What We Learned Modeling a Regen Token Economy

April 1, 2025 · Max Semenchuk
agent-based-modelingtokenomicsresearch

Summary

We built an agent-based model (ABM) of the Regen token economy to test how issuance, burns, rewards, and liquidity interact with eco-credit demand. We ran baselines, parameter sweeps, and stress tests to find policy regions that increase timely retirements while keeping security and liquidity healthy. The result is a governance-ready menu of parameter ranges, code, and reproducible experiments.

Demo session recording:

Key Findings

  • Higher burn share reduces net inflation and shortens retirement lag, but validator rewards need a protected APR floor and sufficient AMM depth.
  • Maintaining liquidity depth during shocks reduces slippage and preserves price stability more than increasing issuance alone.
  • Parameter regions exist where retirements grow faster than issuance without harming validator security, especially when regrowth “r” is capped and AMM rewards are targeted.

Policy Takeaways

  • Cap regrowth r within a defined band and route a share of rewards to AMM depth to offset higher burn share.
  • Maintain validator APR within a floor range in shocks to prevent security erosion.
  • Use an efficient-frontier plot (retirements up vs inflation down) to anchor governance choices.

Problem and Approach

Problem Definition

Core problem: Identify issuance and scarcity policies for $REGEN that most reliably increase timely retirement of eco-credits relative to issuance, while maintaining token security, liquidity, and governance feasibility.

Key questions:

  • What funding distribution to activities (liquidity, credit curation, creation) yields the highest marginal ecological plus economic benefit per $REGEN?
  • Which parameter ranges for regrowth r, burn share, validator APR, and caps minimize net inflation while maximizing retirement rate and reducing retirement lag?
  • How do shocks to liquidity, validator set, or eco-credit demand propagate through staking, prices, and retirements?

Why ABM

  • Heterogeneity: Delegators, speculators, issuers, LPs, validators act with different rules and constraints.
  • Feedbacks and path dependence: Burns on retirements, staking incentives, and liquidity co-evolve with demand, creating non-linear dynamics that static methods miss.
  • Policy exploration under uncertainty: Monte Carlo and stress tests across many micro-rules reveal robust policy regions, not single “optimal” points.

Environment and Interactions

  • Environment: Markets for $REGEN spot, staking, and eco-credits. Treasury/policy module applying issuance regrowth, burns, fees. Exogenous demand processes for eco-credits and liquidity shocks.
  • Interactions: Agents submit orders, staking changes, issuance and retirement events. The environment clears markets, updates prices, applies burns and regrowth, and updates validator/security metrics.

Data and Metrics

Data Sources

  • On-chain: Staking ratios, validator sets, APRs, churn, slashing events. Token price and volume histories. Liquidity depth metrics.
  • Registry and market data: Eco-credit issuance and retirement time series by category and size. Retirement lag distributions.
  • External analogs: ETH post-EIP-1559 burn dynamics, Cosmos staking participation, comparable carbon markets for demand regimes.

KPIs

  • Ecological: retirement rate and lag, issued vs retired, issuance-to-retirement ratio
  • Economic: net inflation, price stability, AMM depth, slippage, fee accrual
  • Security and participation: staking share, validator APR, churn
  • Distribution: ownership concentration and Gini coefficient

Model Overview

Agents and Roles

  • Fund/treasury: mints, splits rewards between holders and AMM, optional sell policy.
  • Holders: hodlers, validators (rewarded), traders (biased micro-flows).
  • Liquidity providers and AMM: constant-product pool with fees and optional external LP in/out.
  • Issuers: create and retire eco-credits.

Key Parameters and Policies

  • Issuance regrowth r, burn share, reward splits (holders vs AMM), validator APR target band, fee bps.
  • Demand: initial external demand in $, elasticity, growth rate, per-tick cap.

Safety Guards

  • Max trade fraction of pool per tick.
  • Demand cap fraction to bound exogenous buy-and-burn.
  • Oracle smoothing to reduce noise.

Experiments

Baselines

Low demand, high demand, validator drop, LP shock.

Parameter Sweeps

  • r ∈ [0.01–0.15], burn_pct ∈ [0.1–0.5], validator APR ∈ [5–20%], elasticity bands.
  • Monte Carlo runs (N=100 per scenario).

Stress Tests

  • Liquidity shock (–50% LPs)
  • Validator exit (top decile)
  • Eco-credit boom/drought

How to Run It Yourself

Run the NetLogo prototype at: https://www.netlogoweb.org/launch#Load

Set sliders: initial-price, issuance-rate, distribution-fraction-of-mint, amm-fee-bps, demand-elasticity, external-lp-flow, etc. Click Setup → Go. Watch price, TVL, cumulative minted/burned.

Results and Interpretation

Tools

  • Fidelity vs speed: NetLogo for prototyping behaviors and UI; cadCAD for batch experiments and Monte Carlo.
  • Reproducibility: notebooks, seeds, CSV outputs.
  • Communication: dashboard built with Replit made findings interactive, but infra costs (~$10/day) pushed us back to a video + source release.

Limitations

  • Single-pool AMM, stylized demand, simplified heterogeneity.
  • No explicit cross-venue fragmentation or cross-chain arbitrage (can be added as network topology later).

Governance Implications

Policy Menu Candidates

  • Issuance regrowth r bands that avoid runaway inflation under typical demand.
  • Reward split ranges that maintain validator APR while preserving AMM depth.
  • Fee bps guidance to support TVL and reduce slippage without killing flow.

Decision Aid

Show an “efficient frontier” plot: ecological retirements up vs inflation down. Use it to select parameter bands that meet minimum validator APR and liquidity depth thresholds.

Possible Next Steps

  • Calibrate with Regen on-chain and registry datasets; publish parameter bands with confidence intervals.
  • Propose a phased pilot: adopt conservative r and burn_pct bands with monitoring triggers and explicit rollback conditions.

References