Fritz 20: A 3,400-Elo Machine and What It Cannot Teach

In a small chess academy on Sampige Road, Malleshwaram, Bengaluru, a laptop fan whines like a mosquito at two in the morning. A thirteen-year-old boy has just pushed his knight to f5. The evaluation bar on the screen drops by 0.6 of a pawn. He is not looking at the bar. He is looking at the coach sitting across from him, a man who is silent in exactly the way a teacher goes silent when he knows his student has just touched something very real. Three weeks later, the boy told me about that move: “I knew it was wrong. But I didn’t know why I still wanted to play it.” The chess board back home has no grandstand, yet every time a piece is lifted, a whole sky of memory rushes back. That boy’s question — why — is precisely the thing Fritz 20 claims to answer.

The name Fritz has been part of chess since 2026, when Frans Morsch wrote the first version and ChessBase in Hamburg brought it to market. In 2026, Fritz 3 beat Garry Kasparov in a blitz game in Munich, and from that night on, Western chess began to treat software as a genuine opponent. In 2026, Deep Fritz drew 4-4 with Vladimir Kramnik in Bahrain. Four years later, again in Germany, Deep Fritz beat Kramnik 4-2 in Bonn. Those markers shaped a belief that lasted two decades: the machine is stronger than the human, and the player’s job is to learn how to endure that.

India is where that belief has been tested most visibly. Viswanathan Anand opened the door in 2026, D. Gukesh took the world title in 2026, and R. Praggnanandhaa and Arjun Erigaisi pulled an entire new generation into academies in Chennai, Bengaluru and Hyderabad. In those academies, chess software is mandatory equipment. In the village clubs farther out, where I have sat for hours listening to people argue about a game nobody can even name the tournament of, a computer is still an expensive thing. Those two worlds share one game, and they are growing up with two different kinds of memory.

Fritz 20 has placed its bet on the first world. The engine inside it combines a neural network with a classical evaluation method, reaching roughly 3,400 Elo at its strongest setting, which means it trails leaders such as Stockfish or Leela by a few hundred Elo on the machine ranking lists. For an amateur player, that gap has no practical meaning. What matters is everything arranged around the engine: cloud analysis to borrow stronger hardware, a sparring mode that adjusts difficulty to the learner’s own style, an opening trainer built around an individual repertoire, a blunder-check function that annotates games in human language, and a database of millions of games bundled in.

Reading that list, I think of afternoons at a roadside chess club not far from Bengaluru airport, where people play on a wooden board set up on the pavement and keep their moves in their heads rather than on a hard drive. Based on my experience following and reporting on chess for many years, one thing is clear to me: what has made young Indian players improve so fast over the past decade is not a stronger engine, but the ability to ask themselves questions. Software only makes that cheaper and faster.

The most valuable thing about Fritz 20 is that it shifts from the role of the one who defeats you to the role of the teacher: instead of handing you a correct move, it tries to point at the silence between what you think and what you do.

That is why the adaptive sparring mode matters more than any Elo figure. A machine that always plays the strongest move teaches memorisation; a machine that plays a nearly-correct move, that leaves just enough of a gap for the learner to decide alone, teaches something else. It does not push the learner into a position of punishment, but into a position of choice. For the students I have sat beside, that difference is bigger than any engine update.

The blunder-check tool also needs to be read properly. It points out the move that collapsed the position, but it cannot point out the hesitation that stood before that move. A thirteen-year-old watching the evaluation bar drop 0.6 of a pawn will learn that the move was bad. He will not learn why his hand still wanted to lift the knight, why his eye went to an empty square on the kingside while his brain had already calculated a sequence on the queenside. Good software records the move. A good teacher records the moment before it.

The opening trainer is where Fritz 20 looks most practical. Amateur players typically spend the first twenty moves in a fog, following a memory of some game online without understanding why. Being forced to build an individual opening system, test it against real results and revise it after every tournament is a discipline rather than a convenience. In India, where the number of open events has grown so fast that a young player can compete three weeks in a row, that discipline directly decides the standings.

What I always tell my students is that a crowded calendar is itself the biggest cause of decline: no software makes up for playing two tournaments in a month and then analysing with eyes already blurred. Fritz 20 can plan training, but it does not know where the learner’s body is.

Then comes the hardest part, the part I believe is the blind spot of every software launch. A database of millions of games, cloud analysis and a 3,400-Elo engine produce a new kind of player: someone who memorises evaluations without understanding why. I have met teenagers who can point out the best move in a Gukesh game within four seconds, then freeze completely when asked how that move changed the meaning of the whole position. Understanding gets replaced by lookup.

Vladimir Kramnik said years after his defeats in Bonn that machines themselves had made elite chess more closed, with every mistake examined under the same light. I do not entirely agree, but I understand the worry. When every move can be scored, people easily forget that across the board, your opponent is also shaking. A machine does not know how to shake. Humans do, and that shaking is exactly what is worth training.

There is one more paradox. The training revolution Fritz 20 advertises only reaches places that already have laptops, internet and money for a licence. In the village clubs I have visited, people still teach chess with a board redrawn on paper and the memory of the oldest person in the village. The distance between those two worlds does not narrow because of an update.

So does Fritz 20 deserve the word “revolution”? For a player training alone, without a coach, who must find an answer for every one of their moves, it is a real step forward. For someone who already has a teacher, it is a better tool, but still a tool. Among the millions of moves stored in memory, the only thing that cannot be stored is human hesitation.

When the machine has shown nearly every move correctly, what remains to make a player different is the seconds they hesitate over the board, and the reason they hesitate. Who will teach that to the thirteen-year-old in Malleshwaram?