We introduce a hierarchical framework for modeling agent risk preferences in environments exhibiting nonlinear payoff structures, stochastic amplification, and regime-dependent inversions between conservative and speculative returns. Unlike classical risk models that assume static utility functions, our framework characterizes agents by their sensitivity to environmental potential—a state variable governing phase transitions between linear and convex return regimes. We define ten agent tiers ranging from stability seekers to chaos maximizers, each implementing distinct switching policies. Our main theoretical contribution establishes that a hysteresis-based switching strategy (Tier 7) achieves strictly superior long-run wealth accumulation compared to both conservative and hyper-aggressive strategies in oscillating environments. This result provides a formal foundation for understanding optimal timing in regime-switching decision problems and has applications in algorithmic trading, evolutionary game theory, and multi-agent systems under environmental uncertainty.