Go to the first, previous, next, last section, table of contents.


Game design

Bluff was designed with the following requirements.

  1. Bluff should be playable on a network, between untrusted players who might resort to cheating. To discourage this, any alliance between two players must not be able to damage players outside the alliance: the total score of any alliance must not exceed, on average, the total score the same players would achieve playing separately.
  2. Trivial strategies, such as never bluffing or always bluffing, should not be optimal. To be true to its name, Bluff should encourage a reasonable amount of bluffing.
  3. There should be a balance between the skill and luck elements.
  4. The minimum duration of a game should not be too long, not to bore players.

While the second and third requirement are not too hard to satisfy, the first one rules out many game variants. If players are not to gain an advantage by communicating in secret with each other, most information must be made public, and each player's private information must not bear any correlation to any other player's.

Thus, Bluff uses dice (which are rolled independently of each other), rather than cards (which are unique in a deck: knowing N-1 people's cards can reveal fundamental information on the N'th player's hand). Rather than passing around a single set of dice, each player uses a different set: this reduces the amount of secret knowledge that cheaters might want to share behind the scenes. Lastly, as much information as possible (amount of dice rolled, bids, who trusts whose bid...) is disclosed to all players on the same footing.

The first phase of scoring (give points to the highest bidder) is meant to satisfy half of the second requirement, by encouraging players to bluff. The second phase (trusting/not trusting other players), on the other hand, makes overbluffing a dangerous habit, especially when other players catch on.

After the trust phase is over, the list of people who bluffed in the current turn is made public, regardless of who trusted whom. This is again a consequence of the first requirement. If bluff information were only revealed to those who call the bluff, players might share this information behind the scenes; if bluffs were only revealed when someone takes a risk and calls them, alliances could agree on a single player to act as "guinea pig" -- thus gaining an advantage against outsiders.

The choice of scoring is open to tuning. The only requirement is that the penalty for a discovered bluff must be higher than the gain for a high bid. When the discovered bluff penalty becoms higher, players are encouraged to call bluffs more lightly, and are discouraged from bluffing. Increasing the cost of calling a bluff when there was none has the opposite effect.

The choice of repeating the game several times, with an increasing amount of dice visible, is due to the third and fourth requirements. Players have to estimate the probability of reaching a certain score starting from certain conditions; this increases the element of skill in the game.

Limiting the game to three turns, instead of five as could seem more natural, is another skill-vs-luck decision. In very long games, lucky players will achieve a high score (say, a full or better) in an early hand; they would then keep this score throughout the match, earning many points from a single lucky dice roll. A three-turn game reduces this problem, and also decreases the chances of the game becoming boring.


Go to the first, previous, next, last section, table of contents.