Find recurring price structures and make them testable.
Pattern Mining scans historical candles for price structures that repeat, measures how similar each new occurrence is to the reference shape, and promotes the strongest candidates into complete strategies with entries, exits and risk settings you can inspect and re-test.
Shape-based discovery Candidates are built from windows of price action, not from indicator formulas you must write.
Similarity you control The minimum-similarity threshold decides how strict a match must be before it counts.
Executable from day one Every result carries SL/TP, money management and limits, so it can be backtested and exported.
The current Pattern Mining surface captured automatically in a safe, empty-result state.
Configure the search
The configuration panel uses the same structure as the other miners:
- Data settings — source, symbols, timeframes and the date mode that defines the research window (a literal range, days back from the last vault candle, or days around an anchor date).
- Direction and window — BUY or SELL candidates and the number of bars that form the pattern window.
- Similarity and evidence — minimum similarity, minimum trades and the maximum number of candidates the job may keep.
- Execution assumptions — spread mode, slippage and signal shift, so matches are evaluated the way an exported bot would trade them.
- Risk grids — SL and TP ranges expressed in ATR multiples, with the ATR period as part of the contract.
- Money management, account limits, trade limits, trading windows and exit stops — identical semantics to the Strategy Lab, carried inside each candidate.
Optional market-regime constraints (trend, momentum, choppiness, Hurst, entropy and related filters) restrict matches to the market context that is part of your idea. Mine from Bank crosses the scan with reusable conditions from your banks. A Macro Channel can be attached so discovery and dates follow the structural channel of the market (see the Macro Channel page).
Run and review
Press Discover to launch the job. Progress is reported per combination, and Past Jobs reopens earlier runs. Results include, for each pattern:
- Symbol, timeframe, direction and generated name.
- Trades, win rate, risk/reward, net profit, profit factor and drawdown.
- The best-performing period and the SL/TP pair that produced it.
- The equity curve preview.
From the results table you can run the candidate on the configured broker, download its bot source, open it in the Strategy Lab with its full mining configuration, combine selected patterns, save it to the Strategy Library or send the whole set to a Board.
Patterns are sensitive to their window. When a candidate survives, re-test it on a different date window and on the target broker before trusting the shape — a pattern that only exists in one period is a coincidence, not an edge.
Recommended path
- Mine with a strict similarity threshold and a meaningful minimum number of trades.
- Open the strongest candidates in the Strategy Lab — the full configuration travels with them.
- Move survivors into a Workflow for filtering, robustness and walk-forward checks.
- Export and compare against the target platform before considering deployment.