Today is the last business day of October, so the 16:00 London fix carries month-end rebalancing. The 10am New York option cut is at 14:00 UTC. Those two events are two hours apart today. Until last Friday they were one hour apart, and from Monday they will be one hour apart again — because London left summer time on 25 October and New York leaves it on 1 November, and for the seven days in between the two most mechanical liquidity events in the trading day sit at an unusual distance from each other. The one month-end of the year that falls inside that gap is this one. If your cut window and your fix window are stored as wall-clock times, one of them is wrong this week and the other is wrong next week.
Two hours today, one hour on Monday, and the arithmetic that does it
The 10am New York cut is a New York wall-clock event. At eastern daylight time, minus four hours, it is 14:00 UTC. At eastern standard time, minus five, it is 15:00 UTC. The 4pm London fix is a London wall-clock event: at British summer time it is 15:00 UTC, and at Greenwich mean time it is 16:00 UTC.
So the gap between them in 2026 goes like this. Through most of the year, cut at 14:00 and fix at 15:00, one hour. From 25 October, when the United Kingdom returned to standard time, the fix moved to 16:00 while the cut stayed at 14:00 — two hours. From 2 November, when the United States returns to standard time, the cut moves to 15:00 and the gap closes to one hour again.
The dates behind that are published. A Japanese broker notice issued at 15:30 Japan time today gives both calendars in a table: American summer time ran 8 March to 1 November and standard time runs 1 November to 14 March 2027; British summer time ran 29 March to 25 October and standard time runs to 28 March 2027. Two further readers, an almanac and a news carry, give the American dates identically. The arrangement follows from those dates and it is our arithmetic, not anyone’s published finding: in 2026 the two offsets are misaligned for roughly three weeks in March, between the 8th and the 29th, and for one week now, between 25 October and 1 November. We are in the shorter window, and the month-end sits inside it.
The practical form of this is dull and it costs money anyway. A window that opens at 13:45 and closes at 14:15 UTC covers today’s cut and finishes an hour and three quarters before the month-end fix. The same window on Monday covers nothing at all, because the cut has moved out from under it. Neither failure announces itself. Both look like a quiet window.
We have to correct our own sentence from Wednesday
On 28 October this desk wrote, in print, that “the 10am New York cut is 14:00 UTC every day and the Fed announces at 18:00 every time.” We used it to argue that a structural finding about expiries dying before their events was, for three of four currencies, a convention restated rather than a discovery.
The argument stands. The sentence does not. Both halves of it stop being true on Monday: the cut becomes 15:00 UTC, and the next Federal Open Market Committee decision, on 9 December, announces at 19:00 UTC rather than 18:00. What we wrote was true for the week we wrote it in and we stated it as a property of the market.
This is a smaller error than publishing a wrong number and a more annoying one, because it is the kind that propagates. A reader who took “14:00 UTC every day” into a configuration file has a correct file until Sunday night. The honest version of that sentence is that the cut is 10:00 in New York every day, and what that is in UTC depends on the date — which is longer, duller, and the only form of it that survives a clock change.
The board we could read, and the board that was nine months old
Today’s expiry board came to us through a syndication mirror, because the original would not open. The mirror, dated 30 October and crediting the original author, describes the arrangement without a single notional: euro–dollar expiries layered between 1.1150 and 1.1200 with the larger figure-level strikes nearer 1.1175, one dollar–yen strike at 158.00 described as roughly forty pips from spot, and one Australian dollar strike at 0.7025. That is the fourth consecutive occasion on which this mirror has carried the levels and dropped the sizes, which is exactly what our source notes say it does. We will not guess at the notionals and we are not going to publish a board without them as though it were one.
The more useful thing we found was a failure. The same expiry publisher maintains an undated address for “today’s NY cut.” It returned a page bylined 29 January 2026 at 07:17 UTC, carrying euro–dollar strikes at 1.1815 and 1.1850 and dollar–yen at 150.00, 152.50 and 153.75 — with notionals, in text, cleanly extractable — wrapped in a sidebar of headlines dated 29 and 30 October 2026. A January board in an October frame.
Now put that next to yesterday’s finding, because the pair makes an argument neither makes alone. Yesterday the same publisher’s dated address for 30 October served a board headlined 28 October whose nine strikes across four pairs matched, to the million, the totals this desk had published the previous day. That board was undetectable from its contents. It was internally consistent, it sat plausibly around spot, and the only reason we caught it was that we happened to have printed the same numbers ourselves twenty-four hours earlier.
Today’s stale board is harmless precisely because it is absurd. Dollar–yen at 150.00 against a spot of 157 is caught by any sanity check you care to write. The one that is nine months out of date cannot hurt you; the one that is one day out of date is the one that gets into your book. Which means a plausibility filter on the levels is not a defence. The only field that separates the two cases is the byline date, and on the dangerous version the byline date is the only thing that is wrong.
