A good logic puzzle is never about knowing something — every fact you need sits in the puzzle itself. It is about the discipline of squeezing those facts: assuming and breaking, eliminating fast, treating even someone's silence as information.
The eleven puzzles below are the classics of four families: truth-tellers and liars, elimination and ordering, river crossings and measuring, and hat puzzles built on silent inference. Each family opens with the one method that cracks it, and every puzzle ends with a full worked solution — including why the tempting wrong answers are wrong.
Try each one before opening the answer. The point is not being right; it is catching the exact moment your reasoning wants to take an illegal shortcut.
Four families, eleven puzzles
Truth-tellers and liars
3 puzzlesEvery puzzle in this family runs on one engine: assume, then check for contradiction. Pick any speaker, suppose they are truthful, and follow the consequences; if you hit a contradiction, they are a liar, and everything they said flips to false. The advanced move is noticing that a liar's answer is still information — a stable error can be reversed, which is exactly what the famous two-door puzzle exploits.
How they're usually asked: "Knights and knaves" · "One guard always lies" · "Who is telling the truth?"
Puzzle 1
On an island where everyone either always lies or always tells the truth, A announces: "We are both liars." What are A and B?
- AA is a liar, B tells the truth
- BBoth are liars
- CA tells the truth, B is a liar
- DCannot be determined
▶Show solution
Answer: A. A is a liar, B tells the truth
🐱 Suppose A tells the truth. Then "we are both liars" is true, which makes A a liar — contradiction. So A is a liar, and the statement is false: they are NOT both liars. Since A is one liar already, the falseness must come from B — B tells the truth. Note what happened: one sentence pinned down two people, because a self-referential lie is still a constraint.
Puzzle 2
Three islanders. A says: "B is a liar." B says: "C is a liar." C says: "A and B are both liars." Who tells the truth?
- AOnly B
- BOnly A
- COnly C
- DA and B
▶Show solution
Answer: A. Only B
🐱 Test C first, because C's claim is the biggest and easiest to break. If C were truthful, A and B would both be liars — but then A's "B is a liar" would be false, making B truthful. Contradiction, so C lies. Then B's "C is a liar" is true — B is truthful. Then A's "B is a liar" is false — A lies. Final check: C claimed both A and B lie; only A does, so C's statement is indeed false. Everything closes. Starting from the loudest claim is a general trick: big statements break fastest.
Puzzle 3
Two doors: one leads out, one does not. Two guards: one always lies, one always tells the truth — you don't know which is which. You may ask one guard a single question. What do you ask?
- A"If I asked the other guard which door leads out, which would he point to?" — then take the other door
- B"Are you the truthful guard?"
- C"Which door leads out?" — asked twice quickly
- D"Does the left door lead out?"
▶Show solution
Answer: A. "If I asked the other guard which door leads out, which would he point to?" — then take the other door
🐱 Route your question through both guards at once. If you ask the truth-teller, he honestly reports the liar's wrong answer. If you ask the liar, he lies about the truth-teller's right answer. Either way the door you are shown is the wrong one — so you take the other. The failed options share one flaw: their answer's meaning depends on who you happened to ask. The winning question is engineered so that both possible speakers produce the same, decodable output. Stable error beats unknown truth.
Elimination and ordering
3 puzzlesThese are the puzzles people mean by "logic grid puzzles": a few categories, a few clues, one consistent assignment. The working method is a table — write candidates, cross out what each clue forbids, and watch forced moves appear. Two habits matter: extract the negative content of every clue ("the red house is next to the blue one" also means red is NOT in certain positions), and re-scan earlier clues every time something new gets fixed, because old clues acquire new teeth.
How they're usually asked: "Who drinks what?" · "Which house is where?" · mini grid puzzles
Puzzle 4
Ann, Ben and Cara each drink exactly one of coffee, tea, juice. Neither Ann nor Cara drinks tea. Cara never touches coffee. Who drinks what?
