Named sessions
Sydney, Tokyo, London, New York and the London and New York overlap are each measured on their own exchange-local clock, so daylight saving moves them for you. Because they are a short fixed list and not a search, the correction runs across those five only. Each row gives local hours, which way the session leans, days measured, edge per bar, adjusted p and a grade. The edge is the session’s average entry result minus the average bar’s, so drift is removed.
Searched windows
A searched window is any run of consecutive slots, up to 12 hours by default, on any of twelve weekday sets. Searching that many windows is a larger family, so the correction is larger. A window has to clear the correction, hold across even and odd weeks, and pay after the Data tab’s costs before it is recommended, and a window whose sign is contradicted on the reserved bars is not.
Which to use
If you already hold a view that the build should only enter in the London session, measure that session and read the result. If you have no view, the searched windows are the way to look, and their adjusted p already counts every window tried. A session that is not significant is a finding too: the tab will show it as not detected, or as too few days.
An early window can look best
A window restricts when an entry may be taken. The trade is held for several bars, so the best entry window often sits a few bars before the move it catches. That is why the window is not a description of when the market moves.
What the Build task can hold
Only the time range is written. The task has no weekday filter beyond its weekend settings, so a result that depends on Tuesday to Thursday cannot be written as such. Exits are not touched, and exchange holidays and early closes are not modelled.