Review Backtest vs Alerts with AI

Basic or Professional AI credits required

AI Insights summarizes a completed Backtest vs Alerts comparison. It can suggest investigations; it cannot establish the cause of a mismatch or certify a strategy.

Before opening AI Insights

• Run the comparison so a report payload exists.
• Check source counts, timezone, and matching settings manually.
• Configure an AI model in Account Settings.
• Confirm the account has AI credits.

The control is disabled when the plan is below Basic, no comparison exists, or AI configuration is unavailable. Its tooltip shows the current reason.

What the request contains

The generated request includes the current symbol, strategy name, chart timeframe, date range, alert source, CSV timezone, matching tolerances, aggregate summary, generated rule-based insights, and a bounded sample of up to 18 problem rows.

It does not send every row from an arbitrarily large comparison. Reproduce any AI claim against the original table and source export.

Generate and review

1. Select Analyze with AI. Opening the drawer starts generation when valid context and credentials are available.

2. Read the response beside the recorded model, token usage, response ID, and any reasoning trail exposed by the drawer.

3. Convert each suggested cause into a check: identify the exact raw rows, source timestamp, strategy setting, or webhook record that would confirm or reject it.

4. Use Regenerate only after changing the evidence or question. A second answer to unchanged inputs is not independent validation.

5. Use the editor handoff when a code investigation is justified. The handoff saves report context when possible and opens Algo Lab with a new chat and script picker.

If the snapshot cap prevents a new saved report, the app can still open Algo Lab with temporary in-memory context. Delete an old snapshot or change plan before assuming that temporary context was preserved for later use.

Verify before editing code

A missing or extra alert can come from data scope, timezone, position-direction normalization, matching tolerance, or strategy behavior. Do not change Pine code until source records rule out the data and configuration explanations.

If generation fails

1. Read the drawer error and disabled reason.
2. Confirm the analysis still has report context after any scope change.
3. Check AI configuration and credit balance in Account Settings.
4. Reduce the analysis window if the useful evidence can be represented by a smaller comparison.
5. Retry once; if the same error repeats, continue with the manual results workflow.

Continue

• Investigate the results manually
• Review matching configuration
• Use AI Pine editing
• Manage AI credits and models