Trade the calendar, not just the chart.
Event Mining builds strategy candidates around scheduled economic-calendar events. Instead of scanning every bar, it focuses the search on the windows before and after the events you select — rate decisions, employment reports, inflation prints — and measures whether a repeatable reaction exists.
Select the events Filter the calendar by impact, currency and event name to define exactly what you are studying.
Shape the window Configure how many hours before and after each event the candidate is allowed to act.
Demand repetition Minimum move and minimum occurrences keep one-off reactions out of your results.
Before you mine: sync the calendar
Event Mining reads the market calendar stored in TR-GAB. Sync the calendar from Market Vault so the event database covers your research window; the job pre-counts how many events fall inside the selected dates so you know the sample size before running.
Configure the search
- Data settings — source, symbol, timeframe and date mode, shared with the other miners.
- Event selection — impact level, currencies and specific event names.
- Reaction window — hours before and hours after the event during which entries are evaluated.
- Evidence — minimum move and minimum number of occurrences for a candidate to qualify.
- Execution and risk — signal shift, SL/TP ATR grids, money management, account and trade limits, trading windows and exit stops, identical to the rest of the product.
Regime constraints, bank conditions and a Macro Channel can also be attached, exactly as in Rule and Pattern Mining. Output KPI filters run during the scan so weak candidates never reach the results table.
Review the results
Each candidate documents the event context it trades and the usual KPI set: trades, win rate, risk/reward, net profit, profit factor, drawdown and the equity preview. From the table you can open the event backtest detail, run the candidate on the configured broker, download its bot source, save it — individually or as a grouped strategy — open it in the Strategy Lab or send the results to a Board.
Event samples are naturally small: an event that happens monthly produces twelve occurrences per year. Prefer more occurrences over more optimization, and treat any candidate built on a handful of events as a hypothesis, not a system.
Recommended path
- Sync the calendar and confirm the pre-count covers enough occurrences.
- Mine a specific event family rather than the whole calendar at once.
- Inspect the per-event backtest detail of the strongest candidates.
- Validate in a Workflow and finish with a test on the target platform.