Scientific comparison
Backtest laboratory
Run every baseline strategy with the same dollars over the same window of stored real prices, then compare what each would have produced.
No real history
Signals use completed sessions; rebalances fill on the next session's open with 5 bps slippage.
Loading real-history coverage…
First composite run on real history is preserved
EXPERIMENT A
NVDA buy-and-hold
—
EXPERIMENT B
Equal-weight AI basket
—
EXPERIMENT C
Relative-value rotation
—
EXPERIMENT D
Momentum rotation
—
EXPERIMENT E
Composite AI baseline
—
Risk-adjusted comparison
Hypothetical results on stored real prices · never a guarantee or forecast
| Experiment | Strategy | Start | End value | Profit | Total return | CAGR | Volatility | Max drawdown | Sharpe | Sortino | Calmar | Trades | Baseline |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
No real market history is stored yet. | |||||||||||||
Saved run history
Every completed run on real history is persisted with its manifest and parameters
| Started | Run | Basket | Lab | Strategy | Period | Start | End value | Total return | Drawdown | Sharpe | Trades | Baseline |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Run a comparison to persist the first results. | ||||||||||||
Walk-forward status
Train 12 months → evaluate untouched 3 months → roll forward. Scheduled for Phase 6 after real point-in-time fundamental coverage is validated.
Needs 15+ months of real history
Robustness status
Threshold, slippage, timing, and factor-weight perturbations remain separate from the preserved composite baseline.
Baseline required first