Fritz 20: The Chess Training Engine and the Amateur Player's Problem

Three in the morning in Saint Petersburg, snow sealing the window frame, and the only sound left in the study is the fan of an old computer. On the screen, a blue line blinks: Fritz suggests a move I had overlooked for twenty minutes. I have sat before lines like that for more than forty years, from the day the first computer was placed in the corner of a chess club in Hanoi to late editing nights in Russia. Tonight there is one small difference I cannot ignore: the machine no longer waits for me to type a command. It speaks first. Fritz 20 is being introduced as a training revolution. It is called a personal coach, the toughest opponent, the strongest ally, built for both those taking their first steps into serious chess and players already competing at tournament level. The promise fits into three adjectives: more efficient, more intelligent, more individual. For someone who has spent most of a career standing at the edge of the board taking notes, that promise deserves to be tested with data. Chess is one of the rare sports in which the machine overtook the human long ago, and the only one in which that machine simultaneously plays three roles: teacher, referee and library. Fritz is not a new name. The engine dates to the early 1990s, tied to two names, Frans Morsch and Mathias Feist, and released through Germany's ChessBase ecosystem. ChessBase began in the 1980s with a game database, and that database changed how players prepared long before Fritz became a familiar word. The engine's biggest landmarks sit in the Man versus Machine matches. In 2026, in Bahrain, Deep Fritz drew 3-3 with Vladimir Kramnik. In 2026, in New York, X3D Fritz drew 2-2 with Garry Kasparov. In 2026, in Bonn, Deep Fritz beat Kramnik 4-2, and the sixth game ended with a blunder the chess world still discusses nearly two decades later. That was the moment chess changed axis: from asking whether a human could beat a machine to asking how a human must learn from one. Thirty years after Bonn, the chess software market has become a genuine marketplace. The transfer market is where people sell the future to buy hope. Here too: players pay for a future of better calculation, and receive a library of moves they may not truly understand. Technically, Fritz 20 inherits what earlier generations built: an adaptive opponent mode, handicap settings for newcomers, opening training, interactive analysis, game database integration, and modules that explain strategic ideas. The most emphasised feature is individualisation: the engine adjusts difficulty to the trainee's level, plays human-like moves to create realistic situations, then points out the error. Fritz was once famous for its handicap mode, which let beginners play against a machine willing to lower itself. That mode sounds small, but it carries an important psychological lesson: an opponent who always wins teaches nothing, while an opponent of equal strength teaches a great deal. Fritz 20 expands that idea into an entire exercise system, where difficulty is adjusted continuously according to the trainee's results. What stands out is that Fritz 20 does not compete on raw strength. In the free-engine segment, Stockfish and Leela Chess Zero have reached a level that forces every commercial product down a different road. That road is experience: interface, structured courses, thematically ordered exercises, and the ability to turn a just-played game into a lesson with feedback. When the stadium holds no ball, memory starts scoring. The same happens on the chessboard. When no real opponent sits across the table, the trainee begins a dialogue with his own memory, and the machine becomes the only companion patient enough to hear it out. A modern training engine does three jobs. It is an opponent: it plays at exactly the threshold that forces you to try. It is a judge: it scores every move in silence. It is an archivist: it records every game you played and every mistake you repeated. Thirty years ago those three roles had to be split among three different people: a stronger player, a coach, and a notebook. But there is a gap no software can close. Today's leading engines are rated above 3,500 Elo, while the human record is 2,882 Elo, set by Magnus Carlsen in May 2026 according to the World Chess Federation. That gap of more than 600 points is not a gap in calculating speed. It is a gap in how the problem is framed. One thing must be said clearly about the nature of that superiority. The machine is stronger at calculating variations, but it is not stronger at choosing what deserves to be calculated. Humans are limited by time and stamina, so they are forced into intuition. The machine has no such limit, so it needs no intuition. This difference explains why a trainee can learn a great deal from an engine while keeping every one of his old weaknesses intact. A typical week for an ambitious amateur contains four activities: playing a few games, reviewing a few games, learning a new opening line, and solving tactical exercises. The machine inserts