Markets put roughly a 70% probability on the Federal Reserve raising rates next Tuesday and Wednesday. At 12:30 UTC today the Bureau of Labor Statistics publishes what consumer prices did in August. Since this desk last quoted an oil price at 06:00 UTC this morning, Brent has fallen from US$108.50 to US$103.76 — four dollars and seventy-four cents, or 4.4%, inside a single London session, and none of it is in the number that prints in an hour. That gap between what the release measures and what the reaction function has to price is the whole of today’s problem, and it is worth being precise about which half of it you are actually exposed to.
What is already priced, so you know what a surprise would have to beat
The September FOMC hike is running at about 70% per FXStreet this morning and 71% on Trading Economics’ read earlier today — consistent with the three-source reading we published yesterday. The Bank of Japan is further along: FXStreet reports a 25 basis point move at the 17 and 18 September meeting as fully priced, with a follow-up in December carrying high probability. The European Central Bank raised for the second time this year yesterday and warned that price pressures may last longer than it had assumed, which has markets pricing another move as soon as October; Commerzbank is on record looking for 2.75% on the deposit rate by December.
Note what that does to the yen leg in passing. This morning we corrected a wire that had the September Bank of Japan meeting on 16 and 17 September, using the Bank’s own published 2026 schedule, which says 17 and 18. FXStreet independently gives 17 and 18. Two readers, one of them the Bank itself. That correction is now closed.
Three central banks priced for tightening in the same fortnight is not a dollar story. It is a story about whether the dollar’s rate advantage widens or merely holds, and a 3.4% headline print does not settle it either way.
The reaction function has a band, and most of today’s outcomes land inside it
Consensus, from FXStreet and TradingKey in agreement: headline +0.4% month on month and 3.4% year on year, core +0.2% month on month and 2.4% year on year against July’s 2.5% core. TD Securities, quoted by FXStreet, looks for core at 2.3% and +0.19% month on month — one house, below consensus. DBS Group Research, also via FXStreet, puts the threshold that would actually make markets price tightening harder at +0.4% headline and +0.3% core month on month.
Read those together and you get the useful shape. Core month on month at +0.2% is the consensus and changes nothing. At +0.3% it clears one named house’s stated threshold. At +0.1% or below it starts to argue against a hike that is only 70% priced to begin with. Between +0.2% and +0.3% — which is where the overwhelming majority of the probability mass sits — a 70% priced hike is still a 70% priced hike, and the dollar has nothing new to do.
That is the uncomfortable part. The highest-probability outcome of the most-watched release of the week is that it resolves nothing, and the market spends the afternoon trading option expiries and a sentiment survey instead. If your system takes a position at 12:29 because CPI is a big release, you have sized for the tails and you will most likely be paid for the middle.
What the 12:30 print cannot see
August CPI measures August. Here is what crude has done since the start of this month, entirely in our own published figures: Brent at US$99 on 8 September, US$100.42 and falling at 06:00 UTC on 10 September, US$102.09 and rising by 11:40 that same morning, somewhere between US$105.44 and US$108.87 at the 10 September close depending on which vendor you asked, US$108.504 and up 0.81% at 06:00 UTC today — and US$103.76, down 3.60% on the day, when we read Trading Economics at 11:45 UTC. OilPrice.com, read in the same minutes, gives US$103.80 and down 3.59%, with a timestamp eleven minutes old. West Texas Intermediate is US$99.07 and US$99.08 on the same two sources, down 3.32% on both.
Two independent vendors agree on the level and on the size of the fall, so the move is real. Brent is down 4.4% from where we quoted it five hours ago, and up roughly 4.8% from where we quoted it three days ago. Six published prices in seventy-eight hours, with the described direction reversing three times.
Five hours ago this desk published an audit of exactly that series and concluded that a correlation measured over three days is a coin, and that the level is what matters rather than the sign. We are going to hold that conclusion and sharpen it: the level has now moved 4.4% inside one session, which means the level is not a slow variable either. What you cannot do is take a headline inflation print for a month that ended eleven days ago and reason forward from it about energy costs, because the energy input has moved further since Tuesday than the entire year-on-year core figure the market is arguing about to a tenth.
The tape and the narrative are contradicting each other on the same page
We could not source a reason for this morning’s fall, and we are not going to invent one. What we can report is that OilPrice.com’s own front page, read at 11:45 UTC, led with a story published half an hour earlier headlined that Hormuz shipping traffic is plunging as the conflict escalates, alongside a second piece arguing prices could top US$120 — directly above a ticker showing Brent down 3.59% on the day. That is not a criticism of the publisher; wires and prices run on different clocks. It is a warning about what happens when a sentiment-scoring layer reads the headlines and a price feed reads the tape and something downstream has to reconcile them.
There is a third reading that will not reconcile with either. FXStreet’s market wrap at 07:36 UTC had WTI “around $97.50” and Brent at “USD 108/bbl” in the same paragraph — a ten-dollar-fifty spread between the two benchmarks, against the roughly four-dollar-seventy spread both live vendors show now. We are publishing all three reads rather than picking one, because for an automated consumer the difference between a benchmark spread of $4.70 and one of $10.50 is not a rounding difference; it is a different market.
What this does not tell you
It does not tell you what CPI prints, and it does not tell you what the Federal Reserve does next week. A 70% probability is not a forecast; it is a price, and it has been wrong before.
We did not establish why oil fell this morning. We looked, we found no dated, readable account of a cause, and one candidate news item that surfaced in search turned out on inspection to be dated April 2026 and was discarded rather than cited. A price move with two corroborating sources and no explanation is still a fact; the absence of the explanation is also a fact, and we would rather say so than fill it.
The Fed probability figures are vendor implied-probability readings, not our own calculation from futures. The TD Securities and DBS estimates are each a single house’s published view reported by FXStreet and are labelled as such throughout; we did not read either firm’s own note.
We did not read the Bureau of Labor Statistics release, for the straightforward reason that it does not exist yet. When it does, it is the source that settles the core-basket question — as it did on Wednesday, when the agency’s own core measure was two tenths away from the one every vendor carried, because it was a different basket. Prefer the agency’s figure, and say which basket you mean.
Related
- FX events calendar — the dated windows, in UTC
- Signals
- EA presets
