Twenty-four hours ago this desk published that the base case three readers were carrying — a divided Congress — was dollar-negative, and that it “would send the pair back through the level it just cleared.” The outcome was not divided government. Democrats took both chambers, which is the same sign and more of it. And the call worked: dollar–yen is back under the 200-day moving average it broke on Tuesday night, and it closed under every reading of that line we have. It also paid nothing. The pair traded 157.38 to 158.67 and settled at 158.46, which is a 129-pip range resolving into a close nineteen pips below the line at the top and two pips above the cloud top at the bottom. Being right about direction and wrong about character is the more expensive of the two errors, because it is the one that keeps you in the trade.
The call was right in sign and wrong in shape
The sequence, in order. Tokyo took the pair down to 157.38 on the result — that is the dollar-negative leg, and it arrived on schedule. New York then bought it back to 158.67, two pips through the highest reading of the 200-day anybody publishes, and let it settle at 158.46. One Tokyo publisher prints the close; it is the only close printed anywhere we read.
So the day did three things: it validated the direction, it reversed the direction, and it ended fourteen to nineteen pips below the level that started the argument. A system holding a short dollar–yen position on the inference we published would have been 108 pips onside at the Tokyo low, roughly flat at the New York high, and ahead by a fifth of a figure at the close. That is a correct macro read delivering a result indistinguishable from noise, and it is worth naming because the temptation after a directionally correct call is to size the next one up.
The useful reframing is that the pair is not trending, it is in a box with two walls fourteen to twenty-one pips apart. The 200-day sits at 158.60 to 158.65 depending on the publisher. One Tokyo reader puts the daily Ichimoku cloud top at 158.44. The close, 158.46, is inside that gap. A breakout model keyed to the 200-day and a mean-reversion model keyed to the cloud are both armed, pointing opposite ways, within twenty pips of spot, in the Tokyo morning. One Tokyo desk publishes a forecast range of 157.800 to 159.000 for the session, and notes that the 159 handle has not traded since 24 September.
The line itself is three numbers, and the break was two pips wide
This desk has now published the 200-day moving average three times in three days, at three values: 158.60 on Tuesday night, 158.63 on Wednesday morning, 158.65 this morning. All three come from Tokyo publishers. All three are presented as the level.
Against a New York high of 158.67, the break is seven pips on the lowest reading and two pips on the highest. A two-pip break is not a break; it is the width of the disagreement between two people computing the same average. Which means that whether Wednesday night contained a 200-day breakout at all is a function of whose moving average your chart draws, and that a filter requiring a close above the line would have declined the trade on all three readings — correctly, as it turns out.
We flagged this line on Tuesday as something that “has capped the pair since late September” on one publisher’s account. It is still capping, on the only test it has had. But the honest statement of what we know is narrower than it was: we know of a cluster of estimates in a five-pip band near 158.6 that price has twice failed to close above. We do not know the level, and nobody who publishes it is telling you which window or which price series they used.
What actually moved it was crude, and crude is a settlement plus a timing spread
The reversal off 157.38 was not a dollar story. Brent and West Texas fell roughly 8 to 10 per cent on the view that the election result shortens the Iran conflict — the White House tied Tuesday’s losses to the war, the Vice President said so explicitly, and the President told a newspaper he had no immediate plans to escalate. Falling crude took long yields down with it, took equities up, and produced exactly the risk-on yen selling that overwhelmed a dollar-negative election result. The dollar index fell about 0.3 per cent on one intraday reading. The dollar was weaker and dollar–yen went up, because the yen was the weaker leg.
The crude numbers themselves are a better-behaved disagreement than this desk usually finds, and the distinction is worth drawing. The settlement-labelled figures agree: West Texas for December at $88.12 on one wire and $88.08 on another, four cents apart, both down 8.3 per cent; Brent for January at $94.62, $94.66 and $94.51 across three publishers, fifteen cents apart, down 8.6 to 8.8 per cent. That is not a dispute. What spreads is everything that is not a settlement: a Tokyo desk had West Texas at $91.56 and called it the lowest since 12 October; a fourth publisher has West Texas below $87 and Brent approaching $93, down more than 9 and more than 10 per cent; the Brent intraday low is given as $93.22. Those are not four opinions about one price, they are four clocks.
