Tournament-Winning Strategy for Iterated Optional Prisoner's Dilemma

Abstract

Iterated optional prisoners dilemma (IOPD) is an adversarial game that can be used to model several real-world scenarios, from mutual grooming between primates to alliances between business firms. This study utilizes simulation techniques to determine winning strategies for IOPD tournaments in a variety of initial conditions. Machine learning techniques are used to iteratively improve upon the winning strategy, culminating in a single undefeated strategy. The outcome of this study is a single strategy that we claim is likely to win an IOPD tournament for most reasonable initial conditions. Iterated optional prisoners dilemma (IOPD) is an adversarial game that can be used to model several real-world scenarios, from mutual grooming between primates to alliances between business firms. This study utilizes simulation techniques to determine winning strategies for IOPD tournaments in a variety of initial conditions. Machine learning techniques are used to iteratively improve upon the winning strategy, culminating in a single undefeated strategy. The outcome of this study is a single strategy that we claim is likely to win an IOPD tournament for most reasonable initial conditions.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 01, 2020
Accession Number
AD1126579

Entities

People

  • Ahmed A. Shamma

Organizations

  • Naval Postgraduate School

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Abstracts
  • Alliances
  • Artificial Intelligence
  • Cold War
  • Commerce
  • Computer Programs
  • Computer Science
  • Department Of Defense
  • Game Theory
  • Information Operations
  • International Relations
  • Learning
  • Machine Learning
  • Mathematics
  • Prisoners
  • Recreation
  • Simulations
  • Social Sciences
  • United States
  • Ussr

Readers

  • Economics
  • Game Theory.
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML