Separate the accusation from the clue
A resident named by an announcement is not automatically an arrest target. The source may be unreliable, the question may be about category rather than alignment, and the clue may treat a special role differently. Fall Guy is the important example: other roles can count it as evil while it is normally a good Meddler.
For a useful final check, say the accusation aloud in precise terms: “At this time, this address is evil because these clues fit these rules.” If your sentence rests on “the card looks evil” or “one speaker said so,” you may still be missing an identity or reliability check.
Choose a separator, not another random clue
If one explanation puts the culprit at 2 and another puts it at 3, look for information that tests the difference. A clue about an unrelated region of the board may consume time without helping. A further investigation is worthwhile when it changes the decision you would make.
This is a default reasoning method, not a promise that waiting is always harmless. Read timed role descriptions before spending time: a late alignment change, a kill or another deadline can change the case. Use the remaining uncertainty and the relevant clock together.
When your clues depend on an assumed truthful speaker, keep that assumption visible. Two accusations built from the same unreliable statement are not two independent confirmations.
Check the mistake cost in this run
The official description says good-aligned arrests reduce reputation and too little reputation can end the run. It also identifies Critic as a role that doubles reputation loss. This changes how expensive a speculative arrest can be even when the basic rules have not changed.
Do not plan your entire run around a remembered fixed recovery amount. The September 17 announcement restored between-case recovery in the Demo and said higher difficulties in the full game would depend on stamps for reputation gains. That is a version and difficulty condition, not a guarantee of free recovery after every case.
Treat the reveal as new information
An arrest can teach you something about a resident, but a hidden reveal is not a proof of innocence. Conman’s Demo description says its arrest does not reveal its identity. Keep the original display and statement in your notes so the action does not erase the reasoning that led you there.
If a result contradicts your model, correct the assumption that failed. Do not keep excluding addresses just to force the remaining arithmetic into one set. Work back through the dependency chain and preserve any conclusions supported independently.