Noise-Induced Regime Transitions in the Iterated Prisoner's Dilemma: A Computational Study of Strategy Robustness

This paper presents a large-scale computational investigation of strategic robustness in the Iterated Prisoner’s Dilemma (IPD) under stochastic action corruption. By systematically varying noise levels from 0% to 50% and evaluating performance across 100 independently seeded tournaments per condition, the study identifies clear regime transitions in strategic dominance. Classical reciprocal strategies dominate in low-noise environments, become unstable in intermediate regimes due to retaliatory cascades, and are overtaken by defect-dominant strategies as signal reliability deteriorates. In addition to canonical strategies, the paper introduces a profitability-adaptive mechanism that conditions behavior on rolling payoff trends, demonstrating improved stability across moderate noise levels. The results highlight the non-monotonic stability of cooperation under imperfect information and provide quantitative evidence that signal reliability functions as a structural parameter governing equilibrium selection in repeated games. All simulations are fully reproducible, with code publicly available.