Your first backtest

The whole loop, once, end to end: open the lab, pick or describe a strategy, refine it in chat, run it against history, and read the report. Your first backtest needs no account at all.

1. 1

   Open the lab

   Head to the lab— no sign-in needed. It has three zones: a left rail with strategy templates and a “New from scratch” button, a center workspace with four tabs, and the AI chat docked on the right. The toolbar on top holds the date range, the bar-timeframe selector, and the Run button — no usage counter anywhere, by design. Daily-bar templates run free without an account — up to 10 a day per visitor — with 3 free AI messages to edit them.

2. 2

   Pick a starting point

   Two ways in:

   • A template.The left rail lists ready-made strategies with their category and historical result — CAGR when a year-plus window backs it, the window's total return for short intraday tests. They are working examples to take apart, not recommendations. Click one and its rules load instantly.
   • New from scratch. The chat runs a guided intake — market, then timeframe, then entry, then exits and stops, then position sizing — and builds a live draft as you answer. It fills gaps with placeholder defaults and tells you which parts are placeholders, so nothing is hidden. The Rules tab shows the draft in plain English at any point.
3. 3

   Edit by chat

   Ask for changes in plain English — the more specific, the better: “tighten the stop from 8% to 5%” beats “make it safer.” Every accepted edit shows change pills listing exactly which fields changed and from what to what, and the spec is validated server-side before it can run. As you build, the AI flags known failure patterns — look-ahead bias, overtight stops, unrealistic sizing — as short teaching notes rather than blocks. Without an account you get 3 free messages a day — enough to feel an edit-and-rerun loop.

4. 4

   Run it

   Hit Run backtest. The default window is January 2016 to today on daily bars — about ten years of data, usually finished in seconds (the elapsed time appears next to the button). If you switch to an intraday timeframe, history is capped at roughly 60 days and the start date adjusts automatically.

5. 5

   Read the report

   Results land in the four tabs of the center workspace — toured below. Rerun after an edit and the tiles show deltas against the previous run, with the old equity curve kept on the chart for comparison.

6. 6

   Create a free account when you're ready to keep it

   Deploying to the public ledger, exporting the code, chatting past the free taste, and running custom or intraday strategies all take a free account — continue with Google or Discord, or enter your email for a sign-in link, no password to invent. Free accounts run unlimited backtests on daily data (a generous fair-use cap exists to stop scripts, not people), no card required; the Plans & limits page explains what each tier unlocks.

The four tabs

Results

Metric tiles — total return, CAGR, max drawdown, Sharpe, win rate, trade count, profit factor, expectancy — each with a plain-English tooltip, above the equity curve with drawdown charted beneath it. After a second run, each tile shows its change labeled better or worse, and the previous run’s curve stays overlaid in a second color. The next page covers every metric in depth.

Trades

Every round trip the strategy took: entry and exit dates and prices, profit or loss, and the exit reason. Click any row to open a candlestick inspector with the exact entry and exit candles marked — the fastest way to see whether a trade made sense or just got lucky.

Chart

The full price history for each traded symbol with every entry (▲) and exit (▼) overlaid. Switch symbols with the pills above the chart to see how the same rules behaved on different names.

Rules

The strategy in plain English — universe, entry, exits, sizing — written so you could trade it by hand, plus the exact spec JSON the engine ran (no black box). Download buttons export a runnable Python file, a Pine Script version for TradingView, or the raw spec.

After the run

Two buttons appear next to the tabs once a run exists: Share copies a link to this exact result (works without an account — anonymous shares carry a watermark), and Deploy to forward test freezes the strategy on the public ledger, where it accrues a track record on data nobody had when it was built. Deploying needs a free account — it puts a name on a public record. That step deserves its own page — Forward testing.