Pick Two: An Adversarial Animal Survival Game

Abstract

The "Pick Two" animal selection puzzle is a popular thought experiment in which two animal species must defend a human against the remaining animal attackers. While typically discussed informally, the scenario presents a heterogeneous coalition-selection problem involving complex interactions among agents with different capabilities and behaviors. In this work, we formalize Pick Two as an adversarial multi-agent optimization problem and develop a biologically inspired agent-based simulation framework to evaluate defender coalition effectiveness. Coalition performance is evaluated through 18,000 Monte Carlo simulations conducted in a Unity-based environment. Results show that coalition effectiveness is not additive and is instead dominated by interaction effects and scaling behavior. Overall, this study demonstrates how agent-based simulation can be used to analyze coalition effectiveness in adversarial environments and highlights the importance of emergent group dynamics in determining collective success.