There is no 12:30 UTC release today. The NFIB small business index already came and went at 10:00 UTC — 98.7 for August, down from 99.8 in July — and the only other US number on the sheet is consumer credit at 19:00 UTC, which has never moved a currency pair in its life. The New York open at 13:00 UTC arrives with nothing scheduled behind it. That is not a dull day. It is the cheapest day of the week to be carrying risk, and there are exactly three more sessions before it becomes the most expensive. If you only decide one thing from this note, decide where in the week your exposure is allowed to sit.
Today’s real windows are not data windows
Two dated events matter today and neither is an economic release. The first is the 10am New York cut at 14:00 UTC, where the day’s FX options expire. The second is the US 3-year note auction at 17:00 UTC — the first leg of a three-day refunding that the Treasury’s own tentative schedule lays out as 3-year today, 10-year tomorrow, 30-year on Thursday, all settling 15 September.
On the expiries, InvestingLive’s list for today makes a point worth stealing. There is a sizable USD/JPY expiry at 158.00 and it is irrelevant, because spot is five figures below it: size alone is not enough. The one that matters is EUR/USD at 1.1650, sitting just above the 200-day moving average at 1.1631, with a larger 1.1700 strike above that only comes into play on a meaningful push higher. An expiry near spot with a moving average underneath it is a magnet; an expiry five figures away is a number in a table. Most people read the table and not the distance.
The auction is the one an FX system tends to ignore, and this week it should not. Tickmill’s summary of the 4 September CFTC data — one broker’s read, so treat it as a single source — has speculative net shorts across the Treasury curve at very large levels, with the 5-year alone above 1.38 million contracts after adding 121,452 on the week. A heavily short rates market meeting three consecutive supply events, into an inflation print, is exactly the configuration that produces a yield move first and a dollar move second. Your pair does not know a note was auctioned. It finds out about twenty minutes later.
Thursday is the block, not today
Here is the week the way an event filter should see it, all times UTC.
- Tuesday 8 September — 14:00 New York cut. 17:00 3-year note auction. 19:00 consumer credit. No tier-one data.
- Wednesday 9 September — no scheduled US data. 10-year note auction.
- Thursday 10 September — 12:15 ECB decision. 12:30 US PPI and weekly jobless claims. 12:45 Lagarde press conference and staff projections. 14:00 wholesale trade and existing home sales. 30-year bond auction.
- Friday 11 September — 12:30 US August CPI. 14:00 University of Michigan preliminary sentiment.
- Then — FOMC 15–16 September, Bank of Japan 18 September.
Look at Thursday. A central bank decision, a US inflation gauge and a claims print land inside fifteen minutes of each other, and the press conference starts thirty seconds after the market has finished digesting the first two. That is not four events. That is one thirty-five-minute window from 12:15 to 12:50 UTC in which EUR/USD is simultaneously an ECB trade and a US inflation trade, and in which any stop you have placed on a technical basis is a stop placed on nothing.
If your news filter is configured by event importance rather than by clustering, it will treat Thursday as three separate medium-to-high impact blocks with gaps between them and let you re-enter into the gap. There is no gap. Configure it as a single exclusion from 12:10 to 13:00 UTC on Thursday and be done.
Spend the risk budget at the front of the week
The useful way to hold this is as a budget rather than a forecast. You have a fixed amount of risk you are willing to have on this week. The question the calendar answers is not what will happen, but which hours are cheap to be exposed in and which are expensive. Today and tomorrow are cheap: no scheduled catalyst before Thursday, an auction that is a rates event before it is an FX one, and a market whose main activity is unwinding a position rather than reacting to news. Thursday and Friday midday are expensive: two macro prints and a central bank inside twenty-four hours, both with real distribution.
Most systems get this backwards. They size uniformly across the week and then discover that two-thirds of the week’s realised risk happened in ninety minutes on Thursday and Friday. If your normal position is what you carry today, your Thursday-lunchtime position should be a fraction of it, or nothing. That is a scheduling decision, and unlike a directional decision you can make it correctly in advance, right now, without knowing anything about what the prints will say.
The corollary is uncomfortable. If your system takes its best trades in the quiet hours and gives the money back in the event windows — and most range and mean-reversion systems do — then the honest optimisation is not a better entry. It is being flat from 12:10 UTC Thursday and 12:25 UTC Friday. You already know that. The reason it does not get implemented is that flat feels like doing nothing, and doing nothing is the hardest instruction to give a machine you have paid to trade.
The one number that could rewrite the week
Friday’s CPI is where the week is decided, and the market is going in split. Roughly 60% is priced for a Federal Reserve hike this month after August payrolls came in at 162,000 against a consensus near 56,000 — ING frames the same pricing as about 15 basis points of September tightening, which is the same statement from the other side. The Cleveland Fed’s nowcast, cited by Kiplinger and single-sourced here, has August headline CPI running near 3.38% year on year and core near 2.38%.
A 60% probability is the worst possible starting point for a systematic trader, because it means the market has no consensus to be surprised out of. Both outcomes are genuinely live, both have a real move attached, and there is no directional edge available in guessing which. There is, however, a very clear edge in not being the person whose stop is 20 pips wide at 12:30:01 UTC on Friday.
What this does not tell you
A calendar tells you when, never how much. Thursday could produce a fully priced ECB, an in-line PPI and a fifteen-pip range; Sunday’s German state election moved the euro by almost nothing and it was a record result. Scheduled does not mean volatile, and the reverse is worse — the largest move of the last week happened on Monday, a US public holiday with nothing on the sheet at all.
The auction argument is the softest thing here. Auction dates and the 17:00 UTC time are solid, from the Treasury’s own schedule and FXStreet’s calendar. The link from a weak auction to a specific FX move is not; it is a plausible chain with a variable lag, and we are offering it as a reason to watch rather than a signal to trade. The positioning figures underneath it come from one broker’s summary of the CFTC data and we have not been able to second-source them.
And none of this is a view on direction. If reading a calendar changes which trades your system takes rather than how large they are, you are running a discretionary book with a backtest attached, and the backtest does not describe you.
Related
- FX events calendar — this week’s releases with times in UTC.
- Signals — how we treat event windows in live positioning.
- Expert advisors — news filters, exclusion windows and clustered events.
