Direct answer
A worked example of event filtering shows how you take a list of upcoming economic calendar items and narrow it to the ones you will treat as potentially relevant for a specific currency pair and time window. The example below uses made-up numbers to demonstrate the mechanics, and it states every assumption so you can independently check the logic.
Mechanism and definition
Event filtering, in an economic-calendar context, is a rules-based process that turns a raw set of scheduled releases into a smaller set of “events to monitor.” A typical input is a table with items such as event name, currency relevance (which economy it belongs to), scheduled time, and a subjective or provided impact level.
A key idea is separation of stable mechanics from variable conditions:
- Stable mechanics: you apply explicit rules to include/exclude events and then compute a simple statistic (for example, post-event move) for the included subset.
- Variable conditions: the true market impact, execution quality, transaction costs, and how quickly prices react can differ from your assumptions.
A concrete filtering rule set (example assumptions)
Assume you are monitoring a EUR/USD-like instrument.
Filtering rules you apply (assumptions you should write down):
- Currency relevance rule: include an event only if it is tagged to be about either the Euro area (EUR side) or the United States (USD side).
- Impact rule: include only events rated “high” impact by the event source.
- Time window rule: include only events scheduled within your chosen time window, for example 10:00–13:00 UTC.
- Uniqueness rule: if two events are scheduled within 5 minutes, you keep both (to avoid hiding overlap), but you will track them jointly.
Note: these rules define the procedure. They do not claim that included events will move prices, only that they are the events you will examine.
Worked example (numerical scenario)
Step 1: Start with an event list (made-up)
Suppose your calendar shows four scheduled releases during 10:00–13:00 UTC:
- A: Euro area “Inflation (high impact)”, 10:30 UTC
- B: United States “Retail Sales (high impact)”, 11:15 UTC
- C: United States “Conference Speech (medium impact)”, 12:10 UTC
- D: Euro area “Consumer Confidence (high impact)”, 12:12 UTC
Step 2: Apply the filter rules
Using the assumptions above:
- Rule 1 (currency relevance): A, B, and D are relevant; C is also about USD, so it is relevant by currency.
- Rule 2 (impact): keep A, B, D (C is medium impact, so exclude C).
- Rule 3 (time window): all kept events are within 10:00–13:00.
- Rule 4 (uniqueness): B at 11:15 and A at 10:30 are far apart; D at 12:12 is close to C at 12:10 but C is excluded, so you only track D alone in the filtered set.
Result: the filtered set is {A, B, D}.
Step 3: Define a measurable outcome (explicit assumption)
To keep this independently verifiable, define a simple outcome you will compute from price data you already have:
- Let post-event move = (price at end of a 30-minute interval after the scheduled time) − (price at the scheduled time).
Assume you collect mid-price values (again, made-up for illustration):
- Price at A time (10:30): 1.0800
- Price at A + 30 min: 1.0830 → move = +0.0030
- Price at B time (11:15): 1.0830
- Price at B + 30 min: 1.0785 → move = −0.0045
- Price at D time (12:12): 1.0790
- Price at D + 30 min: 1.0802 → move = +0.0012
Step 4: Summarize the filtered results
Compute the average post-event move across the filtered events (assumption: arithmetic mean):
- Average = (+0.0030 − 0.0045 + 0.0012) / 3 = −0.0001 (approximately)
This outcome says nothing about future performance. It only shows what happened in this one scenario under the stated assumptions.
What changed because of filtering?
Because you excluded C (medium impact) you did not include it in your post-event analysis. If you had included it, your average and conclusions could change. This is the practical value of event filtering: it makes your analysis focus on a consistent subset, but the choice of rules also determines what you can observe.