Tokyo opens with USD/JPY back at 156.00 after a round trip of about a hundred pips, six hours of thin trading in front of it, and a Bank of Japan decision in the middle of the morning that nobody can price to better than twenty points. Every driver of yesterday’s move was American. Japan supplied a consumer price index at 23:30 UTC and supplies a policy rate around 03:00, and for the fourth session running the yen has taken its instructions from Washington.
The round trip was about a hundred pips and none of it was Japanese
Two publishers give the session high at 156.32 and they agree to the sen: Fisco’s Tokyo note timed 16:11 JST and FXStreet writing at 20:57 UTC. From there the pair traded down to 155.32 on Gaitame’s reading and 155.34 on FXStreet’s — a two-pip disagreement that is as close to confirmation as two publishers get — and closed New York at 156.10, which is Gaitame’s figure and one reader only.
The shape matters more than the levels. Tokyo’s own range yesterday was 155.53 to 156.32, about eighty pips, and the low of the day did not print in Tokyo at all. It printed later, in European and American hours, on American data. Tokyo made the high and then handed the pair over.
The cross tape says the same thing. EUR/JPY ran 178.48 to 179.22 in Tokyo and EUR/USD 1.1456 to 1.1484, both on Fisco’s note. The Nikkei closed at 64,261.00, up 338 points and 0.53%. A rate decision that is supposed to be the largest domestic event in a month produced a half-percent equity session and an eighty-pip currency session the day before it.
The five numbers we said in writing we could not price
At 11:46 UTC yesterday this desk published an article saying that four American releases would land in the same minute at 12:30, that a fifth would land at 14:00 on the New York option cut, and that we had no consensus figures for any of them because the calendar we use returned an access error. We said so in the body and again in the footer, and we declined to invent them. Here is what they were, on one reader — Investrade’s mid-morning note — and we label it as one reader.
- Initial jobless claims 196,000 against a 208,000 consensus. Four-week average 203,250.
- Philadelphia Fed manufacturing 37.8 against 30.5.
- Housing starts 1.275 million against 1.309 million.
- Building permits 1.394 million against 1.410 million.
- Pending home sales +0.3% month on month against a consensus of −0.6%.
The honest verdict on the absence: it cost us less than we feared on direction and more than we expected on one specific thing. Four of the five were readable as a single risk-positive block — labour tighter, manufacturing much stronger, housing construction a little softer. But the fifth is the one where the consensus had the sign wrong. Pending home sales were expected to fall six tenths and rose three. That is a nine-tenths swing in the release that landed on the 14:00 option cut, and it is exactly the kind of thing a desk wants to know in advance and we did not.
The ten-year came back below 5% and the yen closed at 156 anyway
On Wednesday night this desk published a three-way conflict about where the US ten-year closed on Fed day: three publishers, six basis points, two directions, with the Asian handover reading 5.02%. Yesterday it came back. Investrade has the ten-year at 4.944%, down 5.9 basis points on the day; Gaitame’s morning note independently says the yield broke back below the 5.0% mark. Two readers, and this one is not in dispute.
Crude fell with it — WTI at $100.41, down $2.02 — and gold rose to $4,410.40, up $23.00. American equities closed higher across the board: the broad 500-stock index up 0.92% at 7,621, the Nasdaq up 1.56% at 26,382, the Dow up 0.42% at 51,679, the Russell 2000 up 1.10% at 2,890. All of those are the same single reader.
Now hold those two facts next to each other. The American ten-year yield fell six basis points, oil fell two per cent, and USD/JPY closed ten sen higher than the figure. For three sessions this desk argued that the yen was a pure function of the US long end. Yesterday the long end moved down and the yen did not follow it up. If you are running a yen book keyed to Treasury yields, that correlation just stopped paying, and the reason is almost certainly that the market has started trading the Bank of Japan rather than the Federal Reserve — which means the driver changes again in about three and a half hours, in a direction we cannot forecast.
We cannot mark our own 14:00 claim, and we will say why
Yesterday’s article made a near-falsifiable claim: a large USD/JPY expiry sat at 156.00, close enough to spot that the distance condition for a pin was satisfied, but a scheduled release landing in the same minute should defeat it. Both halves of the setup happened. The release landed and it was the sign-flip surprise above.
We cannot tell you who won, because nothing this desk reads publishes a snapshot at 14:00 UTC. We have a session high, a session low and a New York close of 156.10, which is ten sen from the strike. That is suggestive and it is not a mark. This is the second time in three days we have published a threshold our own sources cannot resolve, after a ten-minute window we had to retire for the same reason. The lesson is becoming hard to avoid: if the desk wants falsifiable claims about intraday behaviour, they have to be pegged to a session range, because a session range is the finest thing our sources actually measure.
One thing we can mark. The pre-London note named EUR/USD 1.1450 as the level to watch. Tokyo’s low was 1.1456 and the pair was back at 1.1495 by the American morning. Six pips of room, tested and held.
What Tokyo is actually trading this morning
Japan’s national consumer price index for August was published at 23:30 UTC, three and a half hours before the decision, and we cannot read it yet. What we can tell you is that the two calendars we checked did not agree on what it was supposed to say. FXStreet’s event page gives the headline consensus at 1.9% against a 1.9% prior. OANDA Japan’s preview gives +2.0% against +1.9%. Ten basis points apart on the forecast — and, as usual, the half that can be settled is settled: the Statistics Bureau’s own July release, which we read this morning, puts the headline at +1.9%, the index excluding fresh food at +1.8% and the index excluding fresh food and energy at +1.9%. Both vendors have the prior right. The disagreement is entirely in the forecast, which is the half nobody can settle.
On the measure the Bank actually targets — the index excluding fresh food — both vendors agree: 1.8% expected against a 1.8% prior. That series has now been below the 2% target for seven consecutive months, and the Bank is expected to raise rates this morning anyway.
Gaitame’s morning note puts today’s expected range at 155.00 to 157.00. FXStreet marks resistance at 156.50 and at 157.75, which it gives as the 200-day average, and support at 155.34, 155.00 and 154.00. For context on how wide this month has been: the low on 8 September was 152.89 and the high on 2 September was 160.00.
The practical point for the next six hours is liquidity, not direction. Tokyo will trade a three-figure range on decision morning with most of the real risk sitting on one announcement and one press conference at 06:30 UTC. Position sizing between now and then is a bet on the spread, not on the yen. If your system takes signals in Asian hours, this is the session to widen its stops or sit out — not because the direction is unknowable, which it always is, but because the depth is not there to absorb being wrong.
What this does not tell you
The New York close of 156.10 is one publisher. So is every US index level, yield and commodity price in the third section, and so is the entire consensus column in the second. We looked for a second reader on the American data block and did not find one inside the research window.
We do not have the August CPI actual. The release is minutes old at the time of writing and had not reached any source we can read, including the Statistics Bureau’s own summary page, which was still serving July. We would rather publish the forecast and the confirmed prior and tell you we could not read the print than paraphrase a number we have not seen.
We have no option expiry board for 18 September. That is expected rather than a fault — the publisher we use posts in the European morning — but it means the levels above are technical readings, not flow.
