Dollar-yen opens the Tokyo week in the 157.80s, within about eleven pips of where it closed on Friday. In between sits the weakest non-farm payrolls print this desk has had to mark: plus 29,000 against seven published consensus figures running from 84,000 to 100,000. The pair fell 65 pips on it, took all of that back inside three hours, and handed Tokyo a flat tape. The interesting part is not the round trip. It is that the three feeds publishing the October Fed probability cannot agree on whether anything was repriced at all — one pair of readings has it going from about 28 per cent to below 13, another has it going from 25 to 23 — so the answer to “what did the print do” depends entirely on which vendor you happened to have open.
The round trip, and the two readings of where the top was
September payrolls landed at 12:30 UTC on Friday. Dollar-yen was in the 157.60s going in and traded 156.95 on the release — 65 pips, which is a perfectly ordinary reaction to a number that missed the median forecast by 61,000 jobs. By 15:23 UTC one publisher had the entire decline erased and the pair back in the 157.70s. It closed the New York session at 157.90 or 157.91 depending on the reader, and this morning it is in the 157.80s, which the same publisher describes as around Friday’s close.
Friday’s high has two values. Two readers put the New York session high at 157.91 and 157.92 — one pip apart, which is as close to agreement as this channel ever gets. A third gives the intraday high as 158.22 and the previous New York close as 158.09. Those are not in conflict so much as measuring different windows: if the pair was above 158 before New York, then 158.22 is a full-day figure and 157.91 is a New York-session figure, and both are right about the window they cover. That is a hypothesis about the two pages, not something we have established, and we have not found a reader that prints both windows side by side.
It matters because of what this desk published on Friday morning. We put the one-standard-deviation implied daily move at 90.9 pips and said that was the only actionable number in the article. The day delivered 96 pips on the narrow reading and 127 on the wide one. Both are larger than 91. The vendor disagreement does not need resolving for the mark to come out the same way, which is the most comfortable position a falsifiable claim can be in.
Three feeds, three Octobers, and the conclusion flips
Here is the number that should bother you more than the payrolls print. On Friday morning, before the release, this desk published the October Fed hike probability at 25 per cent, from CME’s tool as carried by one publisher at 03:35 UTC. After the release, three separate readings exist:
- One publisher gives 77 per cent priced for a hold, citing the same CME tool — a complement in the low 20s, though a hold complement is not strictly a hike probability.
- One puts it in “the 20 per cent range”, having been around 70 per cent a month ago.
- One states it fell from around 28 per cent before the data to below 13 per cent after it.
Take the first pair and payrolls moved October by about two points. Take the third and payrolls halved it. Same event, same binary, same afternoon. This is the fourth time this channel has found three feeds giving three probabilities for one central-bank meeting, and it is the first time the spread has been wide enough to reverse the sign of the conclusion rather than just blur it. If you are running anything that keys off a policy probability — a position-size multiplier, a news-window skip, a regime flag — the vendor you pinned it to is a parameter of your system, and on Friday it was worth fifteen points.
We are not adjudicating. We have tried that for a month on consensus figures and the exercise is bankrupt. What we will say is the useful and uncomfortable thing: a reading of the October probability is only meaningful with the feed and the timestamp attached, and a desk that quotes one bare number — including this one, on Friday morning — is overstating what it knows.
The ten-year did the work, and we are not naming a cause
What is well sourced is the yield. The US ten-year was around 5.22 per cent before the release, traded as low as 5.1549, and was back at 5.296 by the New York afternoon — three readers give the low at 5.15 and two give the close at 5.29. The two-year was quoted at 4.73, down six basis points. So the long end gave up six basis points on a bad employment number and then took back fourteen, finishing the day higher than it started. Dollar-yen tracked it both ways. One reader attributes the dollar’s recovery to the expected continuation of the US-Japan rate differential, which is a description rather than a cause.
