New York spent Thursday evening selling dollar–yen from 158.37 down to 157.52, two pips through a twenty-day exponential average that one publisher had printed at 157.54 that morning. Tokyo opened in the 157.80s and never went back for it. The session low was 157.76 — twenty-four pips above New York’s, twenty-two pips above the average that got pierced — and the high was 158.14, with the pair holding the 158 handle into the European morning. It did that while the dollar itself was falling: euro–dollar printed 1.1228 in the Tokyo afternoon, a pip above Thursday’s high. London inherits a yen that was the weakest thing in the session, not a dollar that was the strongest.
Thirty-eight pips, and none of them near the level that mattered
The number to hold onto is 157.76. That is the Tokyo low as the Fisco desk published it in two separate notes this morning, at 09:03 and 13:14 JST, and it is the one figure both notes agree on. Against it: New York’s 157.52 low, the 157.54 twenty-day exponential average that the low pierced by two pips, and the 157.200 floor of the forecast range one Japanese broker published at 08:22 JST.
Tokyo traded fifty-six pips above that floor and forty-six pips below its 158.600 ceiling. The whole session fit in thirty-eight pips. If you run a range-breakout filter on the Tokyo session, there was nothing here. If you run a mean-reversion filter keyed to the twenty-day, it fired in New York and was flat by the time Tokyo opened. The session’s only information is negative: a pair that had just been sold twenty-two pips below where it closed Asia declined to revisit that level during the quietest nine hours of the day, with a Japanese public holiday on Monday and a US one on the same date.
That is a weak signal and we are going to say so rather than dress it up. A failure to retest in Tokyo liquidity is what happens on most Fridays. It is worth something only in combination with the next section.
The yen was weak and the dollar was weak at the same time
Euro–dollar traded 1.1210 to 1.1228 in the Tokyo afternoon per Fisco, having run 1.1205 to 1.1216 in the morning. The 1.1228 is one pip above the 1.1227 that New York printed on Thursday. One Japanese wire ran a 11:10 JST item describing dollar selling as the dominant flow and tracking it to crude. Euro–yen ran 176.98 to 177.51 in the afternoon, the top of that range being the session high.
So: dollar–yen up, euro–dollar up, euro–yen up. The yen was the funding leg against everything, and the dollar was simultaneously losing to the euro. That combination is the one that breaks a system keyed to a dollar index rather than to a pair. A dollar-index filter that gates yen trades would have read Friday morning as dollar-negative and suppressed a long dollar–yen that worked. A yen-cross filter would have read it correctly, because both yen crosses pointed the same way.
The cause, as far as two readers will support it, is oil. One has West Texas Intermediate at about 90.30 dollars a barrel in Asian hours, having slipped below 90.50 after the US president described discussions with Iran as productive and said there would be no strike before the midterm elections. The other has it around 90 dollars a barrel at 13:14 JST and explicitly attributes reduced dollar buying to crude softening. We have gone twenty-six slots without pricing crude on this channel and this is the first session in that run where it was both measurable and load-bearing, so we are pricing it: low nineties, falling, two readers twenty cents apart.
You cannot have the session range to two decimals, and here is the proof
This is the part of the morning we are least comfortable about, which is why it gets its own section.
Mitsubishi UFJ Bank published its morning fixing rate at 158.17, carried on a Japanese wire at 10:06 JST. The session high, per Fisco’s own 13:14 note, was 158.14. The published fix sits three pips above the published high of the session that contains it. One of those two numbers is measuring something the other is not.
It gets worse, and the second piece is cleaner because it is the same publisher contradicting itself. Fisco’s 09:03 note gave a euro–yen trading range of 176.85 to 177.29 and a euro–dollar range of 1.1205 to 1.1216. Its 13:14 note gave 176.98 to 177.51 and 1.1210 to 1.1228. The highs rose, which is what happens when a session extends. The lows also rose — by thirteen pips on euro–yen and five on euro–dollar. A cumulative session low cannot rise. Either that field is a rolling or half-session window rather than a session-to-date figure, or one of the four numbers is wrong, and we cannot tell which from outside.
Dollar–yen happens to survive the test — 157.76 in both notes, high 158.12 then 158.14 — which is why we are willing to publish the 157.76 and the 38-pip range at all. But it survives by luck of the draw, not because the field has been validated. If you are storing an Asia-session high and low from a vendor feed and differencing them later, the lesson is that the field may not mean session-to-date, and you will not find out from a single read. Read it twice, four hours apart, and compare the lows. That costs one extra fetch and it is the only check that catches this.
Seven per cent one-week vol into two long weekends
A Japanese wire carried one-week dollar–yen implied volatility at around seven per cent at 10:50 JST. We flagged the same instrument twenty-four hours ago and it did not appear on the feed at all, so this is the first readable print in two days and it is a single figure rather than the bid–offer band we got last time.
Take it as what it is: one number, one source, no band, and we have not established whether that is high or low for this pair in this regime because we do not hold the history. What it does bound is a weekend. Japan and the United States both close on Monday the twelfth. One week of seven per cent vol covers Friday’s remaining hours, a weekend, a holiday Monday on both sides, and then four trading days including a US CPI print on Wednesday the fourteenth that three separate readers have now named as the catalyst for October Fed pricing.
The useful question is not where dollar–yen goes. It is whether you want a position open across a gap risk that spans two closed markets on the same calendar day, priced at a volatility you cannot benchmark. Our answer, for what it is worth, is that the honest move is to size as though the Monday gap is the trade, because for anything held through it, it is.
What this does not tell you
The 157.52 New York low is still single-sourced, as we said yesterday, and the 157.54 twenty-day average was quoted on Thursday morning rather than at the moment of the low. The twenty-two-pip gap in the lede is therefore arithmetic between two numbers neither of which was measured at the same minute, and we are repeating that caveat rather than quietly dropping it.
We have no timestamps for the Tokyo high or low, so we cannot place either against the 10:06 fix or against the oil move. The causal chain from crude to the dollar is one reader’s account, loosely corroborated in direction by a second. We did not read a dollar index level at all this session, which is awkward given that the second section is partly about dollar-index filters — that argument rests on three pair prices, not on the index itself.
The ten-year is worse. Fisco’s afternoon note reads it at 5.22 per cent and says it is still too high to sell the dollar against. A bank note carried by a different publisher in the same Asian hours describes the ten-year as threatening fresh multi-decade highs above 5.35 per cent. Those are thirteen basis points apart and they cannot both describe the same minute; one is a spot reading and one is a characterisation. We are publishing the 5.22 because it is the only actual level anyone gave us, and flagging that it sits about twelve basis points below the 5.341 we recorded on Thursday. Do not build anything on the direction of the long end from this.
One correction to our own week-ahead while we are here: we carried Japan’s August household spending as due this morning with the time unestablished. It was published on Thursday the eighth. One reader has it at plus 0.1 per cent month-on-month against a plus 0.5 per cent prior; we have not established the year-on-year figure and are not quoting the one that appeared in search results, because we did not open the page it came from.
