Three Ways In: Designing Puzzles That Bend Instead of Break
Photo: Erasmus Kamugisha, CC BY-SA 4.0, via Wikimedia Commons
You built a door with a riddle on it. It’s a good riddle. It rhymes, it has a clever misdirect, and the answer is “a candle.”
Forty minutes later, four intelligent adults have proposed “a mirror,” “a widow,” “a clock,” “the moon,” and “honestly, is it a candle?” and your session has stopped dead in the water. Somebody is on their phone. The tension you were going for has curdled into the specific silence of people who are no longer playing a game but failing a test.
This is not a player problem. It’s a design problem, and it has a diagnosis.
Why Riddles Break at the Table
A riddle is a puzzle with exactly one answer, no partial credit, and no way to observe the mechanism. That’s a fine design for a crossword. It’s a terrible design for a room full of people who can only interact with the world through your description of it.
Consider what the players actually have to work with. They can’t see the door. They can’t touch the carvings. Their entire sensory experience of the puzzle is whatever you said out loud, once, four minutes ago. Meanwhile, you’re holding the answer in your head and suffering from the oldest problem in game design: once you know the solution, you genuinely cannot remember what it felt like not to know it.
And crucially: there are no dice. Most of the game’s other problems get solved by a character being good at something. A riddle asks the player to be clever at 8:45 p.m. on a weeknight after a full day of work. That’s a big ask, and it’s not what most people showed up for.
A Riddle Is a Lock. A Puzzle Is a Room.
The fix is to stop thinking of puzzles as locks and start thinking of them as rooms — spaces with multiple entrances, where exploring one entrance teaches you something about the others.
A well-built puzzle should have at least three ways in:
The intended solution. The elegant one. The one you designed. It should be achievable, and it should reward the players who were paying attention.
The brute-force solution. Something that costs resources — time, spell slots, hit points, a loud noise that attracts attention. This is the pressure valve. It means the puzzle can never actually stop the game.
The weird solution. Whatever your players invent that you didn’t anticipate. Your job as GM is to say yes to it if it’s even slightly plausible, because the memory they take home is the plan they hatched, not the answer you wrote.
If a puzzle has only one entry, it isn’t a puzzle. It’s a password.
Clues Should Be Visible and Redundant
Here’s the operational rule that solves most puzzle disasters: plant at least three clues for every conclusion you want the players to reach.
Not three clues in total — three clues per inference. If the players need to learn that the door opens at sundown, there should be a fresco showing a sun setting, a journal entry complaining that the masons always knocked off at dusk, and an NPC who offhandedly mentions the door “sleeps with the light.” Three different vectors, three different scenes, three different chances for somebody to notice.
Redundancy feels excessive when you’re writing it. It is not excessive when you’re running it. Players miss things constantly — not because they’re careless, but because they’re juggling six characters, a combat round, and the fact that they haven’t eaten since noon.
And put the clues in the world, not in your head. A physical handout. An index card with the inscription written out. A sketch of the mural. Anything the players can look at while they argue, rather than straining to remember your narration.
Failure Should Cost Time, Not Progress
Ask yourself what happens if the players never solve this. If the answer is “the campaign stops,” redesign it immediately.
The healthiest failure state for a puzzle is cost. The door opens anyway, but it takes four hours, and something is waiting on the other side. Or it opens, but it’s loud, and now the whole dungeon knows they’re here. Or it opens, but the mechanism chews through the party’s last healing potion.
That structure does something important: it keeps the puzzle genuinely tense while guaranteeing your Thursday night doesn’t end in a stalemate. Players can feel the difference. A puzzle with a safety valve is fun to fail. A puzzle without one is a hostage situation.
Let Skills Do Work
Investigation, Perception, Insight, History, Arcana — these should be clue delivery, not gates. A successful check never means “you solve it.” It means “you notice something useful.”
“Can I roll Investigation?” is one of the most useful sentences a player can say at a puzzle, because it tells you exactly where they’re stuck. Give them the next clue, not the answer. Point at the second fresco. Mention that the dust on the floor hasn’t been disturbed near the left-hand panel.
A failed check shouldn’t mean nothing happens. It should mean the clue arrives with a complication — you find the inscription, but you have to clear rubble to read it, and clearing rubble takes time you might not have.
The Pre-Flight Checklist
Before a puzzle meets your players, run it through these. Some of them will make you rewrite. That’s the point.
- Can I state the goal in one sentence? If the players don’t know what they’re trying to accomplish, they’re not solving anything.
- Are there three clues per inference, in three different places?
- Can the players physically look at the clues? Handout or prop, ideally.
- Have I listed at least two alternate solutions besides the intended one?
- What does failure cost? If the answer is “the game stops,” go back to step four.
- Is any part of this solvable only by player knowledge, not character action? Trim it or add a skill route.
- Would I solve this myself if I only knew what I’ve told them? Be brutal. You know too much.
- Is there a timer, or can they brute-force it without consequence?
A Quick Worked Example
The Clockmaker’s Door. The players need to open a sealed vault door with three rotating dials and no visible lock.
Intended solution: the dials match three family crests in the entrance hall, arranged by the order the clockmaker’s children were painted — eldest to youngest. Clues: the portrait order, a birthday ledger, and a servant’s complaint about “the master’s fussing over the dials again.”
Brute force: the party’s artificer can force the mechanism in an hour of work, but it rings a bell in the workshop below and takes the party out of stealth for the rest of the floor.
Weird: the party’s bard, who has been flirting with the ghost of the clockmaker’s wife for two sessions, asks her nicely. She tells them. That’s not a failure of the design. That’s a campaign arc paying off.
When to Throw It Out Mid-Session
You will still, occasionally, watch a puzzle die in real time. When that happens, cut it. Thirty minutes is the ceiling. Have the mechanism click, the door swing open, and the sound of whatever’s behind it fill the room. Nobody at the table will ever know you bailed, and the people who feel relief will vastly outnumber the people who feel cheated.
A puzzle that ends early will be remembered as a tense scene. A puzzle that never ends will be remembered as the night the game stopped.