Someone died overnight in a house full of people. You get the household’s statements, a small cast of suspects, a handful of implements, and a set of rooms. Put every person and every piece of evidence back where the statements say they must have been, and the room holding both the killer and the weapon marks itself. You never accuse anyone — you reconstruct the evening, and the accusation is what’s left over.
Written by Shawn Liu, founder & puzzle designer.
Don’t hunt for the killer. Hunting for the killer is the slowest route through an Alibi case, because the killer is not a thing the clues talk about — they talk about where people and objects were. Solve the arrangement and the accusation arrives on its own.
Those are the openers. The full method — including the one move the solver makes on every clue that almost nobody makes by hand — is on the strategy page, worked through a real case. The five clue types, the exact wording conventions, and a case walked end to end are on the how-to-play page. The interactive tutorial runs the same ground on a real, small case.
Alibi cases are built backwards. The generator picks the answer first — who was where, which implement was where — and only then works out what could truthfully be said about it. It enumerates every statement that is true of that arrangement, which is typically hundreds, and starts selecting from that pool, keeping the clues that collapse the possibilities fastest, until exactly one arrangement survives.
Then it does the part that matters most to you: it walks back through the clues it just chose and throws away every one whose removal would still leave the case uniquely solvable. What’s left is a file with no padding in it. If a statement is on the rail, you need it — there is no decorative clue in an Alibi case, and nothing is there to waste your time. A case that somehow comes out ambiguous anyway is discarded and rebuilt from a new seed rather than shipped.
One more deliberate omission: the daily cases never include a statement that simply places someone in a room. That kind of clue is allowed by the engine and generated into the pool, but the daily difficulty bands filter every one of them out before a case ships. Nothing in the file hands you an answer — every placement you make is one you derived. It’s the same commitment behind the other three games, written up in no-guessing puzzles.
A new case lands every day at midnight Eastern. Behind the week sits a ten-band difficulty ladder, in the tradition of the Inspector Parker casebooks; the daily rotation walks Monday to Sunday across seven of those bands, skipping the ones that would put two similar cases back to back. Monday is open-and-shut. Sunday is not. The remaining bands are held back for special days and for a harder mode with a bigger grid — the daily cases stay inside a three-by-four board on purpose, because the pleasure here is depth of reasoning rather than bookkeeping.
The last seven days stay free in the archive. Older cases, back to launch in May 2026, come with a subscription. If you want the same kind of deduction on a grid instead of in a house, Griductive is next door.