How should I define a trading session that crosses midnight?
By Chart Reporter Reviewed 2 min read
The short answer
Define a session by its actual start, end and timezone. A session spanning midnight can include trades from two calendar dates, and Local versus Broker time can change how that window is displayed. Preserve those choices when comparing session results.
A session is a window, not just a name
Names such as evening or London are convenient shorthand, but they do not specify exact boundaries. Write down the start and end you intend and the timezone that gives those times meaning. A local session name can otherwise shift when daylight-saving rules or location change.
For example, a fictional 22:00-to-02:00 window crosses a calendar boundary. Selecting only the first date in a general daily filter may not express that intended session. Check the actual records included rather than assuming that every view interprets the label identically.
Inspect the timestamp basis
A trade can open in one window and close in another. A question about entry timing and a question about realised closing results may therefore require different grouping rules. Read the view's scope note and retain the basis in your review.
When comparing accounts, check their broker timezone settings separately. Use one hub when diagnosing boundary differences. A change in included records after switching Local and Broker time can be explained by calendar boundaries without implying that the recorded execution time was altered.
Make session comparisons reproducible
Keep account, currency, date coverage and strategy filters aligned across the sessions being compared. Record the count of included and eligible trades. If one session has a much smaller sample or incomplete capture, narrow the conclusion accordingly.
For cross-midnight work, a saved description of the window is more useful than a screenshot with an unexplained date label. If you revise session boundaries retrospectively, document the change before drawing conclusions from the revised totals. A named-session comparison can reveal an association in recorded history, but it does not establish that the clock window caused an outcome or will produce the same distribution in future conditions.
Sources and further reading
Related guides
When should I use a screenshot instead of trade replay?
Compare a captured chart image with reconstructed broker-candle context and understand what each can establish.
Why can a journaled trade have no replay candles?
Diagnose absent, insufficient and partial candle history separately from the trade record.