Completed PythonSimulationmatplotlibData Science

Werewolf Simulator

A Monte Carlo simulation of the social deduction game Werewolf, verifying the counterintuitive result that the informed minority (werewolves) consistently wins.

Overview

A simulation of the social deduction game Werewolf (also known as Mafia), built to verify a result I’d heard discussed but never seen properly evidenced: that despite being the minority, the werewolves — as the informed party — hold a structural advantage and tend to win.

What It Shows

The simulator runs thousands of games across varying player counts and werewolf-to-villager ratios, tracking win rates for each side. The results consistently confirm the hypothesis:

The informed minority wins more often than the uninformed majority — the werewolves’ knowledge of each other creates coordination the villagers cannot match.

This is a neat real-world application of information asymmetry theory.

How the Simulation Works

  • Each game initialises a player pool with configurable werewolf/villager ratios
  • Voting and elimination logic is modelled probabilistically
  • Werewolves coordinate (shared knowledge), villagers vote based on limited inference
  • Results are aggregated and plotted with matplotlib

Tech Stack

Python · matplotlib