The Art of War: Beyond Memory-one Strategies in Population Games

TL;DR

This paper introduces an 'information player' strategy for population games that leverages machine learning to identify and adapt to opponents, outperforming all known memory-one strategies in invasion success and cooperation.

Contribution

It presents a novel, uninvadable strategy for population games that uses historical data and learning techniques to adaptively respond to various strategies.

Findings

The information player can invade and outperform all known memory-one strategies.

It naturally learns to cooperate with similar players.

The strategy is effective against a wide range of strategies including TFT, WSLS, and ZD.

Abstract

We define a new strategy for population games based on techniques from machine learning and statistical inference that is essentially uninvadable and can successfully invade (significantly more likely than a neutral mutant) essentially all known memory-one strategies for the prisoner's dilemma and other population games, including ALLC (always cooperate), ALLD (always defect), tit-for-tat (TFT), win-stay-lose-shift (WSLS), and zero determinant (ZD) strategies, including extortionate and generous strategies. We will refer to a player using this strategy as an "information player" and the specific implementation as $IP_0$. Such players use the history of play to identify opponent's strategies and respond accordingly, and naturally learn to cooperate with each other.