How a chart is drawn, and proved fair
3 November 2026
Every day the game needs an island, a lie, and a handful of notes that point to it. The island is generated: around forty-eight hexes, each printed as one of six grounds, with landmarks and readings placed on it and the chart divided into sheets. One hex is then drawn as something it is not. That hex is the lie, and the notes are written about it.
The notes come from a small, fixed grammar. A note can give a bearing from every landmark of a kind, a count of neighbouring hexes of one ground, a sheet, a distance from a landmark, a shortest path between two landmarks, the ground the lie is printed on, or the fact that nothing next to it shares its ground. Each day’s chart carries between four and six of them.
Before anything is published, a fairness gate checks two things. The first is uniqueness: exactly one hex on the island satisfies every note, and it is the lie. The second is deducibility: that hex can be reached by pure elimination, striking out hexes one note at a time, without ever having to guess and see what follows.
The second property comes free, and it is the central design decision. Every note in the grammar is a statement about one hex and the map, never about two unknowns at once. So any hex that is not the answer fails at least one note on its own, and that single note strikes it out directly. There is never a step of the form “suppose the lie is here, and see if that leads to a contradiction”. The solver emits a certificate for each chart listing, for every non-answer hex, the note that rules it out. Sudoku cannot promise this; a unique Sudoku can still require trial and error. This grammar sidesteps the whole problem.
Two further rules keep a chart honest rather than merely valid. A note that eliminates nothing on its own is banned, because that is noise rather than misdirection. And at most one note may be redundant, implied by the others. That one allowance is deliberate: it is the red herring, and it is the setter’s legitimate weapon.
Which brings us to setters. Most charts are drawn by the generator, and the line under the tier tabs says so: drawn by the machine. Some are set by players in the apps, and the best of them go into the daily rotation after other players have rated them. A human setter may mislead you with a note that sounds more useful than it is, or with a red herring. What a setter cannot do is hand you a chart with two answers, or with none, or one that needs a lucky guess, because the gate refuses to publish it. The setter can deceive. The setter cannot cheat.
The same generator and the same solver run on the website, on iPhone and on Android, so a chart is identical wherever it is played and the proof that it is fair was run before you ever saw it.