Collect a dataset

Collect runs your Definition over historical bars and saves every emitted setup as an event in a dataset.

That dataset is the raw material for studies — not downloaded candles, but your setup occurrences.

Prerequisites

• A Definition with working preview in the Definition Builder

• A saved version snapshot

• Plan headroom for event counts — check Settings → Subscription

Start a collect

1. Open Research → Definitions → [your definition]

2. Click Collect or Collect your first events

3. Choose:

• Instrument — ES, NQ, and so on

• Date range — how far back

• Timeframe — must be allowed by the script’s runOn rules

• Profile if you use profiles

1. Confirm and start

Progress appears on the definition detail. Large ranges take time.

While it runs

You can navigate away. Cancel from the collect card if you picked the wrong range.

When complete, the dataset shows an event count. Open Browsing events to spot-check.

What a dataset is

A list of timestamps and payloads — extra fields your Definition attaches to each event (direction, level, indicator values at emit).

Studies read payloads as columns.

What good looks like

Six months of 5-minute ES events, count in the thousands, spot-check shows emits on real breaks not noise. Ready for a study.

Important distinctions

Research Collect = events from a Definition (setup detector).

Playbook Collecting Data = your trades during a session for Historical Intelligence — not Definition events.

Dataset = emitted events. Not a CSV of OHLC you import from elsewhere.

Next

1. Browsing events — inspect payloads

2. Create a study — ask questions over the dataset

3. Conductor overview — agent-driven collect