#poker #poker-cards #card

nightly rs_poker

A library to help with any Rust code dealing with poker. This includes card values, suits, hands, hand ranks, 5 card hand strength calculation, 7 card hand strength calulcation, and monte carlo game simulation helpers.

55 releases (2 stable)

Uses new Rust 2024

new 3.0.0-beta.38 Mar 28, 2025
3.0.0-beta.28 Jan 3, 2025
3.0.0-beta.26 Nov 30, 2024
3.0.0-beta.25 Jun 29, 2024
0.4.0 Apr 26, 2017

#11 in Games

Download history 107/week @ 2024-12-06 65/week @ 2024-12-13 8/week @ 2024-12-20 59/week @ 2024-12-27 163/week @ 2025-01-03 27/week @ 2025-01-10 19/week @ 2025-01-17 16/week @ 2025-01-24 22/week @ 2025-01-31 15/week @ 2025-02-07 11/week @ 2025-02-14 8/week @ 2025-02-21 318/week @ 2025-02-28 200/week @ 2025-03-07 118/week @ 2025-03-14 308/week @ 2025-03-21

946 downloads per month
Used in better-hand

Apache-2.0

395KB
8K SLoC

rs-poker

Crates.io Docs.rs

RS Poker is a rust library aimed to be a good starting place for many poker rust codes. Correctness and performance are the two primary goals.

Core

The Core module contains code not specific to different types of poker games. It contains:

  • Suit type
  • Value type
  • Card type
  • Deck
  • Hand iteration
  • Poker hand rank type
  • Poker hand evaluation for five-card hands.
  • Poker hand evaluation for seven card hands.
  • PlayerBitSet is suitable for keeping track of boolean values on a table.

The poker hand (5 cards) evaluation will rank a hand in ~20 nanoseconds per hand. That means that 50 Million hands per second can be ranked per CPU core. The seven-card hand evaluation will rank a hand in < 25 ns.

The hand evaluation is accurate. rs-poker does not rely on just a single kicker. This accuracy allows for breaking ties on hands that are closer.

Holdem

The holdem module contains code that is specific to holdem. It currently contains:

  • Starting hand enumeration
  • Hand range parsing
  • Monte Carlo game simulation helpers.

Arena

Arena is a feature that allows the creating of agents that play a simulated Texas Holdem poker game. These autonomous agent vs agent games are ideal for determining the strength of automated strategies. Additionally, agent vs agent arenas are a good way of quickly playing lots of GTO poker.

  • Holdem simulation struct for the overall status of the simulation
  • Game state for the state of the current game
  • Agent trait that you can implement to create your more potent poker agent.
  • A few example Agents.
  • Historians who can watch every action in a simulation as it happens

Arena CFR Agent

CFRAgent is an agent that uses the Counterfactual Regret Minimization algorithm to choose the best action. The agent is a good starting point for creating a strong poker agent.

To implement your own strategy you will need to build a new ActionGenerator. The ActionGenerator is responsible for generating all possible actions for a given game state. The CFRAgent will then explore possible results of trying the actions suggested by ActionGenerator. The Agent will choose the action it would most regret not taking.

Testing

The code is well-tested and benchmarked. If you find something that looks like a bug, please submit a PR with an updated test code.

5 Card + Hand iteration is used with fuzzing to validate the seven-card hand evaluation.

Fuzzing is used to validate game simulation via replay generation.

Multi-agent simulations are used to validate correctness and performance.

Dependencies

~1.4–3MB
~52K SLoC