A simple implementation of a variant of the Risk board game for python, designed for playing with AIs.
Runs in python
(2.7 or 3.x) using the curses
library to display the map (but can be run in pure-console mode).
python pyrisk.py FooAI BarAI*2
Use --help
to see more detailed options, such as multi-game running. The AI loader assumes that SomeAI
translates to a class SomeAI
inheriting from AI
in ai/some.py
.
A minimal version of the Risk rules are used:
- Players start with
35 - 5*players
armies. - At the start of the game, territories are chosen one by one until all are claimed, and then the remaining armies may be deployed one at a time to reinforce claimed territories.
- Each turn, players recieve
3 + territories/3
reinforcements, plus a bonus for any complete continents. - A player may make unlimited attacks per round into adjacent territories (including from freshly-conquered territories).
- Each combat round, the attacker can attack with up to three armies.
- Upon victory, a minimum of that combat round's attackers are moved into the target territory.
- The attacker may cease the attack at the end of any combat round.
- The defender defends with two armies (unless only one is available).
- Each attacking and defending army rolls 1d6. The rolls on each side are ordered and compared. The loser of each complete pair is removed, with the defender winning ties.
- At the end of each turn, a player may make one free move
- Victory is achieved by world domination.
Write a new class extending the AI
class in ai/__init__.py
. The methods are documented in that file. At a minimum, the following functions need to be implemented:
initial_placement(self, empty, remaining)
: Return an empty territory if any are still listed inempty
, else an existing territory to reinforce.reinforce(self, available)
: Return a dictionary of territory -> count for the reinforcement step.attack(self)
: Yield(from, to, attack_strategy, move_strategy)
tuples for each attack you want to make.
The AI
base class provides objects game
, player
and world
which can be inspected for the current game state. These are unproxied versions of the main game data structures, so you're trusted not to modify them.