Paper: Algorithmic Monetary Policies for Blockchain Participation Games
Authors: Diodato Ferraioli, Paolo Penna, Manvir Schneider, Carmine Ventre
Date: 18 Dec 2025
Estimated Reading Time: 27 minutes

This paper studies how algorithmic monetary policies can balance performance and decentralization in blockchain participation games. The authors model repeated participation where agents with heterogeneous type and stake decide whether to participate in each round. Token rewards affect both stake distribution and decentralization, which in turn endogenously determines token value. Policies that strongly favor high-type agents maximize short-term efficiency but risk long-term centralization. The paper contrasts outcomes under myopic agents and agents with asymmetric lookahead who partially internalize future effects. It shows that foresight or policy-level simulation of foresight can sustain decentralization above critical thresholds, albeit with volatility. The work also analyzes virtual stake mechanisms that blend type and stake, highlighting persistence of initial inequalities.

Core insights

The model fixes per-round token budgets and focuses on how allocation rules reshape supply distribution rather than total supply. This isolates the effect of reward allocation on decentralization and participation incentives, but assumes a mature system where marginal rewards are less valuable than changes in token value. Under this assumption, decentralization becomes the primary driver of demand, since token value is a non-decreasing function of decentralization among active participants.

Performance-centric policies that allocate most rewards to the highest-type participant shift supply toward already advantaged agents. For myopic agents, this concentrates stake and reduces decentralization, lowering token value and eventually deterring participation by others. The analysis shows that this dynamic is not an edge case but can arise from simple reward rules. A key question raised implicitly is whether any purely performance-based allocation can avoid this concentration without additional constraints.

When agents have asymmetric lookahead, participation decisions incorporate future rounds only when considering abstention. This behavioral asymmetry allows agents to tolerate short-term losses in decentralization to regain higher token value later. The resulting equilibria maintain decentralization above a threshold sufficient to keep at least one participant active. However, decentralization can still hit its minimum, leading to sharp drops in token value. This raises the question of whether token holders would tolerate such volatility in practice, even if long-run participation is preserved.

The paper’s simulated-lookahead policy shifts foresight from agents to the protocol. By occasionally selecting winners that are not the highest type, the policy stabilizes decentralization with myopic agents. This reduces reward variance across participants and flattens stake accumulation, at the cost of reduced throughput in some rounds. The analysis suggests a supply-side tradeoff where smoother stake distribution supports sustained demand through higher perceived decentralization.

Virtual stake mechanisms blend type and stake into a single selection weight, effectively fixing reward shares from the initial state. Because these shares are invariant over time, early inequalities in stake translate into persistent dominance unless type differences are sufficiently large. This implies that virtual stake policies cannot correct unequal initial distributions without strong performance weighting. The result prompts the question of whether bootstrapping phases require fundamentally different monetary rules than steady-state operation.

Overall, the paper frames tokenomics as a dynamic interaction between reward allocation, participation incentives, and endogenous token value. It highlights that reward rules influence demand indirectly through decentralization, not just supply growth. The findings suggest that policies aiming for long-term decentralization must either rely on agent foresight or encode foresight directly into protocol design.