itself into all four at once. It is the opponent in the first, the judge in the second, the textbook in the third, and the examiner in the fourth. That convenience is both the strength and the weakness, depending on whether the trainee answers by himself before switching the machine on. In chess, the patch is called the opening book. Whenever a new engine is released, a set of variations long thought dead comes back to life, and a set currently in fashion is suddenly declared wrong. Professional players do not choose openings by taste. They choose by the engine's evaluation, following the same logic that esports teams use when picking champions according to the patch. That invisible referee has the power to decide championships, and nobody sees it on the scoreboard. Adaptability to whatever is new gets mistaken for real strength. A player who memorises the first thirty moves of a new line is praised for careful preparation. But ask why move thirty-one looks the way it does, and the answer is usually silence. The machine has already moved a step ahead, and the player only manages to keep up. Based on my experience following matches, from open tournaments in Hanoi in the early 2000s to events in Europe later on, I have seen a fairly durable rule. The players who improve fastest are not the ones who use engines the most. They are the ones who split their training into shifts: one shift analysing alone first, one shift switching the machine on to check afterwards. That order matters more than the total hours. In 2026, while both playing and organising tournaments, I stood between two groups of people. One group brought a computer to every game and analysed right at the board. The other brought only a notebook and a pencil. Ten years later, the second group talked about chess far better. They were not stronger in rating, but they understood why a move was right. Software today has become intelligent enough to explain. Alongside the evaluation, it returns a sentence: this move is better because it controls the centre, because it opens a file for the rook, because it pushes the queen into passivity. That is the genuine advance of a new engine generation. The problem lies elsewhere: whether the trainee reads that sentence before looking at the evaluation. Data on training effectiveness in chess is hard to collect, because samples are small and variables are many. But one observation repeats across many clubs: trainees who keep their own notebook improve faster than those who rely only on the software's library. A notebook forces summarising, and summarising forces understanding. The generation of players born after 2026 grew up with an engine from their very first move. They do not know the feeling of waiting three days to borrow an opening book, or of copying a game by hand from a magazine. That emptiness used to be where imagination worked. Now the emptiness is filled with a mouse click, and imagination must find somewhere else to live. At this point I have to say what many in the profession avoid. The engine does not teach you to calculate. It teaches you to recognise. Those two skills differ in nature. Recognition helps you find the right move in ten seconds, but only calculation keeps you standing at move forty, with three minutes on the clock and no familiar pattern left to hold on to. Collective memory of the sixth game in Bonn in 2026 is usually told as a story about human decline. I think that telling misses half the event. Kramnik that year was the most thoroughly computer-prepared player on the planet. He lost in a moment no engine can reach: the moment a human becomes overconfident in his own preparation. I once heard a coach in Saint Petersburg say that an engine is like a mirror that reflects very fast but cannot speak. You look into it and see your wrong move in an instant. But the mirror will not tell you the story of why you played that way, and that story is what actually needs fixing. Esports taught me that youth is not in the muscles, but in the reflexes of the heart. Chess taught me the same with the seat swapped: youth is in the reflexes of the mind before a position it has never seen. A machine can simulate millions of unseen positions, but it cannot pass on the feeling of confusion when you must choose between two moves that both look right. So I read the message of Fritz 20 with some caution and some hope. Caution, because every training revolution sells the buyer a feeling of progressing faster than he really is. Hope, because a patient machine that does not judge and plays at exactly your level is something most amateur chess players have never had in their lives. A Luzhniki night, an old bird understands that landing is also a way of flying. Chess players must learn the same lesson about their machine. Knowing when to switch it off, sit down with an empty board, and find the move alone is the last skill no software can install for you. And if Fritz 20 really is a revolution, that revolution will not be measured by the number of games analysed, but by the number of games players dare to analyse alone.