The catalog audits itself

Buys momentum when QQQ breaks its opening range and rides the first hour's direction on SPY — two thin, separately validated edges on one book, with hard stops and out by close. Skips VWAP-misaligned breaks and macro-event mornings.

validated-thin-edge: small measured edge in honest point-in-time testing. Thin means single-digit basis points per trade and plenty of losing days — discipline is the product, not excitement.

1739 backtest trades over 4 years · net +3.38 bps/trade after 1.5bps round-trip costs

Owns the index on calm days, steps aside when the regime detector reads rough seas. One decision a day at most. The goal is a smoother ride, not beating the market — our live testing's only surviving lane.

validated-risk-tool: proven to reduce drawdown, not to beat buy-and-hold. In live lane testing it matched market-level returns with roughly half the worst-case drawdown.

20 backtest trades over 4 years · net +247.68 bps/trade after 1.5bps round-trip costs

If the market's first hour sets a direction, hold that direction to the close. One entry a day, flat by 15:55. The pattern replicates on both SPY and QQQ, which is part of why we trust it.

validated-thin-edge: +2 to +4 bps per trade net of costs in honest testing, in-sample and out-of-sample both positive.

1001 backtest trades over 4 years · net +2.48 bps/trade after 1.5bps round-trip costs

Fades stretches away from VWAP back toward fair value. Textbook logic, sound risk control — and no edge. Our own testing shows intraday mean reversion on liquid ETFs loses after costs. It stays in the catalog as a learning book and as proof the audit has no favorites.

structure-only: sound structure, no proven edge. If you run it, run it to learn execution discipline, not to make money.

771 backtest trades over 4 years · net -2.52 bps/trade after 1.5bps round-trip costs

How a strategy gets built here

1. You describe what you want in plain English - capital, time you can give, the daily loss that would make you quit.
2. The machine composes from these blocks only. It cannot invent logic; every parameter is bounded by what was actually tested.
3. The audit gates it. Point-in-time backtest, honest costs, Deflated Sharpe, bootstrap - a fatal finding blocks deployment, ours included.
4. You steer on regime. A daily briefing tells you how rough the sea is (our most-validated signal); you decide full size, half, or stand down - or delegate to autopilot or the community pool.
5. Paper first, always. It runs on your Alpaca paper account. Going live means exporting a single-file app you run yourself - your keys never touch us.

The conversational builder is in final testing. Join the founding list to get first access and locked pricing.