README

The code under this directory implements a set of tools for computing
GTO strategies for two players playing heads-up no limit hold'em.

It uses the peval library from pokerstove (see
https://github.com/andrewprock/pokerstove).  Follow the previous link
to see how to build the library.  Once it's done, modify the `PREF'
variable in `Makefile' to the location of the local source of
pokerstove.  And just type:
$ make depend && make

General source file guide:

err.h
        Implementation of some functions for signaling errors.

cfr.h
        Interface to a node in a game tree that also implements the
        counterfactual regret minimization algorithm to compute GTO
        strategies. It doesn't support chance node.  See page 12 of
        the following paper:
        http://modelai.gettysburg.edu/2013/cfr/cfr.pdf.

cfr-inl.h
        Implementation of some template functions that print the
        strategies used by the players under a given node.

range.h
        Implementation of a range of hands that a player can hold
        post-flop.

equi_dist.h
        Implementation of an equity distribution between two players
        post-flop.

preflop_range.h
        Implementation of a range of hands that a player can hold
        pre-flop.

preflop_equi_dist.h
        Implementation of an equity distribution between two players
        pre-flop.

3bet.cc
        Compute the strategies for a 3-bet shove game on the flop in
        an unraised pot.

5bet.cc
        Compute the strategies of a 5-bet shove game that takes place
        in pre-flop.

proto/
        Directory containing a prototype implementation of 5bet.cc in
        Common Lisp.  If quicklisp is installed just create a symbolic
        link to `proto' under `~/quicklisp/local-projects' and type
        from the REPL:
        CL-USER> (asdf/operate:load-system :gto)
        T
        CL-USER> (gto:train)