Debug and replay a Backtest

Full replay is available only when the plan, Backtest configuration, and completed run provide the required debug data. Use it to answer a precise question about a tested version: for example, “Did Condition receive RSI and select Match on this bar?” It is an inspection tool, not a way to edit the strategy or rerun the Backtest.

Open replay

Open the debug or replay action from an eligible Backtest result. The Flow Editor enters a read-only replay view. The Replay Dock replaces editing panels; you cannot change or save the graph there.

A static result overlay can still be available when full replay is not. It summarizes recorded activity and is useful for orientation, but cannot substitute for a missing event-by-event debug payload. If the result says replay is unavailable, check the eligibility explanation before submitting another run with debug options; the old run cannot gain events it did not retain.

Control the timeline

Use reset, previous, next, play, pause, timeline position, speed, and follow controls. Each accepted timeline entry represents a recorded debug event; its timestamp is a point in time, not a unique event identifier. Several events can share a timestamp. Use previous/next to inspect entries, not only the time slider. A Breakpoint can pause replay when the run contains its matching debug event.

Keyboard focus must remain visible as you move between timeline and detail controls.

Try it here

Explore an illustrative replay

Inspect an available debug payload in the read-only Replay Dock and follow events, data, variables, and diagnostics.

Exercise steps

1. Move forward and backward through the fixed illustrative sequence.
2. Inspect the current node and captured values. This read-only trace does not execute your edits.

Start lesson

Inspect details

Use the available tabs for Event, Data, Variables, Margin, Summary, and Diagnostics. The exact content depends on what the Backtest recorded. Per-event data and variable snapshots are optional, so an empty tab is not evidence that the graph never used that category.

For an alert that you expected but did not see, work through one evaluation:

1. Locate the Event Trigger entry for the instrument and period in question. If none was recorded, check the Backtest date range, market-data coverage, trigger settings, and debug range.
2. Follow the highlighted control path to Condition. Compare the active input value with its operator and right-hand source. With RSI 58 and a literal threshold 60, > selects No Match; with RSI 62, it selects Match.
3. If Match occurs, follow that outgoing control edge to Alert. If no Alert event appears, inspect the relevant diagnostics and active path rather than changing the text of the Alert as a fix for the condition.
4. Distinguish “the debug payload has no snapshot for this event” from “the node had no value.” Use other recorded evidence and validation before concluding that an input was absent.

For an order path, Order's Submitted outcome records submission, not a fill. Inspect the later Await Order Event outcome when asking whether the order was accepted, filled, canceled, or rejected. Do not infer a fill from the Order node alone.

Return to editing

Choose Exit editor to leave replay and return to the Backtest detail. To edit, open Strategies, select the strategy and choose Edit in editor. This opens its current version, not the historical graph recorded by the Backtest. Make changes there, save, and submit another Backtest if you need a new comparison. Keep the original run ID and tested version in view when comparing: a later edit never changes what the earlier run tested.