Dupery WikiIndependent case library

Field guide

Dupery truth, lies and taint: can you trust a clue?

The short answer

Check the speaker’s actual role, taint and the clue’s trigger before trusting an announcement. Good alignment and a familiar appearance do not prove truth; a disguise also does not automatically make a statement false.

Detailed role examples use the Demo v0.8.2 text. Check the role and quirk descriptions in your installed build if they differ.

Taint changes information, not the whole identity

The dated Demo mechanics distinguish taint from alignment. A tainted source can give false information while retaining its category and alignment. This matters because arresting a normally good source just for an unreliable statement can be a second mistake.

Think of a clue as a path from the world, through a role’s question, through the speaker’s information state, to the announcement you hear. The weak point can be the speaker. Repeating the same calculation more carefully does not repair a false input.

Role checksits specific questionReliability matterstruth / lie / taintYou receivethe displayed statement
Information reliability sits between the role’s question and your deduction. This is an explanation of the relationship, not an in-game status display.

Two good roles can behave differently

The Drunkard and Copycat illustrate why “good” and “truthful” need separate columns. In the Demo text, the Drunkard lies under a good-role disguise. The Copycat imitates a good role present in the case without adding the Drunkard’s lie rule. Both are normally good Meddlers, but their information does different work.

Normally goodMeddler categoryDrunkardDisguisedLiesCopycatDisguisedCopiesfunction
Alignment alone cannot tell you which information rule applies. Read the actual role and any case override.

Check a contradiction in this order

  1. Question: Are the clues about the same property? “Not Innocent” can still mean a good Meddler.
  2. Scope: Does the ability inspect these addresses, this row, or the surrounding area? An excluded scope is not a negative answer.
  3. Time: Did something change after the original statement? Milkman is a clear reason to preserve the investigation time.
  4. Source: Could a disguise or taint explain why the displayed role’s rule is not giving normal information?
  5. Case rule: Does a quirk alter the assumptions you were making?
Same question?Same scope?Same time?Same target state?Reliable source?Applicable case rules?
Use this checklist to locate the disagreement before choosing a culprit.

Keep a useful clue conditional

Suppose a displayed Mathematician says 7. You can explore the address sets that would fit 7 while keeping the note “if this announcement is reliable.” That calculation is useful: it tells you what another clue needs to separate. It does not establish who the speaker really is.

If the combination search has no result, first check your available addresses, your required/excluded list and any count you entered. Then revisit reliability and special counting rules such as Fall Guy. An impossible input model is a reason to inspect the model, not automatic proof of a game bug.

Do not “correct” a suspect total by guessing one higher or one lower. The materials used here do not establish a universal offset for a lying sum.

Further reading

Next moves

Keep working the case