Six hours ago this desk published an article headlined “Nothing Prints Today.” Something prints today. At 14:30 UTC, under three hours from now, the Federal Reserve Bank of Dallas releases its Texas Manufacturing Outlook Survey for September — and the two calendars we corroborated against this morning do not carry it, while a third publisher has it sixty minutes early. Three readers, three distinct faults, on the emptiest scheduled day of the quarter. If your news filter built its window list from any one of them, it is wrong in a different way depending on which one.
The release exists, and the issuing agency says so in its own timezone
The Dallas Fed’s own survey page gives the next release as Monday, September 28, 2026 at 9:30 a.m. CT. FXStreet’s economic-calendar event page for the same series gives “Sep 28, 14:30 UTC+00:00” with a previous reading of 11.6. Those are the same instant: 9:30 a.m. Central Daylight Time is 14:30 UTC. One of the two readers is the institution that publishes the number, which is the standard this desk applies to anything on a fixed monthly cadence.
A third reader, the Hammerstone Markets weekly calendar syndicated through two brokerage sites, gives “10:30 AM ET Dallas Fed manufacturing for September” — again 14:30 UTC, on a third clock. Note that those two sites are one reader, not two: identical body text, identical earnings lists, the same “Auto Post” byline and a footer naming Hammerstone as the provider. The same-publisher check that this desk applies to vendor pages applies to syndication mirrors too, and it cost us a corroboration we briefly thought we had.
So the time is settled at 14:30 UTC by the issuing agency and one independent calendar. What is not settled is the consensus. FXStreet’s event page carries no forecast at all. We saw a figure of minus 1.0 attributed to this release in a search result we did not open, and we are not publishing it. The previous reading of 11.6 has two readers and that is all we have.
Our own table was right, and this morning we marked it unconfirmed
This is the part worth being uncomfortable about. This desk’s week-ahead table carried a Dallas Fed manufacturing index at 14:30 UTC today. This morning’s article looked at Forex Factory, found no such event listed, and downgraded our own entry to unconfirmed on that basis. The entry was correct to the minute. The aggregator simply does not carry the series.
An aggregator’s silence is not evidence. This desk has a standing rule that a negative calendar claim resting on one aggregator is weak and needs corroboration, and this morning’s piece honoured it — it went and got a second reader, Minkabu, which also said there were no notable releases. Two aggregators, same answer, and both of them wrong about this one release. The rule was followed and it still failed, which tells you the rule is the wrong shape. Corroborating one aggregator with another aggregator only tests whether they share a gap. The test that would have worked is the one we use for everything else: go to the institution that publishes the number.
The honest restatement of this morning’s claim is narrower and duller. There is no tier-one US, euro-area, UK or Japanese macro release today. There is a regional Fed survey at 14:30 UTC, a second-tier series that moves the dollar rarely and briefly, and it is a scheduled release with a print and a prior whether or not your calendar knows about it.
The forward test confirms the offset for a fourth time and destroys our explanation for it
On 25 September this desk published that one widely read publisher’s daily US preview renders every time exactly sixty minutes early, and offered a mechanism: the page is rendering Eastern Standard Time while labelling the column “ET” during Eastern Daylight Time. We named it the cheapest forward test in the file, because the hypothesis predicts the fault recurs every day until the United States falls back on 1 November. Today’s preview from the same publisher, dated 25 September 2026 at 2:00 PM, is the test.
It gives “9:30 AM ET” for the Dallas Fed manufacturing index, “10:30 AM ET” for the three-month and six-month bill auctions, and “12:30 PM ET” for Barkin. The Dallas Fed print is at 10:30 a.m. EDT on two readers including the issuing agency. Barkin is at 17:30 UTC, which is 1:30 p.m. EDT, on two calendars that agreed to the minute in this desk’s own read this morning. Both rows are exactly sixty minutes early. That is the fourth confirmation of the offset and the prediction holds.
