Current documentation: version 0.2007.01.30

The Gambit Users Guide is available in the following formats:

• You can browse it online. (This version is generally the most up-to-date.)
• You can download the HTML files as one tarball. The main page is in gambit/index.html in the archive.
  
• Note that the documentation is already bundled with the Windows installer.

Documentation for libgambit, the game representation and manipulation library, is under development. These are generated from the libgambit source code using the excellent tool Doxygen

• You can browse it online. (This version is generally the most up-to-date.)
• You can download the HTML files as one tarball. The main page is in libgambit/index.html in the archive.
  

Tutorials

Designed as a complement to the manual, these are some tutorials illustrating the use of the graphical interface to analyze some simple games. (These tutorials depict version 0.2005.06.13, but the differences from that version to the current one are minor.)

• A simple game of poker. Illustrates how to build and analyze an extensive form game, using a one-card poker game as an example. [Flash animation] [PDF]
• Quantal response equilibria in a coordination game. Illustrates the construction of a game in strategic form, and calculation of the principal branch of the quantal response equilibrium correspondence. [Flash animation] [PDF]

Sample games

Here are some example games in Gambit format.

• [2x2x2.nfg]. A three-player normal form game with two strategies per player, due to McKelvey and Palfrey. This game has nine Nash equilibria, which is the maximal number of regular Nash equilibria possible for a game of this size.
• [2x2x2-nau.nfg]. A three-player normal form game with two strategies per player, due to Bob Nau. This game has three pure strategy equilibria, two equilibria which are incompletely mixed, and a continuum of completely mixed equilibria. In the interests of full disclosure, this last feature tends to give Gambit fits!
• [bagwell.efg]. Stackelberg leader game with imperfectly observed commitment, from Kyle Bagwell, Commitment and Observability in Games, Games and Economic Behavior 8: 271-280, 1993.
• [bayes2a.efg]. A twice-repeated Bayesian game, with two players, each having two types and two actions. This game also illustrates the use of payoffs at nonterminal nodes in Gambit, which can substantially simplify the representation of multi-stage games such as this.
• [cent3.efg]. A three-stage centipede game of the sort studied by McKelvey and Palfrey, among others. This game features an exogenous probability that one player is an "altruist" who always passes.
• [condjury.efg]. A three-person Condorcet jury game, after the analysis of Feddersen and Pesendorfer.
• [loopback.nfg]. A game due to McKelvey which illustrates that the logit quantal response equilibrium correspondence can have a "backward-bending" segment on the principal branch.
• [montyhal.efg]. The famous "Monty Hall" problem: if Monty offers to let you switch doors, should you?
• [nim.efg]. The classic game of Nim. This version starts with five stones.
• [pbride.efg]. A signaling game from Joel Watson's Strategy textbook, modeling the confrontation in The Princess Bride between Humperdinck and Roberts in the bedchamber.
• [poker.efg]. A simple game of one-card poker used in Myerson (1991), Game Theory: Analysis of Conflict. For a slightly more complex example, check out [4cards.efg], a four-card game contributed by Alix Martin.
• [spence.efg]. A version of Spence's classic job-market signaling game. This version comes from Joel Watson's Strategy textbook.