At 06:22 UTC this morning this desk formally dropped an errand it had carried for weeks: find a dollar-yen 200-day moving average published with an as-of date. Four readings, none dated, so we closed it in print as unavailable rather than carry it a fifth time. Three hours and forty-four minutes later the market settled it for us. A Tokyo publisher timestamped 19:06 Japan time — 10:06 UTC — reported dollar-yen falling to around 158.74 and naming the level it had just broken: the 200-day moving average at 158.87. The low extended to 158.71 by 10:37 UTC. And the figure itself is the audit we said we could not run. We held 158.80 from 17 November and 158.81 twice since. Today three pages give 158.87. The average was advancing at about 2.33 pips a day, which means the single pip we could not interpret yesterday was real movement and not two publishers quoting one stale number. The errand was answerable. Dropping it was the wrong call, and the market did the work.
We closed an open item four hours before it closed itself
The reasoning this morning was defensible and still wrong. Four published values, none with a window or an as-of date, on an object that is by construction a function of its window — so we said we could not audit it and stopped asking. What that reasoning missed is that a series of undated readings is itself dateable, if the readings move. Here is the series we actually had:
- 17 November — 158.80, one publisher, no as-of date.
- 18 November — 158.81, no as-of date.
- 19 November — 158.81 and “near 158.80”, no as-of date.
- 20 November — 158.87, on three separate pages from one publisher, no as-of date.
Seven pips over three calendar days, 2.33 pips a day, 0.123 of a standard day in total. That is a monotone drift at a plausible rate for a 200-day window, and it is the discriminator we told you we did not have. A set of identical undated readings would have meant one number being copied; a set that advances steadily means independent recomputation. We had four readings and needed only to difference them. The lesson is not that undated levels are fine — a level you cannot date is still a level you cannot reproduce. It is that the absence of a date field is not the end of the inquiry when you hold the same field on consecutive days. Drop the errand and you throw away the only evidence you had.
For the record, the figure still carries no as-of date on any of the three pages, and we still cannot reproduce it from its inputs. What has changed is that we can now say it is a live calculation rather than an artefact.
What the break was actually worth, in the unit you size in
Before anyone reads this as a signal, measure it. The low of 158.71 is 16.0 pips below the 200-day, which is 0.282 of a standard day. The next named level down, a 21-day moving average, is given as 158.54 on one page and 158.56 on another from the same publisher on the same day — a two-pip disagreement on a figure neither page dates, and 15.0 to 17.0 pips below today’s low. The whole move is larger: 159.97 on 19 November to 158.71 today is 126.0 pips, or 2.218 standard days, and from the 160.00 figure that stopped yesterday’s high it is 129.0 pips and 2.270 standard days.
So a 2.2-standard-day slide has carried price 0.28 of a standard day through a level, into a 31 to 33 pip gap before the next published marker. That is not a break that has established anything, and the publisher that named it says only that it could add momentum to the decline — which is the appropriate amount of claim. For a system the useful framing is the gap, not the break: you are inside a 31-pip no-man’s-land between two undated averages, going into a weekend, with the New York cut still ahead and Tokyo closed on Monday.
The repricing underneath it: the ten-year went back through 5 per cent
The yen move is a residual of something larger, and this is where the slot’s question gets answered. The American ten-year yield has unwound the whole of yesterday’s push:
- 19 November intraday, above 5.10 per cent (one reader, untimed).
- Overnight, 5.046 per cent, down 2.2 basis points, read at 09:09 UTC.
- 4.98 per cent at 10:01 UTC, having broken below 5.
- 4.97 per cent at 10:37 UTC.
Thirteen basis points from the peak, 7.6 from the overnight reading, with the two-year at 4.762 per cent on the same page — a two-to-ten spread of about 20.8 basis points at the 4.97 print. Alongside it: the S and P 500 down 2.02 per cent and the Nasdaq down 2.25 overnight, European equities with one major index down over 2 per cent intraday, the French-German ten-year spread widened to 154 basis points, and the December gold contract settling at $3,669.00, down $72.90 or 1.948 per cent, at a 52-week low. Spot gold is given separately at $3,660.39 with a 14-month low of $3,657.19; we are not differencing a futures settlement against a spot reading and neither should you.
Tokyo closed down 3,007 yen on the Nikkei 225. At 06:22 UTC we published that fall as about 2,820 points and 4.7 per cent, which was an open-window reading, and the close extended it by 187 points — 6.63 per cent more than the number we printed. An intraday extreme is true only up to the moment it was read, and we were caught by our own rule in the mildest possible way. The implied close is about 5.06 per cent down.
The attribution we decline. The publisher reporting the yen move offers three contributing factors — equity risk aversion, the fall in crude, the ten-year through 5 — and explicitly calls them factors rather than the cause. We cannot separate them and will not pretend to. What we will say is that a yen bid arriving on the same day that American yields fall 13 basis points is the ordinary direction of that relationship, and that it has nothing to do with the Bank of Japan, whose December hike expectation has not changed today.