The part that should bother a headline-keyed system is the headlines. One of these pages is titled “Oil Crashes 9%” over a body that says 8.6 and 8.3. Another is titled “Plunges 10% Near $90” over a body that says approaching $93. Both headlines overstate their own articles, in the same direction, by one to two percentage points. A reader that ingests titles gets a 10 per cent move; a reader that ingests bodies gets 8.5. Positioning scales with that difference.
Two secondary mechanics worth carrying: trend-following funds are reported to have cut Brent length to about 55 per cent from 73 per cent, and one bank put October’s physical surplus at 1.4 million barrels a day. Both are single-sourced and both argue the move had fuel beyond the news.
The option market sold the event, and the Finance Minister said the same thing for the third time
Tokyo implied volatility came down across the curve as the event passed: one-month dollar–yen from 6.81 to 6.63 per cent, three-month 7.60 to 7.52, six-month 8.02 to 7.94, one-year 8.39 to 8.35. The 25-delta risk reversal went the other way, slightly, at every tenor — one-month from 1.87 to 1.89, three-month 1.74 to 1.78, six-month 1.30 to 1.34, one-year 0.61 to 0.62. The publisher attributes that to buying of yen calls as a hedge against a lower dollar–yen, and says option selling dominated as event risk receded.
We are reporting the figures and the publisher’s own characterisation and stopping there, because the page does not state its sign convention and we are not going to tell you which leg is bid on the strength of an inference about notation. What the numbers support without any convention at all is the shape: options got cheaper and more skewed in the same session. Cheaper vol plus a widening skew is a market that has stopped paying for the event and started paying for one tail of it.
Separately, Japan’s Finance Minister repeated the formula: if the election result produces disorderly or excessively speculative moves, the currency authorities will respond, in line with the memorandum confirmed between the Japanese and US authorities. That is the third appearance of that line in three days by our count, and it has now been delivered with the pair at the high, after it had flipped sign inside six hours, and again with the pair in the middle. A sentence that is available at every point in a 129-pip range is not information about the level. It is information about the reaction function at some level further away, and nobody has said which.
Last night’s prints, and a correction we owe on one of them
ADP private payrolls printed 86,000 for October against consensus readings of 67,000, 69,000 and 70,000 across four publishers — a beat of sixteen to nineteen thousand on every one of them. The issuer also revised September from 90,000 down to 86,000, which is October’s figure exactly. So the month-on-month change in ADP private payrolls last night was zero, and the entire “beat” is scored against a forecast built on a prior the same release deleted. The September trade deficit came in at $105.0 billion against a $107.9 billion forecast. The ISM non-manufacturing index printed 54.5 against 54.9 in September, with the consensus given as 54.7 by one publisher and 54.8 by another — the two-labelled-columns fault we have now filed four times in four days, and at a tenth of a point it does not matter here.
The one that does matter: the final October S&P Global services PMI printed 56.1 against a flash of 57.9, a downward revision of 1.8 points on a release whose forecast was the flash copied forward. On Monday we published that the manufacturing equivalent had done the same thing by 1.1 points, and we wrote then that we had conflated “correctly constructed” with “low variance.” This is the second instance in three days, in the larger of the two surveys, and it is 64 per cent bigger. Two instances is still not a distribution, and we are not upgrading it into a rule. But a forecast-equals-flash row is now the only row on this desk’s calendar that has missed by more than a point twice in one week.
And the correction. Yesterday at 11:45 UTC we published that the ADP consensus had “no consensus on any English-language page we can read” and that “the only figure anywhere is a Japanese desk note’s 70,000.” Both halves were wrong. Reuters polled economists and published 69,000, in English, before the release. And the Japanese desk note we labelled as the weakest evidence in the file was within one thousand of that poll. The defensible sentence was “we could not reach it,” and we wrote “it does not exist.” Those are different claims, and the second one is the kind a system acts on by standing down.
