Hedge AI Labs
Methodology & Data
How every analysis and backtest on this platform is computed — the assumptions, the data, and what we refuse to do.
Fill model
- No lookahead. A signal computed on bar i fills at the open of bar i+1 — never on information the strategy couldn't have had.
- Conservative intrabar resolution. If a stop and a profit target are both touched inside the same bar, the stop is assumed to fill.
- One position at a time, no pyramiding, entry signals during an open position are ignored.
- Costs are on by default: commission per side plus slippage (percent of price for stocks, points for futures — e.g. 0.25 pt on ES). Every report states the costs it used.
Data sources
- Stocks & ETFs — split-adjusted OHLCV bars (1-minute to daily), history to roughly 2016.
- Futures — continuous contracts for ES, NQ and RTY; daily history back to ~2002, intraday to ~2016.
- Options — real traded option prices (aggregates of actual trades), including expired contracts. We never reconstruct option prices from Black-Scholes and present them as history.
Honesty rules the analyst must follow
- No silent tuning. If parameters were varied, the report says so and shows the untuned or out-of-sample number as well. A tuned in-sample result presented alone is treated as misinformation.
- Sample size is always reported. Under ~30 trades, results are explicitly flagged as noise-prone.
- Drawdown is mark-to-market — measured bar by bar including open positions, not just at trade exits.
- Long-only strategies are compared to buy-and-hold on the same data. Beating zero is not a result.
- Negative findings are published. "This has no edge" is a valid answer, and you will get it when it's true.
- Every number comes from executed code. The analyst may only report figures its own scripts printed during your session — the code and its output are shown in the chat.
Privacy
Your strategies, uploads, chats and results are private to your account. Each chat runs in an isolated
workspace; other users cannot see or access it. Uploaded files stay on our servers solely to run your analysis.
We don't sell or share your research.
Limits
- Analysis runs are compute-capped; very large studies are split into steps.
- Data entitlements are listed above — the analyst will tell you plainly when something you ask for isn't covered rather than substituting lookalike data.
Disclaimer
Hedge AI Labs is a research and education tool. Nothing on this platform is investment advice, an offer,
or a recommendation to buy or sell any security or derivative. Simulated and backtested results are hypothetical, do not reflect actual trading, and no representation is made that any account will achieve similar results. Past performance does not predict future results. Trading involves substantial
risk of loss. See also the Terms of Service and Privacy Policy.