A published distance is a timestamped quantity
The mirror says the 158.00 strike is “roughly forty pips” from the spot price. That was written in the European morning, when the pair was around 157.60. By 10:10 UTC one Japanese publisher had the pair back in the 157.10s and 157.20s in European trade, which puts the same strike about eighty pips away.
The strike did not move. The board did not change. The number that changed is the one the publisher wrote down, and it halved in value as a description of the market inside four hours. If your expiry handler stores a distance rather than a level, it is storing a quantity that decays, and it will tell you an expiry is in reach for hours after it has stopped being in reach, or the reverse.
Where spot actually is today is less settled than it should be. One publisher’s running coverage had the pair at 157.52 at 11:43 Japan time with the Bank of Japan statement still pending and down to 157.12 by 12:28 — the figures this desk published this morning. A second publisher, writing at 19:10 Japan time, gives a trading range of 157.16 to 157.58 for the period up to then and does not say when that period starts. Those two conflict by four pips on the low unless the second publisher’s window excludes the Tokyo morning, which it may well do and does not state. Four pips is not an interesting disagreement in itself. It is interesting because it is a range quoted without a start time, which is the same class of fault as a distance quoted without a timestamp.
On the euro side, the same 19:10 publisher gives 1.1189 to 1.1201 and an earlier note has the pair briefly above the 1.12 handle. That puts the described strike cluster at 1.1175 just under the market and the 1.1200 strike effectively on it — which is a change in posture. On 27 and 28 October every euro strike on the board sat above the market, the second day with the board more than doubling in size while spot fell further beneath it. Today the cluster is at or below spot for the first time this week. A rule that reads “large expiry nearby, expect a pin” has had three different geometries to work with in four sessions, which is a reasonable argument for not having that rule.
We have no size for today’s fix, and we are not going to invent one
The obvious thing to want on a month-end Friday is a bank’s rebalancing estimate — the standard-deviation signal that says which way hedging flow leans into the fix. We went looking for it and what we found was a trap worth naming.
A note titled “month-end fixing to be mildly supportive of the dollar” is live at a stable address and reads exactly like today’s. It is dated 28 August 2026, and it is about the August month-end: the model pointing to outflows from sterling at minus 1.3 standard deviations and the euro at minus 0.6, into dollars, with a very light expected impact. A second site carries the same August note from a different wire a day earlier. Neither has anything to say about 30 October, and the per-currency table in the original is an image that does not extract in any case.
So we publish no month-end signal today. Not a direction, not a magnitude, not a hedged version of either. The reason is the one from yesterday restated: a document whose address and framing do not carry a date will read as current forever, and the cost of using August’s rebalancing estimate as October’s is not a small error but an uncorrelated one.
What we can say about the window itself is structural. Today’s fix at 16:00 UTC sits on top of two central-bank speakers — Daly of the San Francisco Fed at 16:00 on two readers, Lane of the European Central Bank at 16:10 on one — with Musalem of the St. Louis Fed at 16:30. A month-end fix is the single largest scheduled flow event of the month, and this one has a Federal Reserve president starting to speak in the same minute. If you were going to stand down for one window today, that is the window, and it is not the one at 14:00.
What this does not tell you
It does not tell you the notional on a single strike. The only board we could read gives none, the original publisher’s page returned an error, and the dated address at the same publisher that would have carried it also failed. Everything above about the board is levels and arrangement.
It does not tell you which way the fix goes, and nothing in this note should be read as implying that a two-hour gap between the cut and the fix changes anyone’s direction. The claim is only that the two windows are where they are today and somewhere else on Monday.
We did not establish a Tokyo session high or low. No publisher we read gave a clean pair, so the figures above are point quotes at named times and one range with no stated start. A pre-decision forecast range from one house was 157.000 to 158.800 for the day, and the same house puts the 200-day moving average at about 158.56 in a sentence whose parenthesis is ambiguous about whether that is the average or the price at writing — two pips from the 158.54 this desk published yesterday on two readers, which is within the noise of how such a figure gets quoted and is not a reconciliation.
The clock arithmetic is ours. The transition dates are three-sourced and the consequences for retail session hours are from one broker’s own notice, which is authoritative for that broker and nobody else. We did not read the fix administrator’s or any exchange’s own publication on this, and the one-hour and two-hour gaps above are derived from published transition dates rather than quoted from anyone.
Related
- Economic calendar and event windows — the cut is at 14:00 UTC today and 15:00 UTC from Monday; the London fix is 16:00 UTC in both weeks. Our Systems note this slot covers what the same change does to the American release file.
- Signals
- EA presets
