There is no 12:30 UTC block today. The pre-US slot’s defining event — four American numbers landing in the same minute, half an hour before the New York open — simply does not happen on 18 September. The first US release of the day is at 13:15, which is fifteen minutes after the open, not thirty minutes before it. And one of the two calendars we read has every single one of today’s six US times an hour early, labelled “ET”. We settled three of them at the issuing institution. Each took one fetch.
Today’s actual American schedule, in UTC
Three things, and a fourth for the oil desk.
- 13:15 — Federal Reserve G.17: industrial production, capacity utilisation and manufacturing output for August. Consensus +0.3% on the month for industrial production against a +0.2% prior, and 76.4% capacity utilisation against 76.3%.
- 13:30 — Vice Chair for Supervision Michelle Bowman, in London, on stress testing.
- 14:00 — The Conference Board US Leading Economic Index for August. Consensus +0.1% against a +0.2% July, which took the index to 99.5.
- 17:00 — Baker Hughes rig counts, 450 oil and 591 total prior.
Note what the 14:00 entry collides with. That is the New York option cut, and this desk published yesterday that a release landing on the cut is a different risk from a release landing anywhere else, because the two flows are simultaneous and you cannot attribute the move to either. It has happened again, on consecutive days, with a different print. We flagged it once as a scheduling accident. Twice in two days is a property of this week’s calendar and should be in your rules rather than in your notes.
One calendar has all six of those times an hour early
A widely-read financial portal published a preview of today’s US calendar this morning. It lists industrial production at 8:15 a.m. ET, the leading index at 9:00 a.m. ET, Bowman at 8:30 a.m. ET and the rig counts at 12:00 p.m. ET. Every one of those is exactly sixty minutes earlier than the schedule above, and the offset is perfectly consistent across all six entries. A consistent offset is not a typo. It is a clock.
Three of the six settle at the institution that issues them, and the settlement is not ambiguous. The Federal Reserve’s own G.17 page states that monthly releases are issued at 9:15 a.m., and names 18 September 2026 as the release date. The Federal Reserve’s events calendar puts the Bowman event at 9:30 a.m. The Conference Board’s leading-indicators page gives its next release as, in its own words, Friday, September 18, 2026, at 10 A.M. ET. All three are 9:15, 9:30 and 10:00 Eastern — 13:15, 13:30 and 14:00 UTC — and all three are an hour later than the portal has them.
Here is the part we are obliged to say, and it is the part that makes this worth an article rather than a complaint. We cannot tell you whether that portal’s page is wrong or whether our reading of it localised the times to a different zone. Those are two completely different faults with two completely different remedies, and from outside the page we cannot separate them. That uncertainty is not a weakness in the argument. It is the argument.
Why an hour is worse than a wrong number
If a calendar gives you the wrong consensus, you trade a surprise you did not expect and you find out immediately. If a calendar gives you the wrong time, your news filter suspends trading from 12:00 to 12:30, reopens, and is fully exposed and fully sized at 13:15 when the print actually lands. The failure is silent, it looks exactly like a normal day, and the post-mortem shows an unexplained adverse fill against a release your log says you were flat for.
Worse, it inverts the protection. A window built an hour early does not merely fail to cover the event; it forces you flat during the quietest half hour of the session and then hands you full exposure into the loudest one. You pay the cost of the filter and receive none of the benefit. We would rather have no news filter than one keyed to a clock we have not checked.
The remedy costs about sixty seconds a day and we have now done it twice this week. For anything on a fixed institutional cadence — a statistical release, a central-bank speech, a scheduled publication — the issuer publishes its own time, and the issuer cannot be an hour out about its own release. Aggregators are for discovering that an event exists. They are not for timing it.
Bowman is not the event your filter thinks she is
A second-order point, and the reason the two questions are not the same. The Federal Reserve’s calendar describes the 13:30 event as a chat with the Vice Chair for Supervision in London on the subject of stress testing. That is a bank-supervision appearance, two days after an FOMC meeting, by the governor whose portfolio is regulation.
A filter keyed to the string “Fed speaker” will suspend you for it at high impact. A filter keyed to “monetary policy” will ignore it entirely. Neither is obviously right — a Vice Chair can always say something about rates — but they are different decisions and your system is currently making one of them by accident, based on which vendor tagged the row. The honest version is to keep a small window and size down rather than flatten, and to be aware that the base rate of a stress-testing chat moving USD/JPY is low.
There is also a second Fed speaker today. One calendar has him, the other does not, and the calendar that has him is the one with the demonstrated hour offset. We are therefore not publishing a time for him at all. A name without a reliable timestamp is not a calendar entry; it is a rumour with a job title.
What this does not tell you
We verified three of the six times at the issuing institution, not all six. The rig counts and the two capacity-related lines inherit their timing from the G.17 release and the Baker Hughes schedule respectively, and we did not check Baker Hughes directly; we are relying on the offset being consistent, which is an inference rather than a reading.
We have not named the portal in the body, because we cannot separate a page fault from a rendering fault and it would be unfair to publish the stronger of those two claims when we can only support the weaker one. The source is listed in the footer and you can check it yourself against the three institutional pages, which is the whole point.
And we are not telling you that today is quiet. The Bank of Japan moved the yen close to two hundred pips before London opened, and a day whose largest event has already happened is not a day with no risk — it is a day where your remaining scheduled windows are all low-grade and the unscheduled risk is elevated. Industrial production has not moved USD/JPY in months. That is precisely the environment in which a wrongly-timed filter goes unnoticed for weeks.
Related
- FX events calendar — our schedule, in UTC, with the issuer named.
- Signals — how event windows change our framing.
- EA presets — news-filter and window configuration.
