You found look-alikes. Now test the rule.
A search here answers “when did the market look like this, and what followed?” RLXBT answers the next question: does a rule built on it survive a backtest, walk-forward and Monte Carlo — after costs?
- 1
What you found here
Dated historical analogues of a candle pattern and the spread of what happened next. That is a description of the past, not a forecast, and not a tested rule — twenty look-alikes can share one regime, and none of it counts fees.
- 2
What RLXBT checks
Research Autopilot takes one OHLCV CSV, profiles the market, screens a bounded set of simple hypotheses with trading costs, and pushes the survivors through walk-forward and Monte Carlo. Weak, overfit and cost-sensitive ideas are killed before anything is ranked.
- 3
An honest verdict
A run can end in NO EDGE FOUND or OVERFIT, and that is a useful result: it stops you trading a pattern that only looked good on a chart. A rule that survives comes with the evidence behind it — not a promise about the future.
Free · in your browser · nothing to install
On the chart, run a search, then use Download CSV for RLXBT under the matches. Autopilot accepts 500–10,000 bars (up to 2.5 MB) with the header timestamp,open,high,low,close,volume.
Want to write the exact rule yourself, or let an agent run the loop? RLXBT also ships a desktop app (macOS, Apple Silicon), an MCP server for Claude / Cursor, and a server build — see everything on rlxbt.com ↗
RLXBT is a separate product from the same team. Nothing here is investment advice.