A card gets maybe thirty seconds of a room's attention. In that window it either starts something, a story, an argument, a round of accusations and laughter, or it produces a shrug and gets tapped past, and which of those happens is mostly decided before the night begins, by whoever chose the question. The craft of choosing turns out to be learnable, and it compresses into four questions you can ask of any card, plus two rules that sit above all four.
We apply this test to every prompt before it ships, and nothing about it depends on our decks. Write your own cards for your own table and it works exactly the same way.
1. Does it ask for a story or a fact?
"Your favourite film" is a fact. It gets answered in two words, the room nods, and the card is over. The version that earns its place asks for the story wrapped around the fact:
- Tell us: the film you have defended in an argument?
Same territory, different request. The second version cannot be answered without a scene appearing, who the argument was with, why it mattered, whether they won, and the scene is the conversation. When a card of yours is producing two-word answers, this is usually the fix: keep the topic, ask for the moment.
2. Can everyone in the room answer it?
A question game runs on answers meeting answers, and a card only half the room can answer switches that engine off for the other half. Questions that assume a job, a sibling, a relationship, a sport or a particular childhood all quietly split the room into players and spectators. The test is concrete: go around your actual guest list one face at a time and check the card asks each of them for something they have. The best cards need nothing but a life:
- Would you rather repeat your best year or skip your worst one?
Nobody lacks the raw material for that one, which is precisely what makes it safe to deal to any table.
3. Is there a real cost on both sides?
A would-you-rather with an obviously right answer is a preference wearing the costume of a dilemma. The room answers in unison and there is nothing to discuss, because choosing costs nothing. A dilemma earns its place when both options give something up:
- Would you rather lose the ability to lie or always know when someone is lying?
Both sides of that are genuinely expensive, which is why the room divides, and the division is the card working. Before you deal a choice card of your own, ask which side you would take. If you cannot imagine a person you respect taking the other side, it is not a dilemma yet.
4. Would anyone disagree?
A most-likely-to that nobody would deny is a compliment, and a compliment ends the card: the room agrees, the named person smiles, next card. The format runs on somebody objecting to the accusation, so the accusation has to be deniable, small enough to laugh at and specific enough to argue about.
- Most likely to double-dip and maintain eye contact?
- Hot take: being late is a personality flaw, not a quirk.
Every finger that goes up for the first one will be argued with, and the second reliably splits any room that contains one punctual person and one relaxed one. The same check applies to hot takes: an opinion the whole room already holds is not a take, it is a toast.
The two rules that outrank all four
A card can pass every test above and still not ship, because two standing rules sit over the top of the whole catalogue. Nothing that forces a disclosure, and nothing that makes one person the joke. A card cannot know which room it will land in, so a question that demands a confession, or aims the laugh at whoever is holding the phone, is out regardless of how well it scores on story, cost or disagreement. The longer argument for both rules, and for why they protect the fun rather than limiting it, is in why most icebreakers are boring. Writing for your own table, these two are the ones to keep even if you keep nothing else.
Running the test on a page of your own cards
In practice the four questions collapse into one pass. Read each card and predict the ten seconds after you deal it: who speaks, and what do they say? If the honest prediction is a two-word answer, apply test one and ask for the moment instead of the fact. If the prediction only features half your guest list, that is test two. If everyone answers the same way, tests three and four. You will keep perhaps half of what you wrote, and the half you keep will play better than the whole page would have, because every survivor now starts something. Cut generously: the cards you delete cost you nothing, and the flat card you leave in costs you thirty seconds of a warm room every time it comes up.
The machinery that holds the standard
For the reader deciding whether the catalogue is worth paying for, the honest answer to "who chooses the questions" is: a process, kept deliberately boring. Four parts.
An editorial log, kept since launch. Every prompt that has been removed, reworded or flagged is recorded with its reason, so a decision made in March cannot be accidentally unmade in August by someone who never saw it.
Automated duplicate detection with fixed thresholds. Every candidate prompt is compared against the entire shipped catalogue on its content words, and a pair that shares too much, above 55 percent overlap, or where one prompt's content is three quarters contained in another's, fails the build until a human resolves it. Near-duplicates are the most common way a growing catalogue quietly gets worse, and the scan does not get tired. It also runs new batches against each other, not only against what has shipped, because two writers drafting independently will collide in places neither of them can see.
Blocklists enforced by the test suite. When a prompt is removed, its signature goes into a list the automated tests check on every single change, which means a cut card cannot drift back in through a later batch of new questions. Removal is permanent by machinery, not by memory.
A human approval step. Nothing ships on passing the automated checks alone. Every batch is read, and the writer's own least-confident cards are flagged for a second opinion before release rather than after.
What we do not publish is the list of what was cut, or the reasoning behind any particular removal. That is deliberate, and it is the same courtesy a good host extends at the table: the standard is public, the individual judgements stay in the room. What we can show you is the test, who applies it, and the machinery that stops it eroding.
The time the test was wrong
One reversal is worth telling in full, because a standard you can trust is one that admits its mistakes. "Would you rather sneeze glitter or hiccup bubbles" was cut from an adult deck as filler: nothing at stake, arbitrary answer, no conversation afterwards, a textbook failure of tests three and four. Months later, someone drafting a family deck wrote almost exactly the same question, and our own blocklist rejected it. Looking at it again, the blocklist was wrong. In a room with a nine-year-old and a grandparent, absurd is not filler, it is the one register where the youngest person at the table can win. The question was never bad. It was bad in that deck. So the filler rules are now scoped to the adult decks, the card lives in the family catalogue, and the safety rules stayed global, because safety does not vary by audience the way taste does.
The test tells you whether a question deserves a slot. It does not tell you which slot: that is ordering, covered in how to build a twenty-card round. And if you want to see the standard applied at scale rather than described, the format system and the design notes show the same thinking from two other angles.
The catalogue this test built.
Every deck in Chinwag is the survivors: questions that ask for stories, cost something on every side, and give the whole room a way to disagree.
Read nextWhy most icebreakers are boringThe research behind the standard →