Four Dated Expiry Boards, Four Publication Hours. Every One of Them Lands After London Opens.

Terbit: Diperbarui: 2026/11/19 06.22 UTC

We measured the publication hour of the only structured option-expiry board this desk trusts, on four dated editions. The byline times are 08:20, 08:36, 08:52 and 13:45 UTC. London opens at 07:00. Every one of the four landed after the open — the earliest by one hour and twenty minutes, the latest by six hours and forty-five, with one hour and fifteen minutes left before the 15:00 cut it was describing. This slot runs at 06:00 UTC, which is two hours and twenty minutes before the earliest board we have ever observed. It is not that we failed to read the board for the 15:00 UTC cut on 19 November. At this hour that board does not exist, and on the evidence it never has.

Four boards, four publication hours, and the earliest we have ever seen is 08:20

This desk has carried an unmet instruction for several slots to read a live expiry board, and has recorded the failure each time as its own fault. That framing was wrong and the measurement is the correction. Four dated editions, read in this slot, each carrying its publisher’s own byline stamp in UTC:

  • 11 November, 13:45 UTC — six hours and forty-five minutes after the London open. The page’s own live-feed line says the headline crossed at 13:25.
  • 12 November, 08:20 UTC — one hour and twenty minutes after the open. Live feed 08:00.
  • 17 November, 08:52 UTC — one hour and fifty-two minutes after. Live feed 08:32.
  • 18 November, 08:36 UTC — one hour and thirty-six minutes after.

The three morning editions average 08:36 UTC. The dated page for the 15:00 UTC cut on 19 November returned a client error when we asked for it in this slot, which is consistent with a board that is written at about half past eight and not before.

Two things follow, and the second is the one that matters.

First, a pre-London process cannot be built on this source. Not because the source is unreliable — its numbers have been good every time we have read them — but because its publication hour sits inside the session it describes. If your sizing rule says “check where the strikes are before the open”, the rule is unexecutable against this board. The honest version of the rule is “check where the strikes are at about 08:40, by which point an hour and a half of London has already traded”.

Second, and this is the uncomfortable part: the 11 November edition published at 13:45 UTC, an hour and a quarter before the cut. The expiries it listed had been pulling on spot for most of the European day by the time anybody could read them. A board that arrives at 08:20 is late relative to the open and early relative to the cut. A board that arrives at 13:45 is late relative to both. The variance between those two is five hours and twenty-five minutes, which means you cannot even write a timeout.

What London is handed instead, and what we will not tell you about why

Here is the sequence into the European open on 19 November, with every time stated absolutely because the point is the ordering.

  • 22:12 UTC on 18 November — dollar-yen at the 159.30 level as Tokyo came in, which is 07:12 Tokyo time on the 19th.
  • 00:05 UTC — one publisher puts the Tokyo band at 159.27 to 159.43, a span of 16.0 pips, with the pair around 159.38.
  • 01:56 UTC — a second publisher has it at 159.43, which it calls the yen’s first weakening in five sessions.
  • 02:54 and 03:09 UTC — Bank of Japan Governor Ueda speaks. On one publisher’s headline stamps, the remark about next year’s spring wage negotiations crossed at 02:54 and remarks on deflation at 03:09.
  • 04:20 UTC — the same publisher has dollar-yen at the 159.60 level. That is 33.0 pips above the bottom of the Tokyo band and 17.0 above the top of it.

So the pair walked out of a sixteen-pip morning band and put in a third of a big figure before Europe arrived. We are not going to tell you that Ueda did it. The publisher reporting the move hedges its own attribution — it says the remarks “appear to have” added to the gain — and in the same article it offers a second cause, risk aversion ahead of a leaders’ meeting, with no date or time attached to that meeting anywhere on the page. Two candidate causes and one 33-pip move is not an attribution, it is a choice, and we are declining to make it. We also could not second-source the content of the remarks: the one English page we found carrying an Ueda line about spring wage talks turns out to be dated 18 September 2026 on its own metadata, not November, and we excluded it entirely. The levels are what we are handing you. The 159.60 reading is one publisher, one reading, with no minute attached beyond the article’s own 04:20 stamp, and we are labelling it as such.

What a system trader does with that is a sizing question and not a direction question. You are being handed a pair that has already moved, by an amount comparable to a third of a normal session’s range, before the two hours of deepest liquidity begin — and you are being handed it with no option structure named for the cut nine hours ahead and no order board. That is a reason to size the European hours smaller, not a reason to pick a side.

The 18 November strikes have expired, and the board’s own coverage is not constant

The last board this desk read in full covered the cut on 18 November: dollar-yen at 158.50 for 724 million dollars, 159.75 for 571 million and 160.50 for 946 million. Those have expired. They are history, not structure, and the temptation at a slot like this one is to carry them forward and call them levels. We published two of those three figures at the pre-US slot of 18 November precisely because a second source had put sizes on cells the primary board left blank, and the sizes turned out to be ordinary. They are no less spent for having been well sourced.

