How Chess Puzzle Ratings Work: What the Number in Rook Means and Why It Moves

Open Rook and the first thing you see is a number: your puzzle rating. Solve a puzzle and it goes up, sometimes by three points, sometimes by twenty-five. Miss one and it drops. After a few weeks there is a line on the stats screen that climbs, flattens, dips and climbs again. This article explains where that number comes from, why it moves by the amounts it does, and what it can and cannot tell you about your chess.

A system built for players, borrowed for puzzles

The idea behind almost every chess rating is the one Arpad Elo, a physics professor and a strong amateur player, worked out for the United States Chess Federation, which adopted it in 1960; FIDE followed in 1970. Elo’s insight was that a rating does not need to measure skill directly. It only needs to predict results. If two players’ ratings say one should win three games in four, and over time they do, the ratings are doing their job. When results come out differently from the prediction, the ratings are nudged towards the truth.

A puzzle trainer borrows the whole system with one twist: every puzzle is treated as a player. When you face a puzzle, you are playing a one-move game against it. Find the winning line and you win; miss it and the puzzle wins. Your rating and the puzzle’s rating together say how likely you were to solve it, and the result is compared with that expectation.

Why a puzzle has a rating at all

A puzzle’s rating is a measure of how often players of different strengths solve it. A simple mate in one that almost everyone finds sits low on the scale. A quiet defensive move that strong players miss half the time sits high. In most puzzle systems, the puzzle’s rating is earned exactly the way a player’s is: every time someone fails it, it is evidence that it is harder than its rating said, and every time someone solves it, evidence that it is easier.

That is why a rated puzzle is so much more useful than a puzzle labelled “medium”. “Medium” is someone’s opinion. A rating of 1206 is a summary of how real attempts went, on the same scale as your own number — which is what lets the app compare the two.

The prediction: 400 points is a lot

The heart of the system is a formula that turns a rating difference into an expected score. You do not need to remember the formula, only its shape:

The same numbers run the other way. A puzzle rated 200 points above you is one you are expected to solve about a quarter of the time, and one 400 points above you, about one time in eleven.

How much a result moves your rating

After each puzzle, the change in your rating is proportional to the surprise: the difference between what happened and what was expected. Solving a puzzle you were 91% likely to solve is barely news, so it earns you very little. Solving one you were 20% likely to solve is big news, and your rating jumps. Missing works the same way in reverse: failing an easy puzzle costs a lot, failing a hard one costs little.

Here is a worked example with the classic Elo formula. Say your rating is 964 and you are given a puzzle rated 1206 — 242 points above you. The formula says you were expected to solve it about 20% of the time. You solve it. The surprise is 1 minus 0.2, which is 0.8, multiplied by a sensitivity factor — often 32 for new or fast-moving ratings. That comes to roughly +26. Had you missed it, you would have lost only about 6, because missing a puzzle that hard is exactly what the ratings predicted.

Those are the kinds of jumps you see in Rook: big gains for solving above your level, small losses for missing far above it, and the reverse for puzzles well below you. The exact sensitivity varies between systems, but the shape is always the same.

Why the first try is the one that counts

In Rook a puzzle changes your rating only by how your first attempt goes. If your first move is wrong, the puzzle counts as missed, even if you find the right line on the second try; if you ask for the solution, it is a miss too.

That can feel harsh, but it is what keeps the number honest. In a real game you get one move. If second tries counted, the rating would measure how quickly you can eliminate wrong answers, not how well you can see the right one. It is also why the first-try rate on the stats screen is worth watching: it is the plainest summary of how often you calculate the line correctly before you touch a piece.

Choosing puzzles near your level

The Train mode picks puzzles close to your current rating, and the reason is in the numbers above. Puzzles far below you teach nothing and move your rating by a point or two. Puzzles far above you are mostly missed, which is discouraging and also uninformative. Puzzles near your rating are the ones where your success is genuinely uncertain, which is where both learning and accurate measurement happen.

It is also why the line on your rating chart zigzags rather than climbing smoothly. Near your level you should expect to miss a fair share of puzzles. A run of misses is not a sign that you have got worse; it is the system finding the edge of what you can currently see.

Fast early, slow later

New ratings tend to move quickly. The system knows little about you, so each result says a lot, and a few dozen puzzles can carry your number a long way from where it started. As results pile up, the number settles. That is why the first week in a trainer often shows a steep climb that then flattens: the early climb is mostly the rating finding where you already were, and the flatter part afterwards is where real improvement shows up — slowly.

What the number does not mean

A puzzle rating is not your playing strength. Puzzles tell you there is a tactic and that it is your move; a game tells you neither. Many players have puzzle ratings several hundred points away from their game ratings, in either direction, and the numbers from different apps and websites are not on the same scale either, because each system’s puzzles and players are different.

What the number does measure, consistently, is your pattern recognition for tactics against a fixed pool of positions. That is worth improving: most games below master level are decided by a tactic one player saw and the other did not.

Using the rating well

The rating is a thermometer, not the medicine. Read it now and then, train with puzzles that are hard enough to miss, and let the number catch up.

← Back to Blog