Chess Trainer App Store

Ratings

What a rating actually measures

Four numbers claim to describe how good you are at chess, and none of them converts into any of the others. This page says what each one is measuring, why the puzzle number is always the flattering one, and what Chess Trainer’s own ratings do and do not claim.

Last reviewed 19 August 2026

A rating is not a measurement of you. It is a measurement of you, against a particular pool, under a particular set of conditions — and change any of those and the number changes with it.

Every rating system in chess works the same way underneath: it predicts the result of your next game from the difference between two numbers, then moves both numbers by however wrong it was. What differs between systems is not the arithmetic — it is who is in the pool, how fast the games are, and how much evidence the number is based on.

That is why the same player is 1500 in one place and 1900 in another without having got better or worse in between. Both numbers are correct. They are answers to different questions.

The one that flatters

Puzzle ratings run high, and it is not an accident.

A puzzle tells you there is something there

The single most valuable piece of information in chess is “this position contains a win”, and a puzzle gives it away before you have looked at anything. In a game nobody tells you. Most of the difficulty of playing well is deciding whether this is a position that rewards two minutes of calculation, and a puzzle has already answered that.

And that it is findable now

A puzzle has a solution that works today, in this position, with the pieces where they stand. A game asks a harder question — what improves my position — which has no verifiable answer and no immediate feedback.

So the gap is structural

Expect a puzzle rating several hundred points above your rating in games, on any platform, at any strength. That is not you being bad at converting; it is the two numbers measuring different things. Neither is the fake one.

The table this page does not have

There is no conversion between Lichess, chess.com and FIDE.

You will find tables claiming there is. They disagree with each other, they disagree with themselves at different rating bands, and they are built from small samples of players who happen to have accounts in two places — which is not a random sample of anybody.

What can honestly be said is directional: online ratings generally read higher than FIDE ratings, and Lichess generally reads higher than chess.com. The size of the gap varies by hundreds of points depending on where you are on the scale and what time control you play, so turning that direction into a number produces a wrong answer with a confident face. If you want to know your FIDE rating, the only way is to play FIDE-rated games.

The bundled library

What each band of the puzzle set is asking of you.

The 14,351 puzzles run from 760 to 2800. Most of them carry ratings calibrated against millions of real solving attempts, so the difficulty ordering is earned rather than estimated.

  1. 760 – 1200

    The pieces stop hanging

    One-move tactics, and the habit of checking what is loose.

  2. 1200 – 1500

    Patterns, not calculation

    Forks, pins and back-rank threats seen rather than worked out.

  3. 1500 – 1800

    Two moves deep, reliably

    Deflections, discovered attacks, and the first quiet moves.

  4. 1800 – 2100

    The point is further away

    Long forcing lines, zwischenzugs, and sacrifices you can verify.

  5. 2100 – 2500

    Quiet moves and judgement

    Positions where nothing forcing works and something still wins.

  6. 2500 – 2800

    The top of the library

    Points that lie deeper than most engines are searched at.

Asked often

Four questions with honest answers.

Why is my puzzle rating so much higher than my game rating?
Because a puzzle tells you something a game does not: that there is something to find, and that it is findable right now. That single piece of information is worth several hundred rating points. A puzzle rating measures how hard a position you can solve when you already know it is solvable; a game rating measures whether you notice in the first place.
How do Lichess, chess.com and FIDE ratings compare?
They do not convert, and any table that says otherwise is folklore. Each is a self-contained scale calibrated against its own pool of players, and the pools differ in strength, in time control and in how many games a rating is based on. As a direction rather than a formula: online ratings generally read higher than FIDE, and Lichess generally reads higher than chess.com — but the gap varies by hundreds of points depending on rating band and format, so treating it as a conversion produces a wrong number with a confident face.
What is a good chess rating?
Better than yours was last year. Rating scales are relative to a pool, so a number is only meaningful next to the same number measured the same way at a different time. The useful comparison is always with yourself.
Does Chess Trainer’s rating mean anything outside the app?
No, and it does not pretend to. It is calibrated against the difficulty of the bundled library, which carries Lichess’ human-calibrated puzzle ratings for most of its 14,351 positions. It measures your progress against the library, not your strength against a field of humans on a clock.

Next: the twenty tactical motifs →

Chess Trainer

Tactics, positional judgement, endgame technique and coached play, with Stockfish running on the device. Nothing leaves the phone.

Free software under the GNU General Public License v3. Read the source.

© 2026 Mincho Milev. Chess Trainer is free software.

Stockfish © the Stockfish developers, GPLv3. Puzzles and games from the Lichess database, CC0. Not affiliated with Apple, Lichess or the Stockfish project.