MEZARPOLY

Should you lie in Bluff?

13 November 2026 · 3 min read

The game is named after lying, but people don't lie equally: some never do, some do every turn. Which one wins? We played thousands of games, each time changing a single habit of one player.

1. How the experiment worked
2. Never lying loses
3. Lying every time isn't the answer either
4. Call BLUFF! a little more often
5. When you lead, name the rank you hold most
6. Summary
7. Method and limits

How the experiment worked

At a four-player table one player changed a single habit; the other three were Bluff's normal-difficulty bots. These bots track their own cards, the revealed cards and each player's history; they work out whether a lie will pay and what a challenge costs (the size of the pile). The seat of the player with the changed habit rotated from game to game. Each habit was tested over 3,000 games; the fair share is 25%.

Never lying loses

An honest player who always passes when they don't have the rank wins only 19.0% of games. In Bluff, passing means not playing cards; while your opponents empty their hands, you wait. A small lie at the right moment is what drives the game.

Lying every time isn't the answer either

The opposite extreme doesn't work either: a player who lies every time they don't have the rank wins 26.7% of games, almost the same as the standard player. They tell 9.7 lies per game and 73.7% of them are caught. A player who never passes stands out: when someone plays every turn even though they're unlikely to hold the rank, opponents call more readily. (Mezarpoly's Bluff bots now watch for this too.)

Getting caught is expensive: you take the whole pile. So a good lie is told while the pile is small and your card count looks believable to your opponents. We worked out the maths of this in When to call bluff.

Call BLUFF! a little more often

Among calling habits, the only one that made a difference was calling more often: a player who calls BLUFF! on every play that is more than 30% likely to be a lie wins 29.2% of games. Never calling (23.5%) or calling only when almost sure (23.8%) doesn't differ from the standard player.

The reason: challenging a small pile is cheap. If you're wrong you take a few cards; if you're right the liar takes the pile and you lead the next round. As the pile grows, a challenge gets more expensive and you need to be more sure.

When you lead, name the rank you hold most

The leader chooses which rank to name. A player who names a random rank instead of the one they hold most wins 20.7% of games. If you name your majority rank, you get through that round with true cards and can keep playing that rank all round.

Summary

• Lie, but not every time. A player who never lies stays at 19.0%; most lies of a player who always lies get caught.
• Be bolder about challenging small piles. Calling once a lie is more than 30% likely beats the standard player.
• Lead with the rank you hold most.
• Watch the player who never passes. They're usually lying.

Method and limits

• The games were played with the Bluff rules engine on the Mezarpoly server: the café rule (same rank all round, passing allowed, the pile burns when everyone passes), one deck. 21,000 games in total.
• The opponents were bots. People also read lies from faces, voices and hesitation; results at a real table may differ. The direction should hold, though: never lying and never calling are both costly.

More from the blog