Your one-stop shop for the AIWerewolf Competition
This repository provides the necessary resources for participation in the annual AI werewolf contest organized by AIWolf.org. This includes rulesets, sample clients, and tutorials for each supported programming languages.
Current AI Werewolf Competition:
- Venue: Tenth International Automated Negotiating Agents Competition (ANAC)
- Date and Location: IJCAI 2019, Macao, August 10th to 16th
- Server Version: AIWolf Server 0.5.* (latest bugfix 0.5.5)
- Agent Submission Deadline: July 7th, 2019
What is available at this repository?
This repository contains a summary of English information necessary to join the AIWolf competition. In particular, this repository contains:
- Game Rules and Protocol Specification in the documentation directory.
- Information about the gameserver
- Code samples and HOWTO write an agent in JAVA
- Code samples and HOWTO write an agent in Python
- Code samples and HOWTO write an agent in C#
Where can I find more information?
You can send e-mails with technical and rules questions to the firstname.lastname@example.org mailing list.
If you have any questions, please join our Slack channel! Fill this Google Form to request an invite.
What is AI Werewolf?
Werewolf (also sometimes called Mafia) is a party/social game. The goal of the game is to simulate a town invaded by werewolves, and is based on the dynamics of uninformed majority vs informed minority.
The basic rules are as follow: Each player is a Human or a Werewolf. The werewolf players know who the other werewolves are, but the human players do not know who is a human and who is a werewolf. Each turn (day), the players discuss and vote on one player to be removed from the game. Usually, the werewolves will pretend to be human and try to accuse other humans of being werewolves, in order to get they voted out of the game. After each voting round the werewolves can choose to eat one player at night, removing them from the game.
The game can be further complicated by the addition of roles, such as seers who can divine who is a werewolf and who is a human, and bodyguards, who can prevent a player from being eaten by a werewolf.
AIWolf is a project that aims to build AI agents that can play the werewolf game. The challenges of the werewolf game include: Modeling the behaviors of players, estimating the likelihood of werewolf and humans, discussion and negotiation with other players on voting strategy, and Natural Language understanding and speaking.