early 1990s, developed by Frans Morsch and tied to ChessBase, Europe's largest chess software publisher. For roughly two decades, FRITZ sat at the centre of every argument about the limits of humans against machines. In 2026, at a tournament in London, the Fritz 3 version became one of the first programs to defeat Garry Kasparov in an official game with time controls. In 2026, Deep Fritz drew Vladimir Kramnik in an eight-game match in Bahrain. In 2026, the desktop edition of Fritz drew Kasparov 3-3 in a six-game match in New York. By 2026, Deep Fritz beat Kramnik 4-2 in Bonn. Those milestones placed FRITZ in the role people usually assign it: the challenger, the opponent, the machine that measures human limits. But there is a detail rarely mentioned in the news coverage of those years. Behind every major match, the same software shipped to the market as an analysis tool for amateur players. The same product, the same source code, played both the role of the machine that beat a world champion and the role of the companion sitting beside the board of a fourteen-year-old in the suburbs. That duality defines the whole industry. After Deep Blue beat Kasparov in 2026, the race shifted entirely onto the axis of strength. Stockfish appeared, became the strongest open-source engine, and still holds the lead in raw evaluation terms. Leela Chess Zero came later, carrying a neural-network approach that learns through self-play rather than through a human-written evaluation function. With those two names, the strength race is effectively settled at a level no human player can perceive on a real board. That is precisely why FRITZ 20's positioning becomes interesting. When you can no longer win on engine Elo, you must win on something else. ChessBase chose to change the axis: from selling strength to selling a learning process. In the product description, the publisher places three keywords side by side: train more efficiently, more intelligently, and more individually. Not a single line mentions beating opponents through raw calculation. For Vietnamese chess, this shift arrives at the right moment. For more than twenty years, the game in this country has risen on a golden generation that includes Nguyen Ngoc Truong Son, once the youngest Grandmaster in Vietnam at fourteen, and Le Quang Liem, the first Vietnamese player to cross the 2700 Elo mark according to figures published by the International Chess Federation. Behind them sits a provincial training system, gifted-and-talented centres, and thousands of school clubs. But behind that again, the device almost every young Vietnamese player uses daily to study is an analysis engine with no pedagogical function whatsoever. They are given a machine that supplies the right answer. They are not given a path to finding it themselves. That is where the FRITZ 20 story escapes the frame of a product announcement. THREE TRAINING AXES AND THE PRICE OF EACH I read a product description the way I read a tactical report: strip out the claim, find the mechanism behind it, then ask where that mechanism breaks. The three axes the publisher names efficiency, intelligence, individualisation sound like marketing slogans. Put side by side, however, they form a fairly tight triangle. The first axis is efficiency. In chess training, the greatest enemy is not a lack of material but an excess of data. A young player can access an opening database of tens of millions of games and open any Grandmaster game from the past fifty years. The problem is that the player does not know which of those tens of millions of games is worth studying for them personally. A four-hour evening can vanish into scrolling through games that leave nothing behind. Old engines cannot solve this, because they only answer once you already know what to ask. If FRITZ 20 genuinely shortens the loop from observation to action, its value sits exactly here. It does not give you more data. It cuts away the data you do not need. The second axis is intelligence. In chess analysis there is a truth anyone who has sat before an engine knows: the numeric evaluation bar does not correspond to how humans make decisions. An engine evaluates a position as a win probability. A human evaluates a position through a feel for structure, through memories of similar games, through the fear of a piece buried in the corner. An intelligent training tool is one that translates the engine's language into the player's language. This is where pure engines have failed for thirty years. The third axis is individualisation. This is the most ambitious axis and also the easiest to overstate. A program can adjust difficulty to the user's level, suggest exercises based on repeated mistakes, keep a log of errors. But adjusting difficulty is not the same as understanding a learner. A real coach sees the child's hand shake in the fourth minute of an important game. A program only sees the wrong move. These three axes are not equals. Efficiency can be measured within weeks of use. Intelligence can be measured after months. Individualisation can only be measured across an entire development phase of a player, meaning five to ten years. Anyone selling you the third axis on a shorter timeline is selling something else. I have spent years documenting how young players absorb analysis tools. Based on my experience following games, a fairly stable rule emerges: the stronger the tool, the wider