FOMC (Fed) — scenarios before, scorecard after

Before — what we said

Fed Governor Christopher Waller speaks at 08:30 UTC. There is no consensus figure, so the surprise is measured by tone: does he lean toward hiking in October, waiting for December, or holding? This one matters because the Fed has just turned to hiking, and this week Logan called for '50bp or more' — Waller shows whether that camp is gaining votes. We make calls on gold and EURUSD. We skip BTC: its release-day range is only 1.12 times a normal day's, so the event barely registers, and BTC is trading on risk appetite rather than rates. The dovish branch carries more weight than the hawkish one because weak payrolls hand Waller a ready reason to wait.

The macro picture right now

The market already prices a tightening Fed. The 2y yield at 4.83% (z +2.52) and the 10y real yield at 2.92% (z +3.01) both sit at extremes, while 10y inflation expectations are only 2.36% (z +0.7). The squeeze is therefore in real rates, not inflation. As a result, a hawkish Waller mostly confirms what is priced, whereas a dovish one runs against the extreme and is likely to draw the bigger reaction. MOVE at 113.6 (z +3.61) shows bonds amplifying every signal. On the euro side, speculative positioning at -7.4% OI (z -1.54) is already crowded short, which gives the dovish branch fuel for a squeeze. Even so, EURUSD is still down 3.17% on the month because of French risk. In gold, speculative length at 53.79% OI has dropped 3.07 points in 5 sessions, and GVZ at 23.18% (z -0.78) is cheap: the market is not paying for a big move.

How the last print went

The policy-rate series (a proxy built from IORB) shows the cycle has turned. Rates fell five times in a row, from 4.75% (Nov 2024) to 3.75% (Dec 2025). The latest change, on 17 Sep 2026, was a hike to 4.00% — the first break in that trend. Clues for Waller's tone pull both ways: August CPI at +0.40% m/m supports further hikes, while September payrolls at just +29K (vs 90K expected) give the wait camp its case. On the lane, Reuters reports the Fed may skip October and hike in December; that is the anchor for our inline branch. Where this could be wrong: Trump's public pressure may push officials to sound tougher to protect their independence. Our record on this event: EURUSD has been right 3 of 4 times (on 1 Oct we called bearish and it fell -0.603%); the miss was 16 Sep, when a bearish call produced only -0.269%. On gold, our 1 Oct bearish call met a flat +0.135%.

What it reaches, and how

The event runs on the real-yield lane. For gold, the chain is: Waller -> 2y/real yields -> USD -> gold. Both legs are measured: gold–US10Y r -0.37 ±0.11, and the dominant lane is the dollar (DXY r -0.49, tied with VIX within error), so the path is open. One regime quirk: gold correlates negatively with VIX, against theory, meaning stress pushes gold down rather than up, so the hawkish branch gets no haven cushion. For EURUSD, r with US10Y is -0.27 ±0.12, just above the noise floor, and no axis dominates. The main driver is the French risk premium on the euro side, which this lane does not touch. Waller can only move the dollar leg of the pair, so the impact is capped. For BTC, the link to US10Y (r -0.29) is measured but has to route through the Nasdaq (r +0.39); with VIX low, that route is nearly shut.

Scenarios when the number lands

Three branches, only ONE happens. Each carries its own numeric threshold and an expected direction per asset.

1. Price moved most like this branch 3 of 3 assets agreed

   Numbered threshold

   Waller backs a ≥25bp hike as soon as October and the 2y yield ≥ 4.90%

   If this branch plays out:▼gold▼BTC▼EUR22%likelihood

   Waller siding with Logan forces the market to pull the next hike from December into October. Already-extreme real yields get pushed higher, and the dollar absorbs most of the shock. Gold could test support at 4,137.5 and then 4,120.4, roughly 1–1.5 H4 ATRs away. EURUSD takes a double hit: a stronger dollar while French risk stays in place.

2. Numbered threshold

   Waller keeps a December hike open without committing to October; 2y yield stays within 4.75–4.90%

   If this branch plays out:▬gold▬BTC▼EUR50%likelihood

   This tone matches the Reuters report and is already priced. Gold likely chops around resistance at 4,167.0 without a catalyst to break it. EURUSD falls back to euro-specific drivers, with French risk still weighing on it.

3. Numbered threshold

   Waller says no further hike is needed in 2026, citing labour-market weakness; 2y yield ≤ 4.75%

   If this branch plays out:▲gold▲BTC▲EUR28%likelihood

   Running against the extremes, this branch can travel further than the hawkish one. The dollar falls, gold heads for resistance at 4,188.0, and a crowded EUR short is exposed to a squeeze toward 1.13. Still, every relief leg like this has reversed within a session over the past two weeks, which makes this the branch most likely to fade.

Where we commit — and what would prove us wrong

Unlike the branches above, these carry no “if”. They are the lines graded against real prices below, and they are never edited after the print.

• Gold (XAU/USD)SIDEWAYSConviction 4/10

  Invalidated if: An H4 candle closes above 4,188.0 or below 4,120.4 within 4 hours of the speech

• EUR/USDBEARISHConviction 5/10

  Invalidated if: An H4 candle closes above 1.1300 within 4 hours of the speech

The deciding detail is whether Waller puts a date on the next hike: 'October' versus 'data-dependent' decides whether the 2y yield breaks 4.90% or slips back to 4.75%.

