RESEARCH SETUP
Choose your experiment
Ready. No run has been performed.
Run your own research
This page contains bundled measurements. Start the local app to change costs, controls or input data.
python -m alpha_factory_v1.demos.finance_alphaSetup & methodology →
YOUR EVIDENCE WILL APPEAR HERE
A result you can inspect.
Run the default scenario, then try Gap and recovery. A profitable synthetic run is not a forecast; the loss scenario makes the limits of risk controls visible.
Measured paper outcome
Equity through the experiment
StrategyEqual-weight buy-and-holdCash
Same starting capital, evaluation dates and costs. The reference buys once and holds; it does not use the strategy’s risk exits. Equity includes open positions.
Where the result came from
| Measure | Strategy | Buy-and-hold |
|---|
Decision trail
Signals use information available before execution. Estimated tail risk is based on proposed weights and preceding close-to-close returns.
| Signal / execution | Decision | Target exposure | Estimated CVaR95 |
|---|
Fills, final positions & reproducibility
| Date / symbol | Side / quantity | Fill price | Fee |
|---|
| Symbol | Quantity | Cost basis | Last close |
|---|
Run configuration
Data fingerprint · SHA-256
Engine fingerprint · SHA-256
Hashes identify bytes; they do not authenticate the data supplier. The JSON contains prices, configuration, every fill and every equity mark for exact Python replay.
UNDERSTAND THE BOUNDARIES
Transparent assumptions.
Reviewable decisions.
Prices are evidence inputs
Synthetic examples are labeled. Imported CSV is unverified: review provenance, licensing, splits, dividends and survivorship. Missing bars and invalid values fail visibly.
Costs are part of P&L
Fees reduce cash; adverse slippage changes fills. Realized plus unrealized P&L must equal equity minus initial cash. No borrowing, short sales or hidden leverage.
A limit is not a guarantee
Historical tail risk does not bound tomorrow’s loss. Stops act at the next supplied open. Fixed full fills omit liquidity, market impact, tax and financing.
Calculation conventions
- Signals use prior closes; fractional fills occur at the following open.
- No parameter fitting, automatic optimization or model inference.
- Exploring strategies repeatedly on the same sample is not independent validation.