Investing.com published its preview of today’s US session with nine timed rows in it. Every single one is exactly sixty minutes earlier than the time this desk has on the same event. Not roughly an hour. Sixty minutes, nine times, all in the same direction. If your news filter takes its schedule from that page, it will hold you flat through a quiet hour this morning and put you live in the market for the five releases that land in one minute at 12:30 UTC.
Nine rows, one sign, sixty minutes
Here is the comparison, with our UTC schedule on the right. ADP nonfarm employment change: 7:15 a.m. ET on the preview, 12:15 UTC on ours. Final second-quarter GDP: 7:30 a.m. ET, against 12:30. Core PCE: 7:30 a.m. ET, against 12:30. Chicago PMI: 8:45 a.m. ET, against 13:45. EIA crude inventories: 9:30 a.m. ET, against 14:30. Barkin: 12:30 p.m. ET, against 17:30. Cook: 2:25 p.m. ET, against 19:25. Goolsbee: 4:10 p.m. ET, against 21:10. Kashkari: 5:00 p.m. ET, against 22:00.
Nine rows. Eight distinct clock times. The offset is sixty minutes on every one of them and it never changes sign. That is not a data-entry problem, because data entry produces scatter. A constant offset with no variance is a property of the page, not of the events on it.
We checked one of these against a third publisher rather than trusting our own conversion. Trading Economics, whose indicator pages render in UTC, gives the next US personal spending release as 2026-09-30 12:30. The preview gives 7:30 a.m. ET, which during US daylight time is 11:30 UTC. One hour apart, and the third reader lands on ours.
This is the third same-signed instance this desk has recorded of this exact behaviour on this exact publisher, and it is the first time we have had nine of them in a single document.
What a sixty-minute error costs, in minutes you are not allowed to trade
Take the ordinary rule — flat for fifteen minutes either side of a scheduled release — and lay it over today’s real schedule. The blocks merge into six: 09:45 to 10:15, 11:45 to 12:45, 13:30 to 15:15, 17:15 to 17:45, 19:10 to 19:40, and 23:35 to 00:05. That is 285 minutes, four hours and forty-five minutes of the day.
The part that matters is the middle. Between 09:45 and 15:15 — the London morning, the New York open, and the overlap where the day’s liquidity actually is — there are 330 minutes, and 195 of them are inside a window. Fifty-nine per cent. And 105 of those are one unbroken block from 13:30 to 15:15, because Chicago PMI, the option cut, the EIA number and the quarter-end fix are stacked close enough that fifteen-minute buffers fuse them into a single closure.
Now shift the whole schedule back an hour, which is what the preview would have you do. Your 11:15 to 11:45 block covers nothing. Your 11:45 to 12:45 block — the one containing German CPI, ADP and the five-release minute — does not exist, because your filter thinks those have already happened. You would be flat for half an hour of silence and fully exposed through the densest sixty minutes of the month.
That is the whole argument for reading a calendar’s timestamps as data that can be wrong, rather than as configuration. A directional call being wrong costs you a trade. A timestamp being wrong costs you the trade you did not know you were in.
The window nobody can timestamp
There is a worse case on today’s schedule than an hour-shifted row, and it is the one at 10:00 UTC.
The Ministry of Finance publishes Japan’s monthly foreign exchange intervention total today. This desk went to the Ministry yesterday and read both the page describing the disclosure regime and the monthly data index. Neither states a day of the month and neither states a time of day. The period is derivable from the primary source and our published period is right. The time is not derivable from anything the Ministry puts out. It rests on one aggregator saying 19:00 JST, and on the fact that it has been roughly that before.
So the thirty minutes from 09:45 to 10:15 are blocked on a convention. And that is arguably the single most price-relevant number of the day for anyone holding yen risk, because it settles whether September’s defence of the yen was money or words.
The mechanical answer is not to guess harder. An event whose time you cannot source does not belong in a time-keyed filter at all, because a time-keyed filter fails silently in both directions: it protects you at the wrong minute and leaves you unprotected at the right one. It belongs in a state-keyed rule — a spread or short-horizon volatility threshold that arms itself when the market moves, not when the clock ticks — or in a sizing decision that simply carries less yen risk across the whole morning. Those cost you something real. Guessing costs you something you will not be able to attribute afterwards.
Two entries on today’s clock that are not releases at all
The 13:30 to 15:15 block has two items in it that no economic calendar will show you, and both belong in your schedule today.
The first is 14:00 UTC, the 10 a.m. New York option cut. InvestingLive’s expiry note this morning flags USD/JPY at 157.00 and EUR/USD at 1.1300, and describes 157.00 as sitting less than fifteen pips from spot at the time it was written, shortly after 04:00 UTC, with a comment that it could pin through the European session. We are not publishing a notional against either strike: the size table on that page is inside an image that does not extract, which is a standing limitation we have recorded before and are not going to paper over with a number we did not read.
The second is 15:00 UTC. London is on British Summer Time, so the 16:00 London fix is at 15:00 UTC, and today is the last business day of September and of the quarter. Gaitame’s morning note calls out position adjustment and settlement flow from Japanese real-money accounts as the day’s standing feature. A fix is not a news event and it has no consensus, but it is a scheduled, recurring liquidity distortion with a known time, which is precisely the kind of thing a time-keyed filter is actually good for. If you have blocked the EIA number and not the quarter-end fix, you have the priority backwards.
What this does not tell you
It does not settle why that publisher’s times are an hour early. The obvious explanation is a page rendering fixed-offset standard time while the United States is still on daylight time, and nine same-signed instances in one document is strong. It is not proof. The experiment that discriminates is still the one we committed to in print: the United States leaves daylight time on 1 November, and one fetch of the same preview on 2 November tells you whether the offset vanishes, doubles, or stays. Until then this is a consistent observation with a plausible cause, and we are not upgrading it.
It does not tell you that our UTC times are right and theirs are wrong. It tells you that two independent readers agree with each other and a third does not, on every row we could check. That is what we have.
It does not tell you where spot is now. The expiry note’s description of 157.00 as within fifteen pips was written shortly after 04:00 UTC and the Tokyo session low we have on the tape was 157.23 at 00:03 UTC. Those two are not reconcilable without a price we did not read, so we are reporting both with their timestamps instead of splitting the difference.
And it does not tell you to trade any of the windows above. The arithmetic here is about when you are exposed, not about which way.
