Documentation / Stress Test Methodology — Shock Propagation & P&L

Stress Test Methodology — Shock Propagation & P&L

Run stress tests that compute P&L impact from scenario shocks and portfolio exposures. Five shock methodologies, impact matrices, and attribution analysis.

Table of Contents

What Are Stress Tests?

A stress test calculates the projected P&L impact of a scenario's shocks on your portfolio's exposures. It answers the question: "If this scenario happened, how much would my portfolio gain or lose?"

Stress tests connect two things you've already built — scenarios (with their risk factor shocks) and portfolio snapshots (with their sensitivities). The platform applies each shock to the corresponding exposure to compute factor-level and total P&L impact.

Running a Stress Test

To run a stress test:

  1. 1Navigate to Stress Tests and click "New Stress Test."
  2. 2Give it a name and optional description.
  3. 3Select one or more scenario-portfolio pairs. Each pair combines one scenario's shocks with one portfolio snapshot's exposures.
  4. 4Choose a shock methodology (see below) to determine how historical shocks are selected for each risk factor.
  5. 5Click "Run" — results compute immediately.

Shock Methodologies

When a scenario has multiple historical shocks for the same risk factor, the methodology determines which one is used:

MethodologyHow It Selects Shocks
Worst CasePicks the shock with the largest absolute magnitude for each factor — the most extreme move regardless of direction.
Historical MinPicks the most negative shock for each factor — the worst historical drawdown.
Historical MaxPicks the most positive shock for each factor — the largest historical rally.
DirectionalFollows the net price trend direction. If the factor has been trending up, uses the max shock; if down, uses the min.
MeanAverages all available shocks for each factor — a moderate, central-tendency estimate.
Which to choose?
For regulatory submissions, Worst Case is the most conservative and commonly expected. For internal analysis, Directional or Mean can provide a more nuanced view based on current market trends.

Understanding Results

After a stress test completes, you see several views of the results:

  • Impact matrixA heatmap showing total P&L impact for each scenario-portfolio pair. Color intensity indicates magnitude — click any cell to drill into its factor-level detail.
  • Category breakdownP&L impact grouped by risk factor category (equities, rates, FX, commodities, credit, volatility, macro, inflation). Shows which asset classes drive the most impact.
  • Factor-level detailFor each risk factor: the applied shock, your delta/gamma/vega exposure, and the resulting P&L components (delta impact, gamma impact, vega impact).
  • Impact drivers chartA ranked view of which risk factors contribute the most to total P&L impact, making it easy to identify the dominant risk drivers.

You can also use the configurable pivot table to slice and dice results across dimensions like portfolio, scenario, risk factor, and category.

Comparing Results

StressGen supports two types of comparison:

  • Within a stress test — Compare different scenario-portfolio pairs within the same run to understand how the same portfolio responds to different scenarios, or how different portfolios respond to the same scenario.
  • Across stress tests — Compare items from different runs to see how results change when you update portfolio snapshots or use different shock methodologies.

When comparing two items, the platform decomposes the P&L difference into two attribution components:

  • Shock contributionHow much of the difference is due to different shock values (holding sensitivity constant).
  • Sensitivity contributionHow much is due to different portfolio exposures (holding shocks constant).

Tips

Meaningful Comparisons
The most insightful comparisons use the same portfolio snapshot across different scenarios (to isolate scenario impact) or the same scenario across different portfolio snapshots (to see how risk profile changes affect P&L). Changing both at once makes attribution harder to interpret.