the gap between understanding and execution. Years ago, when engines were still weaker than humans, players had to evaluate options themselves and weigh risk themselves. Today they receive a number ready-made. Learning has been inverted: instead of searching for causes, they start from conclusions. That is not FRITZ 20's fault. But it is the environment FRITZ 20 must exist within. THE BLIND SPOT: WHEN THE STUDENT KNOWS THE ENGINE BETTER THAN THE GAME This is the part I want to spend the most time on, because it bears directly on how Vietnamese chess is training its next generation. My first article on engine-based training was heavily criticised. At the time I argued that introducing engines into the basic learning phase would produce a generation with good memory but weak self-assessment. The community pushed back hard. I spent six weeks reviewing every note, checking every game, and concluded I was half right. The biggest lesson was not about being right or wrong. It was that data never takes offence. If you defend your position with feeling, you will always win in your head and lose on the board. The blind spot of engine training has three layers. The first layer is the illusion of understanding. When a young player watches an engine point out the right move, they feel they have grasped the problem. That feeling is real, because the brain registers a new connection. But the connection is between position and move, not between position and reason. In a real game, no blue line draws options for you. You must rebuild the reason from scratch. If the reason is not stored in your head, the position will drift past your eyes as a meaningless frame. The second layer is the mismatch between opening repertoire and calculation stamina. A fifteen-year-old can memorise the first thirty moves of many complex opening lines, because engines and databases make that cheap in time. But when they reach a middlegame with no memorised move available, they collapse within ten moves. This phenomenon repeats across many junior cohorts. They are strong at the easy-to-copy part and weak at the part that requires original thought. The third layer is the loss of tolerance for uncertainty. An engine always has an answer, and always a clear one. On a real board you spend most of your time in a zone where no option holds absolute superiority. You must choose between two bad moves because the good one does not exist. That skill is a player's survival skill, and it is only forged when you are forced to choose without a prompt. There is a strange parallel between this problem and the youth football academies I once observed. When short-term performance pressure rises, young coaches shift toward physically strong players who win junior tournaments easily. Ten years later that cohort produces nobody at the top level, because the technical foundation was skipped while it could still be built. Chess has an identical version: when medal pressure at under-10 and under-12 level rises, children are taught by stuffing openings and pattern tactics, instead of building calculation and positional assessment. They win junior events. Then, at twenty, they stop. That is why I rate highly the decision of a product like FRITZ 20 to pivot toward training rather than analysis. But I also do not believe a tool can fix a systemic problem. A tool only amplifies the existing habits of its user. If the user is in the habit of tracing causes, a good tool helps them move faster. If the user is in the habit of copying answers, a good tool helps them copy faster and more. People look at the pieces. I look at the whole structure shifting. A training engine only has value when it places the learner inside their own structure, not the machine's. WHAT NEEDS TO BE VERIFIED, AND HOW In my analytical work I always write to a fixed structure: state the hypothesis, describe how to verify it, then conclude. This makes articles longer and sometimes drier, but it protects the writer from the most dangerous thing in this profession: confidence without foundation. Applying that structure to FRITZ 20, three hypotheses need independent verification. The first: individualised training tools shorten the time needed to reach a given Elo level compared with traditional methods. Honest verification means splitting two groups of players at the same level with the same time budget, one using a structured tool, one using ordinary analysis engines, then measuring Elo progress after six and twelve months. Until such data exists, every claim about speed of improvement is speculation. The second: the tool improves decision quality in the endgame. This is the easiest to verify, because endgames have objective standards. Take a set of a player's games, count how often they err in favourable endgames, then compare before and after a training period. The third, and the one I care about most: exposure to a training tool reduces rather than increases dependence on engines during play. It sounds paradoxical, and it genuinely is. If a program teaches you to find reasons yourself, you will need it less. If it teaches you to wait for answers, you will need it more. Two opposite outcomes from the same interface. Here I want to mention a technical detail usually ignored in chess tool reviews. Engine strength is not the most important variable for a learner. The most important variable is the quality of the question the tool puts