Event studies are most useful when the analog set respects the event type and separates the lead-up from the reaction window.
For an upcoming earnings release, the current path ends at today and historical paths end at their own pre-event close. Macro events use the same principle. This keeps information that arrived after the event out of the matching process.
Earnings should be compared with prior earnings for the same company when sufficient history exists. Macro releases should use the same event family. A fallback calendar may help recover dates, but the user-facing study should identify the event and its sample without exposing provider plumbing.
Solid lines before the anchor show what was known during matching. Dashed paths after it show what happened next. The consensus summarizes the selected horizon, while every analog remains visible so disagreement is not hidden.
Companies and event regimes change. Five prior releases can provide context but cannot support a sweeping claim. Read the event-relative base rate, median move, dispersion, worst case, and match quality together, and treat structural changes as a reason to discount older instances.