snowflake-labs Monte Carlo Simulation
When to Activate
Trigger when user asks: "run Monte Carlo", "probability of reaching $X", "forward-looking risk", "simulation for retirement", "what are the odds of..."
Workflow
Step 1: Parse Parameters
| Parameter | Default | User Override |
|---|---|---|
| Number of paths | 10,000 | "run 50,000 simulations" |
| Horizon | 5 years (or from IPS) | "10-year horizon" |
| Initial value | From portfolio or $1M | User-specified |
| Target value | None | "reaching $2M" |
| Contributions | None (lump sum) | "adding $5,000/month" |
| Confidence level | 95% | "99% VaR" |
Step 2: Execute Simulation
Tool: run_monte_carlo
Pass portfolio weights, horizon, paths, and optional DCA parameters.
Step 3: Present Results
Probabilistic Outcomes:
| Percentile | Ending Value | Annualised Return |
|---|---|---|
| 5th (worst case) | $X | X.X% |
| 25th | $X | X.X% |
| 50th (median) | $X | X.X% |
| 75th | $X | X.X% |
| 95th (best case) | $X | X.X% |
Risk Metrics:
- Value at Risk (95%): $X loss over [horizon]
- Probability of loss: X%
- Probability of reaching target: X% (if target specified)
- Expected shortfall (CVaR 95%): $X
Use data_to_chart for:
- Distribution of ending values (histogram)
- Fan chart showing percentile bands over time
DCA (Dollar Cost Averaging) Support
When user specifies regular contributions:
- Model monthly/quarterly additions at specified amounts
- Show how DCA reduces sequence-of-returns risk
- Compare to lump-sum scenario if helpful
Stopping Points
- After Step 1 (parameters confirmed): verify portfolio, horizon, and any DCA settings with user before running simulation
- After Step 3 (results presented): pause for user questions or parameter adjustments
Output
Terminal value distribution table (5th/25th/50th/75th/95th percentiles), probability metrics (loss, double, triple), and methodology note.