Nathravorn / cubirds_study

Study of the card game Cubirds.

Project

This is a Python project to study the board game Cubirds.

The game is accessible through the Game class, which allows one to create a game and play it until the end. When the game reaches the end, text will be printed to indicate this.

Basic analysis functions are defined in game_analysis.py. In particular, the function available_moves(game) returns a dictionary linking lay options (ways to lay down birds on the board) to flock options (ways to make a flock of birds), with a probability mapping for each flock option in the case of a lay which induces a draw of two random cards. A custom estimation of the cards remaining in the deck can be supplied to this function. Otherwise, it supports using the base probabilities of the game or probabilities based on card counting.

Example

The project can be tested by initializing an instance of Game and describing it using the state_summary method.

Output of print(Game(1).state_summary()):

Current turn: 0
Current player: 0
Current phase: lay

Board:
Row 0:  sparrow,     cube,    hibou
Row 1: sandwich,     duck,   parrot
Row 2:   toucan, sandwich,  sparrow
Row 3:  sparrow,  flamant,   parrot


Player 0:
Hand: cube, sandwich, sandwich, duck, duck, hibou, flamant, flamant
Collection: sandwich



Deck:
{
    "cube": 18,
    "duck": 10,
    "flamant": 4,
    "hibou": 8,
    "parrot": 11,
    "sandwich": 15,
    "sparrow": 14,
    "toucan": 9
}

In this example game, our analysis function can be tested using this script:

game = Game(n_players=1, n_rows=4)
print(game.state_summary())
am = available_moves(game)
print_available_moves(am)

Which gives (truncated):

{
  "('cube', 0, 'left')": [
      "small_flamant"
  ],
  "('cube', 0, 'right')": [
      "small_flamant"
  ],
  "('cube', 1, 'left')": {
      "('big_flamant',)": 0.1237698081734779,
      "('small_duck', 'small_flamant')": 0.013010842368640535,
      "('small_flamant', 'small_hibou')": 0.007506255212677232,
      "('small_flamant',)": 0.8557130942452046
  },
  "('cube', 1, 'right')": {
      "('big_flamant',)": 0.1237698081734779,
      "('small_duck', 'small_flamant')": 0.013010842368640535,
      "('small_flamant', 'small_hibou')": 0.007506255212677232,
      "('small_flamant',)": 0.8557130942452046
  }
  ...
}

Tests and benchmarks

You will need to install pytest and pytest-benchmark.

To run tests, run ./test [options]; to run benchmarks, run ./perf [options]. The options are pytest options. To save benchmarks for later comparison, use ./perf --benchmark-autosave

Todo

[X] Make a class for a stack of cards, with methods to draw, add, subtract, get and set in dictionary and list form.
[X] Make a class for a game of Cubirds to store a persistent game state, with methods to play rounds and advance the game to the end.
[>] Study strategies by simulating random player actions.
[ ] Refine analysis by making agents smarter (using a tree search algorithm with simulation of random events).
[ ] Develop an Alpha-Zero-like reinforcement learning approach to learn optimal strategies through self-play.