- AAnn coffee, Ben tea, Cara juice
- BAnn juice, Ben tea, Cara coffee
- CAnn tea, Ben coffee, Cara juice
- DAnn coffee, Ben juice, Cara tea
▶Show solution
Answer: A. Ann coffee, Ben tea, Cara juice
🐱 Clue 1 removes two of the three tea candidates at once — tea must be Ben's. Clue 2 removes coffee from Cara, and Ben is taken, so coffee is Ann's; juice falls to Cara by elimination. The lesson in miniature: a clue that names two people ("neither Ann nor Cara") is stronger than a clue that names one, because elimination puzzles are won by whoever deletes fastest.
Puzzle 5
Four houses in a row (positions 1-4, left to right), painted red, blue, green, yellow in some order. The red house is immediately left of the blue one. The green house is not at either end. The yellow house is somewhere left of the red one. What is the order?
- AYellow, green, red, blue
- BGreen, yellow, red, blue
- CYellow, red, blue, green
- DRed, blue, green, yellow
▶Show solution
Answer: A. Yellow, green, red, blue
🐱 Red-blue move as a glued pair: possible slots (1,2), (2,3), (3,4). Yellow must sit left of red, which kills (1,2). Try (2,3): green must take position 2 or 3 — both occupied, dead end. So red-blue take (3,4), green must be 2, yellow gets 1. One pass, no guessing — notice that the "not at either end" clue did nothing until the pair was placed, then instantly finished the puzzle. Clues fire in the order the puzzle decides, not the order they were given.
Puzzle 6
Someone says, truthfully: "The day before yesterday I was 25. Next year I will turn 28." How is that possible?
- AThey spoke on January 1, and their birthday is December 31
- BThey spoke on February 29 in a leap year
- CIt is impossible; the statement contradicts itself
- DThey spoke on their own birthday
▶Show solution
Answer: A. They spoke on January 1, and their birthday is December 31
🐱 Put the statement on January 1 with a December 31 birthday. The day before yesterday — December 30 — they were still 25. They turned 26 on December 31. This calendar year they will turn 27, and next year 28. Four ages appear across a two-day window without any contradiction. The trap is reading "next year" from today instead of from the calendar — date puzzles are almost always about which reference frame each phrase silently uses.
River crossings and measuring
3 puzzlesProcess puzzles ask for a sequence of moves under constraints. The universal unlock is the same in all of them: the move that feels wasteful — bringing something back, dumping water you just poured, weighing balls you could have "saved" — is usually the move the puzzle exists to teach. If your plan never goes backwards and you are stuck, the backward move is what you are missing.
How they're usually asked: "Wolf, goat and cabbage" · "Measure 4 liters with a 3 and a 5" · "Find the odd ball"
Puzzle 7
A farmer must ferry a wolf, a goat and a cabbage across a river. The boat holds the farmer plus one item. Left alone together, the wolf eats the goat and the goat eats the cabbage. Minimum trips to get everything across safely?
- A7 crossings — the key is bringing the goat BACK on trip 3
- B5 crossings
- CIt cannot be done
- D9 crossings
▶Show solution
Answer: A. 7 crossings — the key is bringing the goat BACK on trip 3
🐱 Take the goat over (1), return empty (2), take the wolf over (3) — and here is the move: bring the goat back (4). Now take the cabbage over (5), return empty (6), take the goat over again (7). Every dangerous pair is only ever alone when the farmer is present. The puzzle is a monument to one idea: progress sometimes requires undoing progress, and the goat crossing three times is not waste — it is the solution.
Puzzle 8
You have an unmarked 3-liter jug, an unmarked 5-liter jug, and unlimited water. Measure exactly 4 liters.
