Agent-Based Simulation for Token Launches

Before you launch, simulate. cadCAD, TokenSPICE, Machinations, and Gauntlet's risk parameter optimization — how serious protocols stress-test their economies pre-launch.

30 min · expert · part of Token Engineering & Economic Modeling

Why You Cannot "Math Your Way" to a Token

Designing a token by closed-form math has a fundamental limitation: the relevant outcomes are emergent. Whether a token holds its peg, whether a lending market liquidates cleanly under stress, whether emissions create a stable yield curve or a death spiral — these properties depend on the simultaneous behavior of many heterogeneous actors (LPs, speculators, arbitrageurs, holders, liquidators) responding to each other's actions. Closed-form equilibrium analysis assumes everyone reads the equations and behaves rationally. Real markets do not. The response is agent-based simulation: build a computational model of the token economy, populate it with synthetic agents that behave according to plausible rules (some speculators, some long-term holders, some arbitrageurs, some bots), run thousands of paths under different parameter settings and shock scenarios, and observe what the system does. This is how serious protocols stress-test their designs before deployment, and how risk-parameter teams continuously monitor live protocols. The tooling that supports this — cadCAD, TokenSPICE, Machinations.io, and the proprietary tooling used by firms like Gauntlet — has matured significantly through 2022-2026. What used to be PhD-thesis level work is now an established part of the token-design pipeline at any protocol that takes mechanism design seriously. This lesson covers the major simulation frameworks, the firm (Gauntlet) that operationalized parameter optimization for live protocols, and the specific scenarios that pre-launch simulations actually test.

Also in this lesson

cadCAD: Block Science's Generalized Framework
TokenSPICE: Trent McConaghy and Ocean Protocol Origins
Machinations.io: Visual Game-Design-Style Modeling
Gauntlet: Continuous Risk Parameter Optimization
What Pre-Launch Simulations Actually Test

Key terms

Agent-based simulation: A methodology for modeling complex systems by representing many individual agents (each with its own behavior rules) and observing emergent system-level dynamics. The dominant pre-launch token-design methodology for serious protocols.
cadCAD: Complex Adaptive Dynamics Computer-Aided Design. Open-source Python framework developed by Block Science for agent-based simulation of token economies. Used by MakerDAO, Compound research, and many other protocols.
TokenSPICE: Open-source agent-based simulator developed by Trent McConaghy (Ocean Protocol co-founder). Opinionated toward crypto-economy patterns with pre-built agent templates for common roles (speculator, LP, arbitrageur, holder).
Machinations.io: Visual node-based modeling tool originally for video-game economies, adopted for crypto token-flow modeling. Strong for early-stage exploration of mechanism designs before detailed quantitative analysis.
Gauntlet: Risk-parameter management firm founded by Tarun Chitra. Primary external risk advisor to Aave (since 2020-2021), Compound, and Uniswap. Provides continuous agent-based simulation of live protocols with recommended parameter updates submitted to governance.
Chaos Labs: Competitor to Gauntlet in the risk-parameter optimization space. Provides parameter analysis for Aave, MakerDAO, Liquity, and others. The category is now competitive with multiple expert firms.
Liquidation cascade: A scenario where one set of liquidations crashes the collateral asset price, triggering further liquidations, which crashes the price further. The March 2020 MakerDAO event is the canonical example. Pre-launch simulation aims to verify the system can handle severe shocks without cascading.
Monte Carlo simulation: A statistical method that runs many randomized paths of a system to estimate distributions of outcomes. Used in cadCAD and Gauntlet's tooling to estimate tail risk, expected returns, and parameter sensitivities under uncertainty.
Risk parameter: A configurable value in a DeFi protocol that controls risk exposure — loan-to-value ratio, liquidation threshold, interest rate curve, debt ceiling, reserve factor, etc. Optimizing these for live conditions is the core service Gauntlet and competitors provide.
Black Thursday: March 12, 2020. ETH crashed 50% in a day, gas prices spiked, and MakerDAO's liquidation auctions cleared ETH at 0 DAI bids because liquidators could not get transactions through. Resulted in $5.4M in bad debt and a fundamental redesign of MakerDAO's auction mechanism. Canonical example of what proper pre-launch stress testing aims to catch.

Continue this lesson — 5 more sections in the CryptoBipto app.

Open lesson

Educational only — not financial advice.