The warden knows who's pardoned. He won't say. But he'll say something.
Three prisoners, A, B and C, are told one of them has been picked at random for pardon. Prisoner A asks the warden — who knows the answer but can't reveal it — to name one of the other two who will definitely be executed. The warden names C. Has A's own chance of pardon improved from 1 in 3?
Reveal the answer
No — it's still exactly 1 in 3, though it feels like it should rise to 1 in 2. The warden's answer only ever eliminates a prisoner who wasn't A, so it carries no information about A specifically; the freed-up probability shifts entirely onto the unnamed prisoner, B. Martin Gardner published the puzzle in 1959, decades before an equivalent version, the Monty Hall problem, made the same logic famous.
— Martin Gardner, Mathematical Games column — Scientific American, October 1959