WhereWholf!

Library to simulate a game of Werewolf. Works like a modified variant of Ultimate.

The goal is that this can be used to write a game with actual human participation (...and, eventually, pass the Turing Test, totally). But we are not there yet.

Design Decisions

Werewolf is a game that relies on social manipulation and trickery. Players lie all the time and that is part of the fun. The code is designed to reflect the deceptive nature of this game: as much as possible, we try to prevent Player objects from being able to introspect¹ other players.

This might result to un-Pythonic code. Wherever this is done, the accompanying documentation will make note of this decision.

¹ In as much as we can do so in Python. Ultimately, classes might be able to read memory segments it should not have access to, should this bug exist in the Python interpreter. See also: Python's "We are all adults here" philosophy.

Game Entities

For a player, a Wherewholf game is divided into two phases: night time and day time.

Night time actions is role-dependent and facilitated largely by the moderator. Most notably, during this phase, the player can learn things about the game (e.g., other players' roles).

Day time is further divided into phases and each player has a set of attributes that could influence their behavior during these phases:

1. Nomination phase. Players accuse each other of being a werewolf. A player's aggression attribute dictates how likely he is to nominate others.
2. Discussion phase. In a real-life game, this is where players argue over the nominations, coordinating their votes on who to lynch. In this simulation, each nominated player can advocate for themselves, the efficacy of this is dictated by the pesuasiveness attribute (DnD players might also call this charisma).
3. Votation phase. The player with the most votes gets lynched. Players who made a nomination in the first phase will automatically vote for who they nominated. In turn, players who were nominated will never¹ vote for them. For everyone else, their suggestibility attribute will dictate how likely will they vote to lynch a nominated player.

At the end of the day time phase, the role of the lynched player is revealed; this is information gained by all the players in the game.

¹ This might change upon introduction of the Tanner role.

Running

A requirements.txt file is included, listing all Python dependencies. This should run with a vanilla Python 3.7 interpreter but for development, the dependencies are conveniences you'd want to have.

To run,

python -m src.main

You can set the following environment variables too, mostly for debugging:

• WHEREWHOLF_MISC_LOG - control the log output of computations that are not strictly part of the main game loop.