texas-poker-computer

---

Description

这个仓库旨在建立德州扑克规则的代码基础，并且构建一套能够使用不同的算法去计算手牌获胜概率的计算系统。

This repository aims to establish a codebase for Texas Hold'em rules and build a system capable of calculating hand winning probabilities using various algorithms.

MileStone

Stage 1

能够使用遍历所有公共牌组合的办法，使用统计的方法去计算两个已知手牌的玩家的胜率。 这个方法非常缓慢，需要遍历的情况有 C(48,5)=1,712,304 种可能的公共牌。 此外，使用的Evaluator策略为Default的（遍历七张牌中所有可能的五张牌组合 C(7,5)=21 ）， 这意味着遍历的量级在千万级别。运行一次的时间在50min左右，有很大提升空间。

This method uses a statistical approach to calculate the win rate for two players with known hands by traversing all combinations of community cards. This method is very slow, requiring traversal of C(48,5)=1,712,304 possible combinations of community cards. Additionally, the Evaluator strategy used is the default one (traversing all possible five-card combinations from seven cards C(7,5)=21), which means the scale of traversal is in the tens of millions. The runtime is about 50 minutes, and there is significant room for improvement.

Stage 2

使用完全规则的方法，设计了一个更好的七张牌判断牌型的 Evaluator，大幅度降低复杂度，运行一次的时间直接降低到了 3min 左右。

Using a method based on complete rules, a better Evaluator for determining hand types with seven cards has been designed, significantly reducing complexity. The runtime has been directly reduced to about 3 minutes.

Stage 3

使用多进程（并非多线程）的方法优化循环结构，大幅提高速度，运行一次的时间降到了 1m30s 左右。

By optimizing the loop structure with multiprocessing (not multithreading), the speed has been significantly increased, reducing the runtime to about 1 minute and 30 seconds.

Stage 4

增加蒙特卡洛模拟，使用样本量在 100,000 量级，误差可以基本控制在 0.3% 以下。 时间有了更大幅度提高，测试两个手牌相对的胜率，多进程的情况下，时间已经优化到5s 左右，是完全可以接受的了。

Monte Carlo simulations have been added, using a sample size of around 100,000, with the error generally controlled below 0.3%. The time efficiency has significantly improved; testing the relative win rates of two hands, with multiprocessing, the time has been optimized to about 5 seconds, which is entirely acceptable.

下面是使用蒙特卡洛模拟 ♥A, ♦A 与 ♣A, ♣K 对局的结果，可以看到在 50_000 次模拟之后已经接近稳定。

Below are the results of using Monte Carlo simulations for a matchup between ♥A, ♦A and ♣A, ♣K. It can be seen that after 50,000 simulations, the results are already approaching stability.

Project

Structure

src/
├── calculator/
│   ├── __init__.py
│   └── calculator.py
├── game/
│   ├── comparator/
│   │   ├── __init__.py
│   │   └── hand_comparator.py
│   ├── evaluator/
│   │   ├── __init__.py
│   │   ├── hand_rank.py
│   │   ├── hand_ranking.py
│   │   └── hand_ranking_evaluate.py
│   ├── __init__.py
│   ├── game.py
│   ├── player.py
│── poker/
│   ├── __init__.py
│   ├── deck.py
│   ├── poker.py
│   ├── rank.py
│   └── suit.py
├── __init__.py
└── main.py

Module

• poker模块用于定义和描述扑克的点数和花色
• game模块定义了牌局的相关信息
• evaluator模块是计算一组牌中最大牌型的模块，可以实现和选取不同的策略
• calculator模块是胜率计算的模块，可以完成不同的情景的胜率计算

---

• The poker module is used to define and describe the ranks and suits of poker cards.
• The game module defines information related to the game.
• The evaluator module calculates the highest hand from a set of cards and allows for the implementation and selection of different strategies.
• The calculator module is responsible for calculating win probabilities under various scenarios.

Unit Tests

项目使用单元测试保证所有规则的准确性以及运算的正确性。

The project uses unit testing to ensure the accuracy of all rules and the correctness of calculations

pytest tests