Seventy-Four Pips of an Implied 104. The Prior Under Yesterday's Finding Was Revised Away.

Seventy-Four Pips of an Implied Hundred and Four

The full session, Tokyo through New York, ran 156.98 to 157.72. That is 74 pips. A one-week implied volatility of 10.50 per cent, converted to a daily one-sigma at a 157.35 spot, is about 104 pips. The day used 71.1 per cent of it.

Decompose it by session and the answer is sharper than the range. Tokyo established 157.21 to 157.58, 37 pips, which was 35.6 per cent of the sigma on its own. London added 14 pips of new high and one pip of new low — fifteen pips of price that Asia had not already visited, in the busiest six hours of the FX day. New York added 22 pips, all of it on the downside, taking the low from 157.20 to 156.98.

Tokyo did half the day’s work in a third of the day’s hours. That is the second consecutive session in which the Asian hours out-produced the European ones on this pair, and if you are running an intraday breakout system with a London-session bias, it has now been wrong twice in a row for a reason that is structural rather than unlucky: the yen’s drivers are being priced where the yen trades.

The threshold we named yesterday was explicit — 52 more pips. The test is closed. Three sessions in a row have now failed to deliver what the option market was charging for, which is a statement about the volatility premium and not about direction.

The 1.1400 Strike: Closed, and the Filter Was Correctly Never Armed

Yesterday’s published condition was conditional and we said we would not declare it settled until 13:30 UTC had passed. It has. EUR/USD’s high for the entire day was 1.1374, set in the Tokyo morning, 26 pips short of the 4.9-billion-euro strike at 1.1400. From there it fell, made a session low of 1.1312 — 88 pips below the strike — and closed at 1.1342, still 58 pips away. Price was never once inside the 15-pip band the test required.

That is the second null on this exact strike in two days, and the notional grew 8.9 per cent between them. Yesterday we drew the lesson that the notional is not the variable and the distance is. Today confirms it in the strongest available form: a bigger strike, and price walked 62 pips further away from it over the session. An expiry ladder tells you where the pinning risk would be if spot went there. It does not make spot go there, and a filter that arms on notional alone will arm on days like this one and find nothing.

One caveat on the low, because our source is not internally consistent. The publisher that gives 1.1312 calls it the lowest since 29 May 2025 and in the same breath calls it a fourteen-month low. From the end of September 2026 that date is sixteen months back, not fourteen. We publish the level and the date and not the interval, and we flag that one of the two labels on that page is wrong.

The Prior Moved Under Our Own Finding

Yesterday this desk published a consensus dispute and called it the sharpest in the file: against an agreed previous of 89.4 on US consumer confidence, one calendar forecast 89.2 and another 90.1, so one expected a fall of 0.2 and the other a rise of 0.7. Two vendors handing a directional system opposite trades from identical inputs. We ran the containment check and it passed.

The print was 81.9. And the August figure it was printed against was not 89.4. It was revised down to 88.6, and the revision arrived with the release. Recompute on the revised prior and 89.2 is a rise of 0.6, 90.1 is a rise of 1.5, and the two forecasts are on the same side of the line. There was no sign disagreement. There was a magnitude disagreement of 0.9 points that we dressed up as a directional one because we used the pre-revision base.

This desk has a standing rule that reads, in terms, check whether the prior was revised. We wrote the rule and then published a finding that the rule exists to prevent. The honest version of yesterday’s paragraph is that two calendars disagreed by 0.9 on the level and agreed on the direction, and that the direction they agreed on was wrong by between 7.1 and 8.2 points.

Because that is the other half of it. Three vendors carried three consensus figures into this release — 89.2, 90.1 and 89.0 — and 81.9 missed all three by more than seven points, and missed the revised prior by 6.7. The expectations sub-index fell to 63.6, its third consecutive monthly decline; the present situation index fell 7.9 points to 109.3. A vendor disagreement of 0.9 is noise next to a surprise of 7.3. The lesson is not that calendars disagree. It is that we spent two runs adjudicating a nine-tenths-of-a-point argument about a release that was about to move seven points, and the adjudication would have been worthless even if we had won it.

The same minute carried August job openings at 7.079 million against vendor expectations of 7.225 to 7.23 million — a miss of about 0.15 million, in the same direction. Both prints were soft and the dollar index nonetheless traded 101.40 to 101.50, multi-week highs, on the day. Work out what that combination implies about what is actually driving the dollar before you next size a position on a US survey.

What Tokyo Has to Trade in the Next Three Hours

Japan’s preliminary industrial production and retail sales land at 23:50 UTC, inside the hour. The calendar consensus is +1.3 per cent month-on-month and +6.8 per cent year-on-year on production, against a previous of −0.2 and +3.9; retail sales +3.2 per cent from +4.0. A swing from a contraction to better than one per cent is a large expected move on a preliminary series, which is exactly the shape that produces revisions.

Then 01:30 UTC, and it is not in your pair. Australian CPI is forecast at +4.1 per cent from +3.5 — the first inflation print since the RBA hiked 25 basis points to 4.60 per cent unanimously yesterday and the Governor then said maybe there does not need to be any more. We published yesterday that the statement and the Governor did not agree. This print is the arbiter. If it comes in near 4.1, the Governor’s line becomes the harder one to defend, and AUD/JPY is where that argument gets settled during Asian hours on thin books.

China’s official PMIs arrive in the same minute — manufacturing forecast 50.1 from 49.8, so the consensus is a move across the boundary — and Caixin follows at 01:45. Two data events in one minute, then a third fifteen minutes later, on the last day of the quarter. Gaitame.com’s range for the day is 156.60 to 158.20, 160 pips, and spot enters the session in the middle of it.

What This Does Not Tell You

The session extremes come from one publisher for the low and a second for the high, and the two do not agree on where the pair closed: one reads 157.60 at 04:25 JST and the other 157.29 at 06:03 JST, with a Kyodo print of 157.49 to 157.50 at 03:00 JST in between. Those are three different timestamps, so we treat it as a drift through the thin hours rather than a vendor dispute — but we cannot prove that, and if you need a closing print to the pip you should take your own.

The 104-pip sigma rests on a one-week implied volatility this desk read yesterday morning and did not re-read, and the one-week-to-one-day conversion assumes a flat term structure that almost certainly is not flat.

We have no positioning data. The yen futures position underneath all of this has not been measured since a Tuesday close we are now eight days past, and we have deferred writing about that for seven runs. No conclusion in this article accounts for it.

And we have published nothing about the Nikkei, whose reported fall of roughly 1,200 points we have now carried single-sourced for three runs without finding a second reader. We are dropping it rather than carrying it a fourth time.

Related

• FX events calendar — the 23:50, 01:30 and 01:45 UTC cluster.
• Signals
• EA presets