Portfolio stress test

Portfolio Stress Test

Purpose: Quantify portfolio-level downside exposure across multiple adverse scenarios to assess resilience and identify vulnerabilities.

Trigger: User asks "stress test my portfolio", "worst case scenario", "portfolio in a crash", or uses /portfolio-stress-test.

Data Flow

1. Load Portfolio

• Check <current_portfolio> first — skip tool call if present.
• Otherwise: load_portfolio(portfolio_id).
• Extract holdings list with tickers, shares, and current market values.

2. Fetch Live Quotes & Metadata

• For each ticker: call GLOBAL_QUOTE(ticker) to get current price.
• For each ticker: call OVERVIEW(ticker) to get sector, beta, market cap.
• Batch in parallel — single plan step for all fetches.

3. Define Stress Scenarios

Use these standard scenarios unless user specifies custom:

Scenario	Market (S&P)	Rates (10Y)	Sector Impact
Market Crash	-30%	-50bp	Tech -35%, Financials -40%, Defensive -20%
Rate Shock	-10%	+200bp	Growth -25%, Banks +5%, Utilities -15%
Recession	-20%	-100bp	Cyclicals -30%, Consumer Staples -10%, Healthcare -5%
Sector Rotation	+5%	+50bp	Growth -15%, Value +10%, Small Cap -20%
Inflation Spike	-15%	+150bp	Energy +15%, Materials +10%, Tech -20%

4. Run Stress Test for Each Holding

For each ticker, call scenario_stress_test(ticker, current_price, beta, sector, scenarios_json).

Input structure:

{
  "ticker": "AAPL",
  "current_price": 185.00,
  "shares": 50,
  "beta": 1.2,
  "sector": "Technology",
  "scenarios": [
    {"name": "Market Crash", "market_move": -0.30, "rate_change": -0.005},
    {"name": "Rate Shock", "market_move": -0.10, "rate_change": 0.02}
  ]
}

Output from tool:

{
  "Market Crash": {"price": 120.25, "value": 6012.50, "pnl": -3237.50, "pnl_pct": -35.0},
  "Rate Shock": {"price": 157.25, "value": 7862.50, "pnl": -1387.50, "pnl_pct": -15.0}
}

• Batch all scenario_stress_test() calls in parallel — single plan step.

5. Aggregate Portfolio-Level Results

For each scenario:

portfolio_pnl = sum(pnl for all holdings)
portfolio_pnl_pct = portfolio_pnl / total_portfolio_value × 100
new_portfolio_value = total_portfolio_value + portfolio_pnl

6. Render Output

Stress Test Table — use render_table():

Scenario	Portfolio P&L ($)	Portfolio P&L (%)	New Value	Worst Holding	Best Holding
Market Crash	-$6,500	-34.1%	$12,550	AAPL (-35%)	Cash (0%)
Rate Shock	-$2,800	-14.7%	$16,250	AAPL (-15%)	XOM (+5%)
Recession	-$4,200	-22.0%	$14,850	TSLA (-30%)	JNJ (-5%)
Sector Rotation	-$1,900	-10.0%	$17,150	AAPL (-15%)	BRK.B (+10%)
Inflation Spike	-$3,100	-16.3%	$15,950	AAPL (-20%)	XOM (+15%)

Vulnerability Analysis:

• Worst scenario: {scenario} → portfolio drops {X}% (${Y}).
• Best scenario: {scenario} → portfolio drops only {X}% (${Y}).
• Average stress impact: {X}% decline.
• Most vulnerable holding: {ticker} (average {X}% decline across scenarios).
• Most resilient holding: {ticker} (average {X}% change across scenarios).

Chart — use generate_chart():

• Type: bar
• Series: P&L % by scenario
• X-axis: Scenario name
• Y-axis: Portfolio P&L %
• Title: "Portfolio Stress Test — Scenario Impact"
• Colour: Red bars for negative, green for positive (if any).

7. Thesis Upsert — MANDATORY

This is an advanced skill → append stress test results to portfolio thesis. ⚠️ Extended thinking is discarded — copy the complete results into the thesis field; it is the ONLY output that reaches the Thesis panel.

Thesis structure:

## Portfolio Stress Test — {Date}

{Stress test table}

{Chart spec}

### Vulnerability Summary
- **Worst case**: {scenario} → portfolio value drops to ${X} ({Y}%).
- **Average stress**: Portfolio loses {Z}% across 5 scenarios.
- **Most vulnerable**: {ticker} ({sector}) — average {A}% decline.
- **Most resilient**: {ticker} ({sector}) — average {B}% impact.

**Recommendation**: {One-line guidance — e.g. "Reduce Technology allocation to limit downside in Market Crash scenario."}

8. Chat Response

State in chat field:

• Number of scenarios tested.
• Worst-case scenario and impact ($ and %).
• Average portfolio decline across scenarios.
• Most vulnerable holding.
• Tools used: load_portfolio, GLOBAL_QUOTE, OVERVIEW, scenario_stress_test (N times), render_table, generate_chart.

Cost Controls

• Skip load_portfolio() if <current_portfolio> is present.
• Batch all quote and metadata fetches in parallel — single plan step.
• Batch all scenario_stress_test() calls in parallel — one plan step for all holdings.
• Reuse session cache — don't refetch data already in <acquired_data>.

Error Handling

• If scenario_stress_test(ticker) fails, assume that holding declines by the market beta × market move for that scenario and continue.
• If sector data missing from OVERVIEW(), classify as "Unknown" and apply market beta-only stress (no sector overlay).
• If all stress tests fail, fall back to simple beta-based calculation: pnl = shares × price × beta × market_move.

Output Standards

• All dollar amounts formatted with commas: $1,234.56.
• Percentages with one decimal: -12.3%.
• Table must include all scenarios and a summary row (if applicable).
• Chart must show negative values as red bars, positive as green.
• Vulnerability analysis must name specific tickers and sectors.