From hypothesis to live capital.
One pipeline, one engine. What you research is exactly what trades — nothing translated, nothing rebuilt, nothing lost on the way.
Any signal you can write.
Direct, programmable access to the raw options data — IV, greeks, term structure, skew, minute bars to 2012. Write signals in the built-in DSL, compose them in the visual builder, or describe them in plain English and let the AI draft the code. No menu. No ceiling.
Does it predict anything?
Before you risk a backtest on it, interrogate the signal itself. Time-series, distribution, scatter, and decile plots against forward returns. The same exploratory rigor a desk quant would demand, one click in.
Combine signals. Survive the future.
Blend signals into regression models validated walk-forward — trained on the past, scored on data they've never seen. Fold tables show you exactly where the edge holds and where it decays. Memorizing history doesn't count.
Broker-real, or it doesn't count.
Minute-resolution simulation: market-accurate fill, impact, and slippage models, stress-aware stops, commissions, and margin calls under both Reg-T and portfolio margin. The engine's job is to disappoint you now instead of your broker doing it later.
Build the book, not just the trade.
Blend validated backtests into one allocation-weighted portfolio: short premium beside long vol, index flows beside earnings. The engine nets greeks across sleeves and computes true cross-margin (Reg-T or portfolio) on the combined book, so diversification pays where it's real: lower margin, shallower drawdowns.
Go live. Then hold it to the numbers.
Snapshot a strategy, or the whole book, into a paper or live account. Swaps queue safely and every action lands in an append-only ledger. We resample the backtest into a distribution of expected outcomes, then test live results against it as trades fill. When realized performance drifts outside that envelope, you get an early stop alert, so you catch a decaying edge before your P&L does.
An analyst on the desk, not a chatbot.
Describe the idea; it drafts the DSL. Ask why the drawdown clusters; it runs sandboxed analysis on your actual results and shows its work. Every claim links back to data you can open.
Your best idea deserves a real test.
Join the waitlist, bring the strategy your current tool can't express.