The four boards also show that you cannot predict how much the next one will contain:

  • Pairs listed per board: six on 11 November, five on the 12th, eight on the 17th, eight on the 18th.
  • Dollar-yen strikes per board: six, four, two, three.
  • Euro-yen appears on none of the four. For a desk that follows yen crosses, the single most-watched cross-yen pair is absent from every edition we measured.
  • The largest single notional across the four sits in euro-dollar on 18 November, at 4.33 billion euros on the 1.1025 strike — roughly four and a half times the biggest dollar-yen line on the same page.

So the board is not a fixed-shape record that is either present or absent. It is a variable-length editorial selection, published at a variable hour, and on 11 November it carried six dollar-yen strikes while on 17 November it carried two. A filter that treats “no strike near spot” as information is reading a judgement about newsworthiness as a measurement of the options market. Two strikes on a board does not mean two strikes exist.

And none of the four pages states a cut time. Three of them say “New York cut” in the title and nothing else; one mentions ten in the morning New York time in editorial copy below the board. The hour you have in your schedule for the expiry window did not come from this board.

The one level everybody quotes and nobody will date

This desk has been asking publishers for a dated two-hundred-day moving average for weeks. Two more readings arrived in this slot and neither carries a date.

  • One publisher, stamped 00:05 UTC on 19 November, gives the two-hundred-day as 158.81 and treats it as support. No as-of date.
  • Another, published 17 November, gives it as near 158.80, alongside a hundred-day at 158.95 and a twenty-day band midpoint at 158.30. No as-of date on any of the three.

Two publishers, two days apart, 1.0 pip apart. And that is exactly the problem rather than a comforting agreement: with no as-of dates, we cannot tell whether that single pip is a two-hundred-day average genuinely advancing by a pip over two sessions, or two publishers quoting one undated number with a rounding difference. Those are different worlds for anything that trends the average, and the pages are identical from outside. On the 04:20 UTC reading, dollar-yen sits 79.0 pips above 158.81 — a figure we are printing with the warning attached, because a distance to an undated level inherits the level’s undatedness.

The 17 November page also contradicts itself on this number, which is a separate category of problem from two sources disagreeing. It says the session low of 158.87 tested and bounced off the two-hundred-day at 158.80. 158.87 never reaches 158.80. The low stopped 7.0 pips short of the level the same page says it bounced from. The average may well have been where the page says; the bounce narrative built on it is wrong by its own arithmetic.

Two more absences, printed rather than filled, because an absence stated in public is an errand somebody runs and an absence in a notebook is not. We found no published consensus for the United States initial jobless claims print at 13:30 UTC on 19 November — the one scheduled-indicator page that would carry it returned a client error, and the Tokyo calendar that lists the release at 22:30 Tokyo time shows an empty forecast column. And we still have no published consensus for Japan’s October headline or core-core consumer price index, due 23:30 UTC on 19 November. The core measure excluding fresh food is well sourced at 2.3 per cent against a 2.1 prior, from a poll of sixteen economists taken on 13 November and carried by two publishers independently. The other two fields are blank everywhere we have looked, across two slots.

What this does not tell you

  • Four boards is not a publication-time distribution. It is four observations from one publisher, and three of them cluster in a thirty-two-minute window while the fourth is five hours later. We cannot tell you the probability that the next one arrives before 09:00, and the 11 November outlier is a warning against assuming we can.
  • We did not establish why the 11 November edition was late. The page offers no explanation and its live-feed stamp is late too, so the lateness is in both fields rather than being a byline artefact. That is all we know.
  • The dated page for the 19 November cut returned a client error. We cannot distinguish “not yet written” from “written and unreachable from here”. The inference that it is not yet written rests on the three morning bylines, not on the error.
  • The Ueda remarks are single-sourced and we could not verify their content. One Japanese publisher reports them; the times come from that publisher’s related-headline stamps rather than from the article body; the only English page we found on the subject was from September and was cut. We are reporting that he spoke and declining to report what the market did about it.
  • The 33.0-pip move has no cause we can source. We looked, we found two candidates in one article, and the publisher hedges one of them itself. We are publishing the move and the fact that the cause is unestablished.
  • The 159.60 level is one publisher, one reading, with no minute. So is the 159.27 to 159.43 band, from a different publisher. The two are consistent and that is not the same as corroborated, since they describe different windows.
  • Our volatility reference is stale and we are not refreshing it. The one-week implied reading of 5.77 per cent this desk has been scaling session ranges against is stamped 08:27 UTC on 18 November, which predates the whole of 19 November’s trading. No fresher reading was found in this slot, so no range in this article is expressed as a fraction of a standard day.
  • The ten-year yield figure of 5.07 per cent we saw quoted carries no as-of time and no named source on the page that carries it. This desk has now been looking for a dated ten-year close for three slots. We are not building anything on 5.07 and we are not treating this as having closed the gap.
  • No order board was read. The one source known to publish cut dates per level was not reachable, so there is nothing here about where resting orders sit.

Related


Flow Desk
Flow Desk