Skip to content
z1chhPublic

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

1 watching

Forks

Repository files navigation

Colosseum Survival!

Setup

To setup the game, clone this repository and install the dependencies:

pip install -r requirements.txt

Playing a game

To start playing a game, we need to use agents. For example, to play the game using two random agents (agents which take a random action), run the following:

python simulator.py --player_1 random_agent --player_2 random_agent

This will spawn a random game board of size NxN, and run the two agents of class RandomAgent. You will be able to see their moves in the console.

Visualizing a game

To visualize the moves within a game, use the --display flag. You can set the delay (in seconds) using --display_delay argument to better visualize the steps the agents take to win a game.

python simulator.py --player_1 random_agent --player_2 random_agent --display

Play on your own!

To play the game on your own, use a human_agent to play the game.

python simulator.py --player_1 human_agent --player_2 random_agent --display

Autoplaying multiple games

Since boards are drawn randomly (between a MIN_BOARD_SIZE and MAX_BOARD_SIZE) you can compute an aggregate win % over your agents. Use the --autoplay flag to run $n$ games sequentially, where $n$ can be set using --autoplay_runs.

python simulator.py --player_1 random_agent --player_2 random_agent --autoplay

During autoplay, boards are drawn randomly between size --board_size_min and --board_size_max for each iteration.

Notes

  • Not all agents supports autoplay. The variable self.autoplay in Agent can be set to True to allow the agent to be autoplayed. Typically this flag is set to false for a human_agent.
  • UI display will be disabled in an autoplay.

AI agent

I implemented the ai_agent.py file in agents/ directory (which extends the agents.Agent class) using a Monte-Carlo tree search algorithm. The agent can be tested using the information below

```
python simulator.py --player_1 random_agent --player_2 ai_agent --display 
```

It is possible to autoplay as follows:

```
python simulator.py --player_1 random_agent --player_2 ai_agent --autoplay
```

Full API

python simulator.py -h       
usage: simulator.py [-h] [--player_1 PLAYER_1] [--player_2 PLAYER_2]
                    [--board_size BOARD_SIZE] [--display]
                    [--display_delay DISPLAY_DELAY]

optional arguments:
  -h, --help            show this help message and exit
  --player_1 PLAYER_1
  --player_2 PLAYER_2
  --board_size BOARD_SIZE
  --display
  --display_delay DISPLAY_DELAY
  --autoplay
  --autoplay_runs AUTOPLAY_RUNS

Game Rules

Game Setting

On an M x M chess board, 2 players are randomly distributed on the board with one player occupying one block.

Game Moving

In each iteration, one player moves at most K steps (between 0 and K) in either horizontal or vertical direction, and must put a barrier around him or her in one of the 4 directions except the boarders of the chess board. The players move in a round-robin way.

Note:

  • Each player cannot go into other player's place or put barriers in areas that already have barriers.
  • Currently the maximal number of steps is set to K = (M + 1) // 2.

Game Ending

The game ends when each player is separated in a closed zone by the barriers and boundaries. The final score for each player will be the number of blocks in that zone.

$$S_i = \#\text{Blocks of Zone}_i$$

Goal

Each player should maximize the final score of itself, i.e., the number of blocks in its zone in the endgame.

Example Gameplay

Here we show a gameplay describing a $2$-player game on a $5\times 5$ chessboard. Each player can move at most $3$ steps in each round.

The final score is $A:B = 15:10$. So A wins the game.

License

MIT

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages