Guide
How to win at Snap Judgments: deduction when the clues are opinions
Strategy for Snap Judgments and other guess-who style games. Pick questions that split the board, read your opponent, and know when to guess.
Classic guess-who games are pure logic: either the person wears glasses or they don't. Snap Judgments is messier. The question cards are opinions, like "This person has a Pinterest board for a life they are still building," and your opponent answers based on their own read of their secret face. You're not just narrowing down a list. You're modeling another person's judgment.
This guide covers both halves: the math of good questions, and the psychology of subjective answers.
Part 1: the math of a good question
Aim for an even split
Every match starts with 16 faces. If every question cut the remaining faces exactly in half, you'd need only four questions: 16 to 8 to 4 to 2 to 1. That's the best any yes-or-no question can do, because each answer gives you at most one bit of information.
Lopsided questions are much weaker than they feel. Here's how many faces you should expect to have left after one question on a full board, depending on how the question splits it:
| Faces that would answer "yes" | Faces that would answer "no" | Expected faces left |
|---|---|---|
| 8 | 8 | 8.0 |
| 6 | 10 | 8.5 |
| 4 | 12 | 10.0 |
| 2 | 14 | 12.5 |
| 1 | 15 | 14.1 |
The expected count is (yes² + no²) ÷ 16, because a question leaves you with the "yes" group or the "no" group, in proportion to their size. A question that fits only one face usually leaves you with 15. You'll occasionally get lucky, but on average it barely moves the needle.
Before you play a card, estimate how many of the remaining faces it fits. Pick the card closest to half.
Use the redraw
Once before each question, you can swap your whole hand for five new cards without losing your turn. If every card in hand fits almost none or almost all of the remaining faces, redraw. A balanced question is worth more than any one card you'd lose.
Write your own when the cards don't fit
You can always ask a custom yes-or-no question instead of playing a card. Late in a match, when three or four faces remain, the cards rarely split them well. A plain, objective custom question like "Is your person wearing something on their head?" can split the last few faces cleanly, and nobody can argue with the answer.
Part 2: the psychology of subjective answers
Predict their answer, not yours
When you play "This person would order the most expensive thing on the menu," what matters isn't which faces you think fit. It's which faces your opponent would say fit. If you know them, use it. If they're a stranger, assume they'll go with the most obvious visual read.
Flip only what you're sure about
A "yes" answer tells you to flip down faces that don't fit, but with subjective questions some faces are honestly borderline. Flip down only the faces you're confident about. A face you wrongly flip is gone for good from your thinking, and if it was the secret, you'll end up guessing wrong. Keeping two or three extra "maybe" faces costs you a turn or so. Losing the real answer costs you the match.
Watch for patterns in their answers
Over a few questions you'll learn how your opponent judges. Some players answer very literally. Others go with gut feel. If they said "yes" to two roast cards about confidence, they probably read their face as a big personality, and later answers will follow that read.
Answer honestly and consistently
You're answering questions too, and the game only works if both players answer in good faith. Decide early how you see your character and stay consistent. Random or contrarian answers make the game worse for both players, and an observant opponent will notice.
Part 3: when to guess
Under the default rule, wait
By default, any wrong face guess loses the match immediately. With two faces left, a guess is a coin flip. One more well-chosen question almost always gets you to one face, so guess only when you're down to one face, or when you're sure your opponent will win on their next turn.
With three strikes, guess earlier
If the host turned off "wrong final guess loses," you can be wrong twice and keep playing. Then a face question is often the most efficient question on the board:
| Faces left | Chance a face question is right | Guesses needed in the worst case |
|---|---|---|
| 2 | 50% | 2 |
| 3 | 33% | 3 |
| 4 | 25% | 4 (too many: you'd lose on the third miss) |
With two or three faces left and no strikes used, guessing directly is usually faster than asking another card question.
Watch the race
Snap Judgments is a race. If your opponent seems close to the answer, a risky guess can be better than a safe question that leaves them one more turn. If they're far behind, take the extra question and guess with certainty.
A sample match plan
- Questions 1–2: play the most balanced cards in your hand. Redraw if none splits the board reasonably.
- Questions 3–4: use what you've learned about how your opponent answers. Pick cards they'll answer predictably.
- Three or four faces left: switch to objective custom questions to split the remainder cleanly.
- One face left: make your final guess.
The same principles apply to any guess-who style game. Ask questions that split the board in half, respect uncertainty, and don't guess until the odds are on your side.