The Bank of England closes at the issuer — and the lone dissenter wanted more
At 06:26 UTC we recorded that one reader had Bank Rate raised to 4 per cent in November, the first rise in more than three years, and that we had not second-sourced it. It is now closed at the primary source, and the detail that matters is the one no summary carries. The Monetary Policy Committee voted 8 to 1 to raise Bank Rate by 0.25 points to 4 per cent at the meeting ending 4 November, published 5 November. The single dissent, Catherine L Mann, voted for 0.5 points, to 4.25 per cent. The dissent was more hawkish than the decision, not less.
That is the second central bank in two days where a dissent count read as dovish pressure has the sign backwards. On 19 November we published the Bank of Japan’s equivalent: one member voted for the September increase and then said it was not enough, while two voted against that increase and then for October’s hold, so the dissenters across forty-two days do not overlap at all. A filter that counts dissents, or watches the count fall, is reading a number whose sign it cannot determine. Two instances is two instances — we are not upgrading it to a rule — but it is no longer an anecdote.
Three further things from the issuer, each of which a vendor row flattens. Inflation was 3.5 per cent in September and the Committee expects it to rise to around 5 per cent by the middle of next year before falling back, which is 1.5 points of further rise written into the document that raised rates. The market-implied median path quoted in the minutes reaches 4 per cent in November and 4.25 per cent in February — not December. One vendor said this morning that a 25 basis point December hike is already expected; we are not calling that a contradiction, because the issuer’s path is a 4 to 5 November vintage and fifteen days old, and a path quoted inside minutes is a dated object like any other. And one member states in the minutes that this is not the start of a hiking cycle while two others expect a further rise in the near term, so the committee does not agree about what it just did.
The next decision is 17 December, a Thursday, with the meeting ending 16 December and the minutes published the same day as the decision. Note the clustering, because it is this slot’s actual job: the Federal Reserve meets 8 to 9 December with a Summary of Economic Projections attached, eight days before the Bank of England, and the Bank of Japan’s meeting runs 17 to 18 December — starting on the same calendar day as the Bank of England decision. Three policy events in ten days, two of them sharing a date.
Sterling, meanwhile, beat on four fields and made a new low for 2026 on one publisher’s headline. The London range we can time is 1.3079 to 1.3122 at 10:37 UTC, 43.0 pips, with sterling-yen’s low at 207.94 and euro-yen’s at 174.04. If you needed a demonstration that the print is not the trade, a four-field beat into a new annual low is it.
The volatility reading we need exists as a headline we could not open
Every standard-day figure above is computed from a one-week implied volatility of 5.68 per cent, observed at 08:27 UTC on 19 November — over 27 hours old at the time of writing. There is a newer reading. A headline stamped 17:22 Tokyo today, 08:22 UTC, says dollar-yen one-week option volatility has risen into the 6 per cent range. We could not open that page, and our own standing rule is that we never publish a figure from a headline whose body we have not read. So we are telling you the reading exists, that we do not have it, and what it would do if it is right: at 6.00 per cent a standard day at this spot is 60.02 pips rather than 56.82, a factor of 1.056, which makes every normalised figure in this article about five and a half per cent too large. The direction of that error is the uncomfortable one — it makes today’s move look slightly more significant than it was.
We also have no option expiry board for today’s 15:00 UTC New York cut. The dated page failed for a third consecutive slot, and this was the slot positioned to read it: on four measured editions the earliest publication time was 08:20 UTC, comfortably before this run. The board did not fail to be early today. It failed to exist where we look for it, and the one source we know of that publishes cut dates per level has nothing on its index either.
What this does not tell you
The 200-day at 158.87 still has no as-of date, we still cannot reproduce it from a high, low and close, and the drift rate of 2.33 pips a day is computed across publishers rather than within one. It is consistent with recomputation; it does not prove it. Two readings in the series are from a single publisher each.
No single traded dollar-yen print in this article carries a clock time of its own. The 158.74, 158.81 and 158.71 lows are each described as early London by the publisher with no minute attached; what we can date is the page, not the tick. The dollar index we have as a sequence — 103.953 Tokyo high, 103.895 prior close, 103.704 London low, back to about 103.88 by 09:40 UTC — and that same publisher says the day has stayed inside yesterday’s 103.588 to 103.974 range, which contradicts a separate page calling 103.93 a fresh high for the year. We take the levels and leave the framing.
We cannot give you a crude oil level with a contract month. One page has an unmonthed 83.21 at 09:09 UTC and a Brent settlement of 92.73; a Tokyo publisher has New York crude in the 81-dollar area at 10:37 UTC. A $9.52 spread between two unidentified contracts on one page is not a number we will publish as a spread, and crude is one of the three factors offered for today’s move, so the gap matters.
One more, on our own archive. Yesterday we published the 19 November close as a band, 159.69 against 159.61, attributed to two publishers. A single publisher carries both figures in two of its own items an hour apart. The band was never a disagreement between houses; it was one house disagreeing with itself, which is a different and smaller finding than the one we printed.