The day ahead, and two calendar faults in it
Times UTC. Bank of Japan document 23:50, on which see below. Japan services and composite PMI 00:30, with one calendar printing 52.0 and 52.3 unlabelled. Australian trade balance 00:30, consensus a surplus of 1.00 billion Australian dollars. Japan ten-year JGB auction 03:35, about 2.6 trillion yen, on one calendar only. RBA deputy governor Hauser 06:45. German factory orders 07:00. Norges Bank 09:00, expected to hold at 4.50 per cent. Euro-area PPI 10:00. Bank of England and MPC minutes 12:00, Bailey 12:30. US nonfarm productivity, unit labour costs, initial jobless claims and a Czech National Bank decision all at 13:30. Hammack 13:50, Barkin 14:10, Macklem 15:30, Lane 15:40, Lagarde 16:00. Mexico 19:00, expected to hold at 6.50 per cent.
Two faults inside that list. The initial jobless claims consensus is printed as 1.94 million on one calendar and 196 thousand on another — a factor of ten, on a weekly release, in a four-event minute. The second figure is the plausible one and the first is not far from that page’s own continuing-claims number, so a mislabelled field is the likely explanation; we are publishing the discrepancy rather than the resolution. And at 08:05 one calendar lists a speech by the Swiss National Bank’s chairman while another lists one by an ECB Executive Board member, in the same minute. Two calendars, one slot, two different central banks. A name-keyed news filter fed the wrong one of those opens a window for the wrong currency, which is the failure mode this desk filed in September over a central banker’s identity and has now seen again.
Also worth knowing: no page we read lists a Bank of England Monetary Policy Report alongside today’s decision, and no page we read gives a time for tomorrow’s US employment report. Our 13:30 for that is a fact we carried from Monday, not something we re-read today. Reuters’ survey has nonfarm payrolls at 65,000 after 29,000, private payrolls at 64,000 after 46,000, and the unemployment rate at 4.2 per cent.
What this does not tell you
It does not tell you the result. Two US publishers that called both chambers for the Democrats published no seat count at all, and the only numeric Senate tally anywhere in our file is a Tokyo reader’s 54 to 44 — which totals 98 of 100 and which neither US source corroborates or contradicts. The pre-election baseline was 53 to 47 in the Senate, and two publishers give the House baseline as 219 to 214 and 218 to 214, totals of 433 and 432 against a chamber of 435. We are not publishing a seat count. We note that the Senate race still being tabulated is Alaska, which is the race we said on Monday was the only one on any calendar with a dated resolution step, 18 November, and that several races including one in Oregon were uncalled at 2 a.m. Eastern.
It does not give you a US close. No page we could read was published after the US cash close. One intraday checkpoint has the Dow up 1,501 points or 2.9 per cent and the Nasdaq up 2.1; a later line in the same article says more than 1,600 points with the S&P around 1.7; an untimed ticker shows the Dow at 53,357.19, up 3.03 per cent, and the S&P at 7,688.87. Gold is given at $4,058.90, up 3.31 per cent, on a vendor clock reading 15:45, inside a session range of $3,919.00 to $4,071.40, against an untimed ticker of $4,072.27. Those are checkpoints, not closes, and we are not dressing them up.
It does not settle what the Bank of Japan publishes in the next half hour. Three calendars carry a Bank of Japan document at 23:50 UTC. One calls it minutes, two call it a summary of opinions — different documents with different lags and different information content — and the meeting the pages we read attribute it to is dated September, which we cannot reconcile against a publication today and are not going to assert either way. We did not reach the Bank’s own release calendar. If your schedule has a window on this, you do not currently know what is landing in it.
It does not give you an expiry board, for the second session running. There is no 5 November board on any publisher we can open. One vendor’s orders index still shows 26 October as its most recent expiry item and still labels that item “17 hours ago” — the identical state we published yesterday, twenty-four hours later. A relative timestamp that does not advance in a day is not a stale feed, it is a stopped one, and the two failure modes need different handling. A search result that looks like a 5 November board carries 2024 in its own address.
And it does not tell you what the pair does from 158.46. The box is twenty pips wide, and the day’s real event is a Bank of England decision at 12:00 UTC where consensus is a hike to 4.00 per cent from 3.75 — after London is already open and after Tokyo has gone home.
Related
- FX event calendar — the 23:50 Bank of Japan document, the 12:00 Bank of England decision, and the four-event 13:30 minute.
- Signals — what a correct direction that pays twenty pips should do to the next position size.
- EA presets — a breakout filter keyed to a level three publishers give three values for.
