Who’s the Spy — Strategy and Clue Tactics

What this page assumes

• Everybody has the same word except one player, who is given nothing.
• That player is not told they are the odd one out — they have to work it out from what everyone else says.
• The category is public before the first clue. There are 9 of them.
• Everybody gives one clue aloud, in seat order, then the table votes.
• A caught player gets one guess at the word from a shortlist of 10.
• A clue turn runs to about 15 seconds, and the whole clue phase has a length the host sets.

If any of that is new, the rules page is the place to start; this one is about what to do with them.

The clue problem, stated properly

A good clue has to fail one test and pass another at the same time: obvious to somebody holding the word, useless to somebody who is not. Most players optimise only for the second half, produce something so broad it could not identify anything, and are then surprised to be voted for — because a clue that gives nothing away also proves nothing.

The reliable move is to describe a relationship the word has rather than a property it has. Properties are guessable from the category alone. Relationships — what it sits next to, what it replaces, who complains about it — require the specific thing, and they are much harder to fake from the outside.

Why the category being public changes everything

This is the detail most tables underuse. Because the 9 categories are announced before the round, the odd player out already knows which world the word lives in. A clue pitched at category level — the kind that would be true of any answer on the list — hands them a free turn and costs you nothing they did not already have.

Pitch below the category, not at it. And note the corollary: if the table has agreed on a narrow category, everybody’s clues get harder at once, and the round gets shorter and sharper.

Turn order is not fair, and you should plan for it

First is hardest. You are speaking into silence: nothing to lean on, no established register to match, and whatever you say becomes the foundation the odd player out builds their whole round on. Say less than you think you should.

Last is most suspicious. Not because late clues are worse, but because the table has been listening for a while by then and a clue that arrives after all the evidence looks assembled from it. If you have the word and you are last, be more specific than feels comfortable — you can afford it, and the alternative is being voted for on vibes.

The middle is where the odd player out should aim. Enough cover to borrow a register, early enough not to look like waiting.

Spotting the one with nothing

The common heuristic — vagueness — is close to useless, because nervous players with the word are vague too, and confident players without it are not. The better tell is direction:

• A clue that reacts rather than adds. It agrees, echoes the shape of an earlier clue, or reframes something already said. Ask yourself whether they could have gone first with it. If not, that is the signal.
• A clue that would fit the whole category. Not vague — broad. Those are different, and only the second is evidence.
• Over-specificity with nothing behind it. A player who has worked out the word late will sometimes over-correct into a detail nobody else thought worth mentioning.

And the anti-tell: hesitation is not evidence. It is the most-voted-for behaviour at every table and it correlates with being shy, not with being the spy.

When you are the one with nothing

First, do not try to deduce the word before you speak. You usually cannot, and the attempt is what produces the long pause that gets you voted for. Take the first clue you hear, move it sideways — same register, different angle — and say it without hedging.

Second, commit. A confident wrong clue survives a vote far more often than a careful one, because a table votes on delivery long before it votes on content.

Third, if you are caught, you still have the guess from 10 — so spend the clue phase listening for the word rather than only for cover. Being voted out having worked out the answer is a different outcome from being voted out blind.

The one thing civilians get wrong at the end

When a vote already looks settled, the remaining clues stop being useful and start being a liability. A last, unnecessarily precise clue does not change a vote that is already going the right way — it just hands the answer to somebody who is about to be asked to guess it. Once the room has decided, say less.

Common questions

What makes a good clue when you know the word?

One that would be obvious to somebody holding the same word and almost useless to somebody who is not. The test is not how clever it sounds — it is whether a player with nothing could repeat it convincingly. If they could, you have given away nothing and also proved nothing, which is the trap most players fall into for the first few rounds.

Why is going first harder than going last?

Because you have no clues to hide behind. The first clue is given into silence, so it has to be vague enough to survive being wrong about what everyone else will say, and it is also the clue the odd player out will build their entire round on. Later speakers get to sound like the ones before them, which is exactly the cover the first speaker is denied.

How do you spot the player who does not have the word?

Not by vagueness, which is the common mistake — nervous players are vague too. Look for a clue that reacts rather than adds: one that agrees with what has already been said, restates the shape of an earlier clue, or lands so broadly it would fit any answer in the category. The tell is not what they said, it is that they could not have said it first.

What should you do if you are the one with no word?

Speak in the middle, not last. Going last looks like you waited, and the table is already primed to notice it. Take an earlier clue, move it sideways rather than repeating it, and commit — a confident wrong clue survives far longer than a hedged one, because the table reads hesitation and not accuracy.

Is the category announced before the round?

Yes, and this is the single most useful thing to plan around. The 9 categories are public the moment the round opens, so everybody knows which world the word lives in before a clue is given. That is what makes a lazy clue fatal: "you can sit on it" narrows nothing when the whole table already knows the category.

What happens if the spy is voted out?

They get one guess at the word from a shortlist of 10, so being caught is not the end of it. This changes how the civilians should play the last clue of a round: if the vote already looks settled, an unnecessarily specific clue does nothing for you and hands the caught player the answer.

Does the strategy change on one phone versus separate screens?

A little. On separate screens up to 8 people see the word privately and simultaneously. Sharing one device seats up to 12, and the phone going round means there is a gap between seeing the word and speaking — which is thinking time the odd player out gets too, and it makes them noticeably harder to catch.