This is a sample of Python code from a larger No-Limit Texas Hold'em (NLHE) poker analysis project. It may be of interest to someone building their own poker library.
You can find many other examples of poker programs online, most of which deal with random dealing, ranking hands, keeping track of betting etc. This project has a different focus. It is designed to solve more unusual analysis problems that I encountered as a professional poker player such as finding isomorphic/canonical representations of the flop, and finding nuanced hand properties (e.g. having 'three-to-a-straight' and 'two overcards') rather than 'which hand ranks higher'.
- pokertools.py: providing the basic building block: a Card type
- properties: many predicates of tuples of cards, e.g. is_twopair, is_rainbow
- examples/bluffing.py: finding nuanced flop hand properties for bluffing
- examples/translation.py: parsing PokerStove-style card notation using a tokeniser
- examples/isomorph.py: working with suit isomorphisms on the flop
The word 'hand' in poker is ambiguous. It could mean: the two hole cards I have been dealt; the best 5-card poker hand I can make with any combination of my hole cards and the board; the unit of a poker game starting with dealing hole cards and ending with the winner collecting the pot.
In this library, the first two meanings are important to distinguish. I reserve
holecards for the former and hand for the latter.
Some of the techniques in the above have been used to create an approximate Nash Equilibrium strategy solver for a simplified version of NLHE. One day I might release more code from that project, such as:
- a tool to construct simplified NLHE game trees recursively
- an iterative constraint-satisfaction algorithm using NumPy to split ranges of holecards into e.g. betting and checking actions according to given criteria such as equity and polarity
- a tool using the 'sklearn' machine learning framework to analyse hand history data