Reinforcement Learning Produces Dominant Strategies for the Iterated Prisoner's Dilemma

TL;DR

This paper demonstrates that reinforcement learning techniques can develop highly effective strategies for the Iterated Prisoner's Dilemma, outperforming traditional strategies in tournament settings, including noisy environments.

Contribution

It introduces novel reinforcement learning-based strategies that outperform existing methods in the Iterated Prisoner's Dilemma, validated through extensive tournament results.

Findings

Trained strategies outperform all opponents in standard tournaments.

Reinforcement learning strategies excel even in noisy conditions.

The strategies are effective against a diverse set of over 170 opponents.

Abstract

We present tournament results and several powerful strategies for the Iterated Prisoner's Dilemma created using reinforcement learning techniques (evolutionary and particle swarm algorithms). These strategies are trained to perform well against a corpus of over 170 distinct opponents, including many well-known and classic strategies. All the trained strategies win standard tournaments against the total collection of other opponents. The trained strategies and one particular human made designed strategy are the top performers in noisy tournaments also.