Main repository for the Axelrod Tournament of the Web-based Autonomous Systems course that uses the Axelrod 4.13.0 Python library.
The library requires Python 3.6 or greater. Install with pip:
pip install axelrod
Each player submits their strategy for the Prisoner's Dilemma. The tournament is a round robin run (one iteration) so that each player is paired with every other player. Each player is also paired with its own twin and with RANDOM, a strategy player that cooperates or defects in each round with equal probability. Each match between two players consists of 10 rounds. The player whose strategy accumulated the greatest score wins.
Scores are attributed based on the Prisoner's Dilemma where each player has two possible actions, Cooperate (C), and Defect(D). The following negative payoffs indicate how many years each player will spend in prison (in absolute value):
| Cooperate (C) | Defect (D) | |
|---|---|---|
| Cooperate (C) | (-1,-1) | (-5,0) |
| Defect (D) | (0,-5) | (-3,-3) |
Run python3 was-tournament/was_tournament.py (was_tournament.py) to run a simple tournament of 10 rounds (1 round robin) between a TitForTat, a Grudger (i.e. a Trigger), a Defector, a Cooperator, and a Random.
The tournament produces three files analysing the results:
- was_results.png: visualizes tournament normalized scores
- was_win_distributions.png: visualizes tournament win distributions
- was_payoff_matrix.png: visualizes tournament payoff matrix
- was_tournament_analysis.csv: summarises tournament results
Additionally, the program prints the morality metrics calculated for each player of the tournament:
- Cooperation Rate: The fraction of interactions in which the player cooperated.
- Good-Partner Rate: The fraction of interactions in which the player cooperated at least as much as its opponent (excluding interactions between a player and its own clone).
- Eigenjesus Rate: A metric that always favors cooperation, and gives more (positive) weight to cooperations with moral opponents than to cooperations with immoral opponents.
- Eigenmoses Rate: A metric that favors cooperation with moral opponents, and defecting immoral oppontents. It gives a positive weight to cooperations with moral opponents, and a negative weight to cooperations with immoral opponents.
Before implementing your strategy, you can run python3 was-human-inclusive-match/was-human-inclusive-match.py (was-human-inclusive-match.py), and try out different startegies by
playing as a human against a TitForTat player.
Create a new branch, and implement your strategy in the was-tournament.wasstrategies package. For further examples, see an additional example strategy that considers the history of interactions, or these variations for Tit-for-Tat.
Your module name should have the format {your_name}_strategy.py and your new strategy class name should have the format {Name} (e.g., Danai()). See danai_strategy.py as an example.
You can always modify the provided simple tournament to try out your strategy against basic Axelrod strategies.