- AFill 5, pour into 3 (leaves 2), empty 3, pour the 2 in, refill 5, top up 3 — the 5-jug now holds 4
- BFill both jugs and estimate half of the total
- CImpossible without a third container
- DFill 3 twice into 5 and the overflow is 4
▶Show solution
Answer: A. Fill 5, pour into 3 (leaves 2), empty 3, pour the 2 in, refill 5, top up 3 — the 5-jug now holds 4
🐱 Fill the 5 and pour into the 3: the 5 now holds exactly 2. Empty the 3, transfer the 2 into it — the 3-jug now has 2, meaning it can accept exactly 1 more. Refill the 5 and top up the 3: exactly 1 liter leaves, and 5 − 1 = 4 remains. Every number here was created by a difference between containers, never by eyeballing. That is the whole genre: unmarked jugs can still compute, because pouring-until-full is exact arithmetic.
Puzzle 9
Eight balls look identical, but one is slightly lighter. Using a balance scale only twice, how do you find it?
- AWeigh 3 vs 3 first; the answer splits eight balls into groups of 3, 3 and 2
- BWeigh 4 vs 4, then 2 vs 2 — three weighings needed
- CIt cannot be done in two weighings
- DWeigh 2 vs 2 repeatedly until it shows
▶Show solution
Answer: A. Weigh 3 vs 3 first; the answer splits eight balls into groups of 3, 3 and 2
🐱 The instinct is 4 vs 4, but that first weighing only halves the field — leaving 4 candidates, too many for one more weighing. Weigh 3 vs 3 instead. If one side rises, the light ball is among those 3: weigh 1 vs 1 of them, and either one rises or it is the third. If the scale balances, the light ball is in the leftover 2: one weighing settles it. A balance has three outcomes — left, right, equal — so each weighing can cut candidates into three, not two. Eight ≤ 3², hence two weighings suffice.
Hats and silent inference
2 puzzlesThe deepest puzzle family: what others do NOT say is data. Each player reasons about what another player would have concluded if the situation were different — and when that conclusion doesn't come, a possibility dies. Solving these means simulating someone else's reasoning inside your own, one level deep, sometimes two. Slow down and narrate each person's viewpoint in turn; the puzzles collapse once every silence has been translated into a sentence.
How they're usually asked: "Three prisoners, five hats" · "Why does his silence tell you the answer?"
Puzzle 10
Three prisoners stand in a line. C sees A and B; B sees only A; A sees no one. From a bag of 3 white and 2 black hats, one is placed on each head. C is asked his color: "I don't know." B is asked: "I don't know." A, who sees nothing, then says: "I know my color." What is it, and how?
- AWhite — C's and B's ignorance each eliminated the worlds where A wears black
- BBlack — two black hats were used on B and C
- CA guessed with 3/5 probability
- DIt cannot be deduced from silence alone
▶Show solution
Answer: A. White — C's and B's ignorance each eliminated the worlds where A wears black
🐱 C would know his color only if he saw two black hats (then his must be white, as only 2 blacks exist). He doesn't know — so A and B are not both black. Now B knows this. If B saw A in black, B could conclude "then I am not also black — I'm white." But B doesn't know either — so A is not black. A hears both admissions and, seeing nothing at all, concludes: white. Two people's ignorance, publicly declared, carried all the information A needed. "I don't know" is never a null statement in these puzzles — it is a report about what the speaker sees.
Puzzle 11
Two players each have an integer from 1 upward on their forehead, visible to the other but not to themselves. They are told the two numbers are consecutive. They alternate saying "I don't know my number" — until suddenly one of them announces it. Why does this process ever terminate?
- AEach "I don't know" eliminates the smallest remaining possibility, so knowledge climbs one rung per round
- BIt doesn't — the puzzle is a paradox
- COne player eventually guesses at random
- DThey secretly communicate through timing
▶Show solution
Answer: A. Each "I don't know" eliminates the smallest remaining possibility, so knowledge climbs one rung per round
🐱 If a player saw the number 1, they would instantly know theirs is 2 (numbers are consecutive and start at 1). So a first "I don't know" announces: I do not see a 1. Now someone who sees a 2 knows their own must be 3 — if it were 1, the other player would have known already. Each round of shared ignorance raises the floor by one, and whoever's number is smaller eventually gets caught by the rising floor. This is the cleanest example of common knowledge doing physical work: nobody learned anything private; they only learned that the other didn't know — and that was enough.