The AI strategy lab for momentum & swing traders

Test the setup you saw online — before you risk a dollar.

Describe any strategy in plain English. Our AI writes the exact rules, backtests 20 years in seconds, and tells you where it breaks. Then prove the survivors on a public, independently verifiable track record.

Run a backtest free →See the live records →

Free to start · no card · results in seconds

Golden Cross

2016-01-04 → 2026-07-30 · daily · real data, live crosshair

Total return

+709%

CAGR

21.87%

Max drawdown

31.0%

Trades

Templates · The starting point

Famous strategies, actually tested

Real numbers from real historical data — computed before you clicked, never cherry-picked after.

The backtest service isn't reachable right now — refresh in a moment.

The Lab

Build anything

Describe a strategy in plain English — any symbols, famous templates, or from scratch. The AI writes the rules, runs two decades of data in seconds, and teaches as it edits.

The Proof

Freeze it, forward-test it

One click deploys your strategy to the ledger. The spec freezes — hash-verified — and every trading day it's evaluated on fresh data. Timestamped. Append-only. Unfakeable.

The Board · Live now

Receipts, in public

The leaderboard ranks live forward performance, not backtested hindsight. Losers stay on the board. That's the point — a track record no one can rewrite.

The Strategist · An AI that tests, not guesses

Every opinion this AI has is a backtest

Other platforms sell an “AI that understands markets.” Ours is wired directly into a research engine — when it has an opinion, it runs the simulation and shows you the result, so every claim comes with a test you can inspect.

Speaks fluent strategy

Plain English in, exact executable rules out — entries, exits, stops, sizing. Nothing hidden behind an arrow: you can read, edit, and export every rule it writes.

Validated against decades

Every edit re-runs the full simulation — up to 5,000 US symbols, twenty years of history, intraday down to one-minute candles. The AI's suggestions are measured, not imagined.

Warns you first

It flags look-ahead bias, overtight stops, oversized positions, and too-few-trades significance before you spend on a doomed run. An AI that says your idea is flawed is worth more than one that says it's brilliant.

The strategy library was built the same way — every house strategy was drafted with the AI, refined through the engine run after run, and is now forward-testing on the public ledger. Browse the library →

Honesty · Printed on every result

What a backtest can't tell you

Past ≠ future

A strategy that worked for ten years can stop working tomorrow. Every result here is a historical simulation, never a forecast.

Survivorship bias is real

Full-universe tests use currently listed stocks — companies that went to zero aren't in the data, which flatters results. We label it when it applies.

Costs are assumptions

Every simulated fill includes a slippage assumption, shown with the run. Real trading costs vary — results are approximations, not receipts.

Questions, answered straight

How is this different from indicator platforms?

Indicator platforms sell black-box signals and pre-computed lookup tables. Here every strategy is a readable rule you can inspect, every “what if” actually re-runs the simulation, and every claim carries receipts.

Where does the data come from?

Daily US equity prices, split-adjusted, going back two decades, with a slippage assumption on every simulated fill. Full-universe tests note their survivorship bias right on the result.

Will this make me money?

No. This is a research and education tool. It shows how ideas would have performed historically — and history does not predict the future. Nothing here is financial advice or a recommendation to trade.

What makes the forward ledger trustworthy?

Deployed strategies are frozen — the spec's cryptographic hash is public, and the signals ledger is append-only at the database level. We could not rewrite a track record if we wanted to.

Can it predict what a stock will do?

No, and it won't pretend to. Ask it to predict prices and it politely refuses and offers a historical test instead. That's a feature.

Is your AI “trained on market data”?

We won't hide the ball: the strategist is a frontier AI model wired into our research engine — two decades of price history across 5,000+ US symbols. The intelligence you feel is that loop: it never hands you an unvalidated opinion, because everything it proposes gets simulated before you see a number. Platforms claiming their AI was “built to understand markets” are describing the same architecture with less honesty.

Curious what “buy the dip” actually returned?

Find out in about four seconds.

Test it now