to the learner after each error. A good question forces you to review the chain of decisions that preceded it. A poor question points out the right move and stops there. Fifteen years ago I began drawing diagrams for every analysis of mine and kept the habit of clearly annotating data for each claim. The reason is not aesthetics. When you are forced to cite a source for every assertion, you automatically discard most of the sentences you cannot prove. Chess works the same way. A student required to record the reason for every move progresses more slowly in the first three months and far faster from the twelfth month onward. If FRITZ 20 achieves that, it will have done something three decades of chess engines did not. WHEN THE VENDOR CHANGES AXIS, THE MARKET WILL TEST IT There is one thing I want to put on the table plainly, even if it is not pleasant for chess software publishers. For most of this industry's history, engine strength was the commodity. You sold a program stronger than the previous version and users upgraded. That model died the moment Stockfish crossed the threshold where no human player can perceive a difference. When every engine is strong enough to point out the right move, competitive advantage no longer lies in pointing out the right move. It lies in making the user understand why the right move is right. FRITZ 20's move toward a training identity is a correct strategic decision, and also a forced one. But it carries a risk I see in many educational products across sport generally: when you sell a process, you become accountable for the process's results. Selling strength is unverifiable, because nobody can play an engine at its limit. Selling training is different. Users will ask after six months: how much stronger am I? That is why I believe the real test of this product will happen in youth clubs and talent centres, places where a good or bad process reveals itself after a few cohorts. For Vietnamese chess, this is an opportunity worth taking. Our training system is strong in mobilisation and organisation, strong in early talent identification. The historical weakness lies in the transition from promising junior to internationally consistent performer. That is exactly the phase where the quality of the learning process matters most, and where a well-structured training tool can contribute most. WHAT I GOT WRONG Following a habit established after a serious misjudgement at a World Cup, I always add a self-audit at the end of a deep analysis. In this article, what I may have wrong is the weight I place on the third hypothesis, that training tools reduce dependence. There is another possibility I have not ruled out: every tool, whatever philosophy it was designed around, ends up reshaped by the user's existing habits. If that is true, the difference between a pure analysis engine and a training tool will be far smaller than the marketing claim, and the decisive variable remains the person sitting beside the student, not the software. I may also be wrong to underweight enjoyment. Across many years in this profession I have tended to treat discipline as a more important variable than motivation. But for a fourteen-year-old sitting before a screen at nearly three in the morning, what keeps him there is not discipline. It is the feeling of unanswered curiosity. A tool that knows how to nurture that feeling may have more impact than any Elo progress chart. I am sixty-nine. I still learn from the young every day. Chess has no concept of retirement, and no concept of a final tool. CLOSING: WHAT WILL BE TESTED ON A REAL BOARD Back to the apartment in Hanoi, where the fourteen-year-old is still sitting before the screen at nearly three in the morning. He has just seen the move he missed for two hours. The next question he asks himself will decide everything. If he asks the machine what the right move is, he will gain one more line of memory and forget it in two weeks. If he asks why he never thought of it, he will gain something usable for ten years. No engine, not even the strongest version, can answer the second question for him. But a properly designed tool can force him to ask it. A system does not lie, but it can only be heard when the data is thick enough. With FRITZ 20, the data is not yet thick. The product has just launched, and every assessment of it right now is an assessment of a promise, not of a result. What will be verified over the next two to three years does not lie in technical specifications. It lies in whether a generation of young players enters international events with better self-assessment than the previous one, or with a thicker opening memory and a emptier middlegame. For Vietnamese chess, that question is not academic. We have proven our ability to produce outstanding individuals. What remains unproven is our ability to produce an even generation, where the tenth player understands the game at the level the first player understands it. No engine can do that. A good, repeatable training process can. If in three years young Vietnamese players enter continental events with better endgame calculation than today, we will have our answer. If not, we will have one more strong product sitting idle on a hard drive, and one more generation taught to wait for answers. Patience is not stillness. Patience is waiting for the exact moment the opponent is forced to decide for themselves instead of copying.