Properties of winning Iterated Prisoner's Dilemma strategies. A meta tournament analysis.

Authors: @Nikoleta-v3 @drvinceknight @marcharper

This repository contains the source code and analysis for the paper: "Properties of winning Iterated Prisoner's Dilemma strategies" which is available at: arXiv:2001.05911.

Software

To clone the repository locally run the following:

$ git clone https://github.com/Nikoleta-v3/meta-analysis-of-prisoners-dilemma-tournaments.git

A conda environment specifying all versions of the libraries used is given in environment.yml. To create and activate the environment run:

$ conda env create -f environment.yml
$ source activate axlml

Reproducing the results

The raw data containing the summary results of each tournament analysed in the paper is achieved at: https://zenodo.org/record/3753498

To download and unpack the archived data run the following command whilst the conda environment is activated:

$ inv data

To process the raw data run:

$ inv process

Note that this will take a few minutes.

The process task creates four data sets located in the folder data/:

• standard_3_processed.csv
• noise_3_processed.csv
• probend_3_processed.csv
• noise_probend_3_processed.csv

These are also available at https://zenodo.org/record/3753565. To download the processed data sets without carrying out the process task run:

$ inv prodata

Analysis

Most of the analysis presented in the paper is carried out in the Jupyter notebooks found in the folder nbs/.

The analysis is also carried out the merged data set of the entire data sets considered in the paper. To generate the merged data set run the command:

$ inv merge

To carry out the regression and correlation analysis, presented in the notebooks 3. Linear Regression and 2. Correlation Table and Heatmaps, for the merged data set run the commands:

$ python src/correlation.py
$ python src/regression.py

Tests

Several scripts have been tested with pytest. To test run:

$ inv test

The software is released under an MIT license.