Getting started

Begin with a small strategy whose decisions you can explain. The SJS getting-started example provides a complete crossover strategy for this workflow.

Create or import an experiment

Open Agent Lab in your workspace. Choose New experiment, or select a saved app from the import list and choose Import app.

Importing copies the saved app into a new experiment. It does not copy a running session's positions or keep the experiment synchronized with later app edits.

Set up the first run

1. Open Settings and give the experiment a recognizable name.
2. Select one instrument and a short historical period with available data. The start and end inputs are labeled UTC.
3. Choose the interval, capital, shares or contracts, and fill delay. Review timezone, regular-hours filtering, and end-of-day closing.
4. Open SJS and add the example source. Keep advanced layers out of the first experiment unless they are part of what you are testing.
5. Choose Save revision, then Run backtest.

Run backtest also saves pending draft changes before starting. Saving explicitly gives you a clear checkpoint before execution.

Inspect the baseline

Open History, select the completed run, and examine Chart, Trades, and Logs. Confirm the data covers the expected period and that the first entry and exit match the source rules. A completed run can have no trades; completion alone does not establish strategy correctness.

Choose Use as baseline on the run. You now have a reference for the next change.

Make one improvement

Edit the strategy yourself or connect an assistant. Ask for one specific change, run it, and compare the results with your baseline.

Use SJS for the main strategy logic. The L3 guide explains when advanced coordination is useful; SJS does not currently expose the policies module.