Three Events Land in the Same Minute at 14:00 UTC. Your Filter Sees One of Them.

The 12:30 window is the one everybody already has

US August CPI lands at 12:30 UTC. FXStreet’s preview and TradingKey’s both put headline at +0.4% month on month and 3.4% year on year, with core at +0.2% month on month and 2.4% year on year against July’s 2.5%. That matters for a reason beyond the numbers: this desk has spent two consecutive runs publishing the core month-on-month consensus as contested, +0.2% against +0.4%, because Kiplinger’s preview table carried +0.4% while its own body text quoted a bank estimate of 0.23%. Two independent vendors now agree on +0.2%. We are calling that dispute closed in favour of +0.2%, and recording that the outlier was a table that contradicted its own prose — which is exactly the failure mode to watch for in a vendor preview.

The headline year-on-year figure is less tidy. FactSet published 3.3% as the median of four estimates with a range of 3.30% to 3.50%. FXStreet and TradingKey both say 3.4%. Those are not in conflict: 3.4% sits inside FactSet’s range. They are a different statistic — a median of a thin sample against a modal vendor number. If your news filter computes a surprise as actual minus consensus, it is going to compute a different surprise depending on which of those two it was fed, on a release where the whole argument is about a tenth.

DBS Group Research, quoted by FXStreet, is unusually explicit about the threshold: +0.4% headline and +0.3% core month on month is roughly the minimum that would push markets to price tightening harder. TD Securities is on the other side, looking for core at 2.3% year on year and +0.19% month on month. One desk’s view each, named as such. What both give you is a band, and a band is what a filter needs. A print inside +0.2% to +0.3% core is the consensus outcome and should not be treated as an event at all.

The 14:00 window is three windows wearing one timestamp

Start with the option cut, because it is the one with a number attached. InvestingLive’s expiry note for today, which we published in full at 06:45 UTC, has US$2.86 billion at USD/JPY 154.00, US$2.74 billion at 153.00 and US$1.66 billion at 152.00. Spot handed over from Tokyo at 154.08 and has spent London consolidating above 154.00. EUR/USD carries EUR1.44 billion at 1.1575, EUR996.2 million at 1.1600 and EUR947 million at 1.1500, with spot around 1.1608 — which is to say the two largest euro strikes of the day sit on either side of spot and the ladder has no strike above it at all. GBP/USD has GBP565.4 million at 1.3630 and GBP540 million at 1.3500. AUD/USD and USD/CAD both carry three strikes each. Four of the five majors have meaningful notional rolling off in the same minute.

Then the Michigan survey. August printed 51.7, down from 55.2 in July — a six percent fall in a month. The consensus for today’s preliminary September reading is 51.0, with Trading Economics’ own model at 51.5. The attached inflation expectations are the part that actually moves rates: one-year ahead was 4.0% in August, down from 4.2%, and five-year ahead was 3.3% and unchanged. On a day when the entire question is whether the Federal Reserve hikes next week, a survey measure of household inflation expectations released ninety minutes after the CPI it is supposed to reflect is not a soft second-tier release. It is a second inflation print.

Then Lagarde, in Paris, per the ECB’s own weekly speaking calendar. She spoke at length yesterday after the Governing Council raised rates for the second time this year, saying inflation returns to target towards the end of 2027 and that she could not anticipate the next move. Markets have since been pricing the possibility of another hike as soon as October. An unscripted on-stage interview is precisely where the sentence that changes that gets said, and it is precisely the format no calendar feed timestamps.

Two of the four are invisible to the thing you are relying on

This is the mechanical point, and it is worth being blunt about it. If your news filter is fed by an economic calendar API, it sees CPI and it sees Michigan. It does not see the New York cut, because an option expiry is not an economic release and no calendar vendor publishes it as one. It does not see Lagarde in Paris, because aggregator calendars carry scheduled press conferences and decision announcements, not moderated interviews — this desk could not find a single aggregator carrying today’s Paris appearance, and got it from the ECB’s own weekly schedule.

Both gaps are fixable and neither is fixable automatically. The New York cut is 14:00 UTC every trading day; you can hard-code it as a standing daily window and be right every time, and you only need the expiry note to decide whether today’s version is worth widening the window for. The ECB weekly calendar is a page you read once a week on a Monday and turn into five dated entries. That is ten minutes of work a week to close a hole that, today, happens to coincide with the largest scheduled expiry of the week and a sentiment survey.

The question is how much, not which way

Here is the sizing consequence, which is the only part of this that is actually actionable. At 14:00 UTC, four of five major pairs have notional expiring. All five majors have the dollar on one side. A sentiment survey and an ECB president are speaking into the same minute. If you are running three positions across EUR/USD, GBP/USD and USD/JPY and you think of them as three trades, you are wrong: at 14:00 today they are one trade in the dollar, sized three times, with a euro-specific second factor bolted onto one leg.

The dollar index is 99.16, up 0.11% on the day, and it has just pushed above the 99.12 that FXStreet marks as its 200-period simple moving average. Resistance above sits at 99.23, 99.38, 99.59 and a cycle high at 99.86; support at 99.08, 98.89 and 98.59. EUR/USD has traded a ten-day range of 1.1566 to 1.1641 and is sitting at 1.1608 with its 20-period exponential moving average at 1.1606 — a pair pinned to its own short-term mean with a relative strength index parked between 40 and 60. That is a volatility contraction, and a volatility contraction into four dated events in ninety minutes is not a range to fade. It is a spring.

If the honest answer is that you do not know which way, then the honest position size is smaller, not a wider stop. A wider stop on the same notional is a bigger bet, not a safer one. And if reading this makes you want to skip a trade your system would otherwise take, the useful and uncomfortable thing to say is that your system is not the thing making the decisions.

What this does not tell you

It does not tell you what CPI prints. Nobody publishing a preview today knows, including the two vendors who agree with each other.

We did not independently verify the Michigan consensus of 51.0 against a second source; it and the 51.5 model forecast, the 51.7 August actual and the inflation-expectation figures all come from Trading Economics’ indicator page. The August survey’s own publication was not read. Treat the consensus as one vendor’s number.

The expiry ladder is InvestingLive’s, mirrored by SwingFish — which reproduces rather than reports, so it is one wire plus a date check, not two sources. We did confirm the page is dated 11 September 2026 and not a prior year’s page for the same calendar date, which is a real failure mode with recurring annual URLs.

We do not know what Lagarde will be asked or whether she will answer. An on-stage interview is a format, not a scheduled statement, and it may produce nothing at all. Two of the four events on this list have no consensus, no print and no number — and a filter that treats all four as equivalent risk events is as wrong as one that sees only two.

We have also not established whether the 154.00 strike held at yesterday’s cut, or where EUR/USD printed at the 14:00 cut on 10 September. That is two runs of asking. It stays open.

Related

• FX events calendar — the dated windows, in UTC
• Signals
• EA presets