One publisher’s headline attributes the recovery to tariffs. We could not read the measure, the announcement or the timing in the body of that page, and this desk does not publish a cause it cannot read in full. So: the pair erased a 65-pip decline while the ten-year erased a six-basis-point decline, the two moved together, and the thing that turned the yield around is not established here. We looked.
The option surface has a second reader at last, and the curve is inverted
A housekeeping note that is more important than it sounds. On Friday this desk published the dollar-yen implied volatility and 25-delta risk-reversal curve from a single Japanese desk note and said plainly in the article that half the piece failed if that one page was wrong. It was the largest single-source dependency on this channel. It now has a second reader, from a different publisher, and the two agree.
The first publisher had the one-month 25-delta at 2.47 in favour of yen calls at 17:41 UTC on 1 October. The second publisher had the one-month risk reversal at 2.495 at 08:12 UTC on 2 October, and a 2.30–2.67 quote — midpoint 2.485 — five hours before that. Two and a half hundredths of a vol point apart, from two desks, with opposite sign conventions for the same thing: one writes the yen-call premium positive, the other writes it negative. If you are parsing either page, the sign is the publisher’s convention and not the market’s direction.
The second reader also gives a tenor the first does not, and it is the one that matters for a week like this. At 08:12 UTC Friday the curve read: one week 10.00 per cent, one month 9.45, three month 8.92, six month 8.95, nine month 8.92, one year 8.86. That is inverted by 1.14 vol points from the front to the back, and the one-week had already retreated from 10.5 per cent earlier that morning. An inverted vol curve is the market saying the risk is in the next few sessions rather than the next few quarters. It said that on Friday about payrolls and it was right by about five pips.
What Tokyo actually has to trade this week
The levels first. One reader puts the 200-day moving average at 158.51 as resistance, the 26-day Ichimoku baseline at 156.64 as support, and the conversion line at 157.71 — which is where spot is. Another frames the week as a 156 to 158 box with the 158 area carrying intervention concern. The 160 line this desk has carried for weeks is not in play at 157.80.
The events, in UTC. Japan services and composite PMIs at 00:30 this morning, forecast 51.6 and 52.5. Japanese consumer confidence at 05:00, forecast 35.5, in the same minute as the Prime Minister’s policy address to the Diet — a special session convened today. The Bank of Japan publishes its output gap and potential growth estimates at 05:00 as well. Then US services PMI final at 13:45 and the ISM non-manufacturing index at 14:00, forecast 55.2 at one reader and about 55.0 at another, which is a gap small enough that this desk is explicitly not opening an item on it.
Tomorrow is the one to put in the schedule: the Finance Minister and the Bank of Japan governor both speak at a securities industry convention, alongside a ten-year JGB auction of roughly 2.6 trillion yen. Tokyo-area core inflation printed 2.7 per cent against a 2.4 forecast ten days ago and core-core reached 3.0, and this desk has now failed twice to find a post-print reading of October Bank of Japan pricing. The governor speaking is the next chance to find out what the market thinks the 3.0 means, and if the answer arrives in a speech rather than a probability feed, that is still better than the nothing we have had.
Chinese markets are shut through 7 October. Thin crosses, a policy speech and a central-bank governor inside 48 hours: size for the gap risk, not for the trend.
What this does not tell you
It does not tell you which of the three October probability readings is right, and we have no way to settle it — the CME tool is quoted by two of the three, which means either one of them read it at a different moment or one of them is not quoting what it says it is quoting. It does not tell you what turned the ten-year around on Friday afternoon. It does not resolve Friday’s high, and the window hypothesis above is a guess about two pages rather than a finding. The option figures are from Tokyo-session readings on Friday and are therefore pre-payrolls; the post-release vol reading existed in a headline we could not open, so it is not in this article. The Monday morning level is from one publisher at 22:10 UTC and has no second reader. And none of this tells you direction: a pair that opens eleven pips from its close after a 61,000-job miss is a pair with no information in it, which is an argument for smaller size and not for a view.
