Previously, I released a package for solving Leduc Hold'em with chance sampled counterfactual regret minimization. This can be thought of a generalization and extension of that package. In particular, this code allows one to solve small zero-sum extensive-form games using either linear programming or counterfactual regret minimization. It can be downloaded here (55kb tar.gz).
A brief list of features:
- General extensive-form games, as described by game tree enumeration
- Linear programming and counterfactual regret equilibrium-finding algorithms
- Best response and exploitability computations
- Some support for chance abstraction
- Extensible design
To use the linear programming functionality, which is required for high precision equilibrium solutions and exploitability calculations, you will need GLPK. The counterfactual factual regret solver benefits from ATLAS. Neither of these packages is required to compile the software.