By David Anderle Implemented project Finance Simulation Python

Volatility Cascade Engine

I built a simulation of how an initial shock in one asset can propagate through a leveraged financial system through mark-to-market losses, margin breaches, forced liquidations, and impact-driven secondary declines.

Download project View Python file

Problem

Most people talk about market stress as if losses scale smoothly with the size of the initial shock. In leveraged systems, that assumption can fail. Once enough portfolios cross their margin thresholds, forced selling can create a second wave of losses that is larger than the first.

Model logic

Each fund has equity, leverage, a margin threshold, a liquidation rule, and overlapping exposures to a shared set of assets. After an initial shock hits CHIP, the engine revalues portfolios, checks for breaches, forces distressed funds to sell, applies price impact, and repeats until the system stabilizes.

Shock range tested

2% to 30%

First distressed fund

16% shock

Three distressed funds

20% shock

System loss at 30%

34.3%

What the outputs show

The system looks stable under smaller shocks. Then the shape changes. At 16%, the first distressed fund appears. At 18%, there are two. At 20%, there are three. System loss rises from 11.1% at a 14% shock to 16.9% at 16%, 21.6% at 18%, and 25.7% at 20%. The exact numbers are not the point. The visible nonlinearity is.

Shock size versus system loss chart from the Volatility Cascade Engine — Final system loss steepens once liquidations begin feeding back into prices.

Shock size versus distressed funds chart from the Volatility Cascade Engine — Distress stays at zero until the system crosses a boundary, then begins to spread.

Fund-asset network after an initial CHIP shock — A network view makes overlapping exposures and spillover paths easier to inspect.

Why this project matters

I can turn an abstract systems idea into a working model.
I think in terms of thresholds, propagation, and feedback loops.
I care about how systems fail under pressure, not just how they look in calm conditions.

What it is not

This is intentionally simplified. It is not a production risk engine, a forecasting product, or a claim about real institutions. Its value is that it makes a clear systems argument visible in code, charts, and interpretable outputs.