Multi-Asset Regime Backtest¶

Walk-forward HMM-based regime allocation across multiple asset classes with stress testing.

Overview¶

Backtests a regime-aware allocation strategy across multiple assets, comparing it against buy-and-hold and 200-day SMA trend-following benchmarks. Uses walk-forward validation: train HMM on a fixed window, test on a rolling period, advance forward.

Strategy¶

Regime-based allocation scales exposure based on market volatility regime:

Regime	Allocation
Low Vol	95%
Medium Vol	77.5% (linear interpolation)
High Vol	60%
Uncertain	50%

Walk-Forward Backtest¶

for each window:
    train_days = train_years * 252
    test_days = test_months * 30
    train data → train HMM
    test data → forward-filter labels → compute allocation
    strategy return = allocation * daily_return
    advance window by train_days

Default: 1-year train window, 6-month test window.

Benchmarks¶

Three equity curves compared per asset:

Regime Strategy — allocation-weighted returns based on HMM regime
Buy & Hold — raw buy-and-hold returns
200-Day SMA — trend-following: only long when price > 200-day SMA

Metrics¶

Annualized return — total return annualized via (equity_end / equity_start)^(252/days) - 1
Sharpe ratio — annualized daily Sharpe
Maximum drawdown — peak-to-trough maximum
Sharpe improvement — strategy Sharpe minus buy-and-hold Sharpe

Stress Tests¶

Maximum drawdown during three crisis windows:

Crisis	Period
2008 Crisis	2008-09-01 to 2009-03-31
2020 Covid	2020-02-01 to 2020-04-30
2022 Rate Hikes	2022-01-01 to 2022-10-31

Visualizations¶

Equity Curves¶

Per-asset line chart: regime strategy (colored by asset), buy-and-hold (muted dashed), 200-day SMA (faint dotted).

Regime Timeline¶

All assets stacked vertically as horizontal colored bars. Each bar divided into regime segments (Low Vol / Medium Vol / High Vol). Hover shows regime name and date range.

Asset Comparison Table¶

Dark-styled table with all assets side by side. Sharpe improvement column colored green/red. Best-performing asset highlighted with cyan left border.

Stress Test Bar Charts¶

Grouped bar charts per crisis: strategy drawdown vs buy-and-hold drawdown per asset.

Usage¶

Add assets via sidebar (default: SPY, BTC-USD, GLD, TLT)
Set date range (default: last 5 years)
Set walk-forward parameters (train period in years, test period in months)
Click "Run Backtest"

URL Routing¶

Supports shareable links:

?selected=SPY&start_date=2019-01-01&end_date=2024-01-01&train_years=2&test_months=6 — auto-runs with SPY, custom dates and windows
Clicking an asset tab updates the selected param for sharing