snowflake-labs Historical Backtest
When to Activate
Trigger when user asks: "backtest this portfolio", "historical performance of 60/40 AAPL/MSFT", "test portfolio over 3 years", "run a backtest"
Workflow
Step 1: Parse Portfolio Weights
Extract weights from natural language:
- "60% AAPL, 40% MSFT" -> {"AAPL": 0.60, "MSFT": 0.40}
- "Equal weight tech stocks" -> equal split across mentioned tickers
- "60/40 equity/bond split" -> map to appropriate tickers/ETFs
- Validate: weights must sum to 1.0, no negative weights
Step 2: Determine Parameters
| Parameter | Default | User Override |
|---|---|---|
| Benchmark | S&P 500 (SPY) | User-specified |
| Date range | 3 years | "over 5 years", "since 2020" |
| Rebalance frequency | Monthly | "quarterly", "annual" |
| Initial investment | $1,000,000 | User-specified |
Step 3: Execute Backtest
Tool: run_backtest
Pass parsed weights, benchmark, date range, and rebalance frequency.
Step 4: Present Results
Summary Metrics:
| Metric | Portfolio | Benchmark | Difference |
|---|---|---|---|
| Cumulative Return | |||
| Annualised Return | |||
| Annualised Volatility | |||
| Sharpe Ratio | |||
| Max Drawdown | |||
| Calmar Ratio |
Use data_to_chart for:
- Cumulative return chart (portfolio vs benchmark)
- Rolling 12-month return
- Drawdown chart
IPS Comparison Workflow
If comparing two candidates (e.g., for IPS compliance):
- Backtest Candidate A
- Backtest Candidate B
- Present side-by-side comparison table
- Follow with Monte Carlo for forward-looking validation
Stopping Points
- After Step 1 (weights parsed): confirm portfolio composition and parameters with user before running backtest
- After Step 3 (results presented): pause for user questions before proceeding to IPS comparison or Monte Carlo
Output
A backtest summary table with key metrics (annualised return, volatility, Sharpe, max drawdown, VaR, CVaR) plus methodology note.