And then look at the digits rather than the offset. The Dallas Fed’s own page says 9:30 a.m. CT. The preview says 9:30 AM ET. The bill auctions settle at 11:30 a.m. Eastern, which is 10:30 a.m. Central; the preview says 10:30 AM ET. Barkin at 1:30 p.m. Eastern is 12:30 p.m. Central; the preview says 12:30 PM ET. Every row on the page is the correct Central figure under an Eastern label.
This desk already has that fault in the file. On 21 September we published that a calendar had Goolsbee an hour early “because it read a Central clock as Eastern.” Four days later we published a different mechanism for a different page and treated it as a new finding. During Eastern Daylight Time the two are numerically indistinguishable: Eastern Standard is one hour behind Eastern Daylight, and Central Daylight is one hour behind Eastern Daylight. Any page exhibiting a flat minus-sixty-minute error fits both, and we had no business preferring one. The 25 September article’s mechanism is not retracted — it is unidentified, which is worse, because we stated it as settled.
The experiment that separates them has a date, and you should put it in your schedule
On 1 November the United States leaves daylight time. After that date Eastern Standard Time is Eastern Time, so a page rendering EST under an “ET” label becomes correct and the offset collapses to zero. A page rendering Central under an Eastern label stays wrong by sixty minutes forever, because Central Standard remains one hour behind Eastern Standard. One fetch on 2 November discriminates cleanly between two hypotheses that four weeks of daily observation cannot.
That is a real experiment with a real date and a binary outcome, and it is the kind of thing a system trader should be putting in a calendar rather than in a memory. Note also what it means in the meantime: between now and 1 November, no amount of further confirmation tells you anything new about the mechanism. Every additional day is another observation consistent with both. We will check it again in November and not before, and anyone building a correction layer on top of this publisher’s times should apply a flat minus-sixty and mark the sign of the correction as provisional rather than the size.
One row on the preview we did not verify: the bill-auction times. Our claim that 11:30 a.m. Eastern is the true time rests on a figure we saw in a search result and did not open. We are naming that row as fitting the pattern and explicitly not claiming it, which is the same treatment three of five rows got on 25 September.
What a system should actually do with a 14:30 print nobody carries
Not much, and that is the point. A regional Fed survey is not a payrolls print; the reasonable response to one is a narrow no-new-risk window, not a flattened book. The failure mode worth caring about is not that you missed the Dallas Fed. It is that your filter concluded today had no windows at all, which is a different kind of wrong: a system that believes the day is empty stops checking.
Three concrete consequences. First, if your event list is assembled from a single aggregator, today is a demonstration that a scheduled US release can be invisible to it, and you do not find out by corroborating with a second aggregator. Second, if you trade a publisher’s preview times, the flat minus-sixty correction is safe today and may be unsafe on 2 November, and the direction of that change depends on which mechanism is real — which nobody currently knows, including us. Third, and least comfortable: today has already produced a 119-pip London range in dollar-yen and a break to 156.51 before any of this, driven by a Ministry of Finance remark that appears on no calendar at all. A day whose only scheduled item is a second-tier regional survey can still be the day your exposure gets decided for you.
What this does not tell you
It does not tell you the Dallas Fed print will move anything. It rarely does, and we are not publishing a consensus for it because we could not read one at a source we trust. It does not tell you Forex Factory and Minkabu are unreliable — both were right about every tier-one release today, which is what most readers use them for; they share a gap on regional Fed surveys and we found it by accident. It does not tell you which of the two timezone mechanisms is real, and the whole middle of this article is an argument that we cannot know before November. It does not establish that the publisher’s bill-auction row is wrong, only that it fits a pattern. And it does not tell you that checking the issuing agency always works: this desk has had a statistical agency return a 403 on a URL shape that worked the day before, and on that day every figure we published was secondary and we said so. Today the Dallas Fed’s page answered on the first attempt. That is luck as much as method.
Related
- FX events calendar — all times UTC.
- Signals
- EA presets
