Quant Risk Modeling

Contract

• Input: problem description and inputs defined by the skill body.
• Output: Markdown artifact with completed process steps.
• Side effects: none.
• Dependencies: none.
• Stop condition: all process steps executed; artifact saved with required sections.
• Risk: low.
• Boundary: produces reasoning artifact only; no system changes.

Quant Risk Modeling

Build a risk measure — VaR, CVaR, drawdown, stress — with explicit distributions, assumptions, and historical validation. A risk number without context is misleading.

When to use

• The user needs a risk metric for a portfolio or strategy.
• A backtest needs a drawdown / stress profile.
• A strategy needs a position-size or leverage cap based on risk.

Process

1. Define the risk object

State exactly what needs measurement: portfolio-level P&L, a single factor, a strategy, an instrument. Time horizon (1-day / 10-day / monthly) matters.

Completion criterion: risk object and horizon named; measurement unit (absolute / % / $) specified.

2. Choose distribution / model

• Parametric: Normal, Student-t, GARCH, multivariate Normal / t.
• Non-parametric: historical simulation, bootstrap.
• Scenario / stress: specific shocks (rates +200bp, equity crash 2008).
• Credit / default: CVaR for defaults, PD/LGD/EAD framework.

Name the choice and why it fits the regime (e.g. "t-distribution for fat tails").

Completion criterion: model named; assumption (distribution, correlation, independence) listed.

3. Compute

Run with explicit parameters. Report:

• VaR at chosen confidence (e.g. 95%, 99%).
• CVaR / ES (expected shortfall) — more stable than VaR.
• Drawdown: max, average, recovery time.
• Stress: impact of named historical or hypothetical shock.

Completion criterion: all requested measures present with exact parameters.

4. Validate

• Backtest coverage: does VaR 95% cover ~95% of out-of-sample days? If not, recalibrate.
• Stress test: does the model predict known historical losses within a factor?
• Stability: does the risk measure jump wildly on small data changes?

Completion criterion: out-of-sample coverage or stability check completed; discrepancy addressed.

5. Deliver

Markdown artifact: risk definition, model, measures, validation, and a note on what the number means operationally (e.g. "VaR 99% = $1.2M; expected loss beyond is CVaR 99% = $2.5M").

Completion criterion: deliverable includes measures, validation, and an operational interpretation.