Before — what we said

After the scenarios were published: release time moved 18:00 · 07/10 → 17:00 · 07/10; price-moving moment moved 18:00 · 07/10 → 17:00 · 07/10.

Minutes of the 16–17 September FOMC meeting — the meeting where the Fed raised rates 25bp, its first hike after the cutting cycle. Minutes carry no consensus number; the market is reading how broad the support for further hikes is and how many participants weighed a larger step. Scenario thresholds are therefore anchored on the 2y yield reaction, checkable right after release. Calls cover gold and EURUSD only, keeping the running bearish bias. BTC is skipped: on minutes days its range is just 1.12 times a normal day, essentially no different, and the yield channel to BTC has to detour through Nasdaq — a channel currently ignoring yields.

The macro picture right now

10y real yield at 2.92% (z +3.0) while 10y breakevens sit flat at 2.36%: the bond selloff is purely real yields, the channel gold is most sensitive to. The 2y yield at 4.83% (z +2.52) sits 83bp above the policy upper bound — several further hikes are already priced, so the minutes must be more hawkish than that to surprise. MOVE at 113.6 (z +3.61) shows bonds under strain, so any given surprise gets amplified. Speculative EUR positioning at -7.4% OI (z -1.54) is crowded short: a dovish branch could trigger short covering, a hawkish one has little room to add. Gold is the mirror: specs still hold 53.8% OI (z +1.04) after a 7.0% monthly drop, leaving room to liquidate if yields climb. VIX at 15.52 shows equities are not sharing the bond stress.

How the last print went

The policy-rate series (proxy built from IORB +10bp) fell one way from 4.75% (Nov 2024) to 3.75% (Dec 2025), then the 17 Sep 2026 print broke the trend up to 4.00% — these minutes belong to that reversal. Clues lean hawkish: August CPI at 0.40% m/m was on the table at the meeting, and Logan just called for hikes of '50bp or more'. Where this could be wrong: the minutes predate September payrolls of only +29K, so the market may treat hawkish language as stale; Trump's call for Powell to resign adds pressure. A skip-October, hike-December path (Reuters) is largely priced. Our scored releases: 1 Oct gold bearish came out flat (+0.135%), EURUSD bearish correct (-0.603%); 16 Sep gold bearish correct (-1.136%), EURUSD bearish flat. Gold has repeatedly failed to react within 4 hours, so confidence stays moderate.

What it reaches, and how

Gold: minutes → 2y/real yields → USD → gold. The path is open: correlation with US10Y r=-0.37 ±0.11 clears the noise floor, and the dominant lane is USD (DXY r=-0.49); both links point the same way, so the force is not cancelled. On release days gold's range runs 1.37 times a normal day. EURUSD: minutes → US–euro front-end rate differential → EURUSD. Its release-day reaction is the largest of the three (1.68 times), yet over 60 sessions no axis clears the noise floor (US10Y r=-0.27 ±0.12) because the euro is being driven by French fiscal risk — the path exists but is narrow, and part of the reaction may be two-way volatility. BTC: no direct path; it must run yields → Nasdaq (r=+0.39) → BTC, and Nasdaq is rising regardless of yields.

Scenarios when the number lands

Three branches, only ONE happens. Each carries its own numeric threshold and an expected direction per asset.

1. Numbered threshold

   Minutes show 'most' participants favouring further hikes, with a 50bp step discussed; 2y yield ≥ 4.90% within 2 hours of release

   If this branch plays out:▼gold▼BTC▼EUR25%likelihood

   Minutes show a majority wanting more hikes and a 50bp step genuinely debated, beyond what the 2y yield prices. Real yields extend higher and lift USD; gold likely tests 4,105.8 then 4,088.9 (5 touches each), EURUSD heads to 1.12 — an untested level. With MOVE elevated, the move can exceed 1 H4 ATR.

2. Numbered threshold

   Minutes repeat the September tightening bias without new signals on a 50bp step; 2y yield holds within 4.76%–4.89% within 2 hours of release

   If this branch plays out:▼gold▬BTC▼EUR55%likelihood

   Minutes echo September's tone and the December-hike path already in prices. Reaction stays within about 1 H4 ATR (32.5 USD for gold, 0.003 for EURUSD) and the existing H4 downtrends keep control.

3. Price moved most like this branch 2 of 3 assets agreed

   Numbered threshold

   Minutes show a split, with only 'some' participants seeing further hikes as needed and stressing labour-market risk; 2y yield ≤ 4.75% within 2 hours of release

   If this branch plays out:▲gold▲BTC▲EUR20%likelihood

   Minutes reveal a split and labour-market worry; paired with weak payrolls, the market trims hike expectations. A lower 2y yield pulls USD down; crowded EUR shorts mean EURUSD could rebound more than the surprise alone implies, and gold tests 4,137.5 resistance (8 touches).

Where we commit — and what would prove us wrong

Unlike the branches above, these carry no “if”. They are the lines graded against real prices below, and they are never edited after the print.

• Gold (XAU/USD)BEARISHConviction 4/10

  Invalidated if: Gold closes the post-release H4 candle above 4,137.5 (8-touch resistance)

• EUR/USDBEARISHConviction 5/10

  Invalidated if: EURUSD closes the post-release H4 candle above 1.1250

How many participants are recorded as weighing a 50bp step: if it is more than Logan alone, current 2y pricing is not the ceiling.

Before — what we said

Fed Governor Waller speaks as the ISM Manufacturing PMI prints, with consensus at 54.8 against 54.6 prior. The speech is the part that matters: there is no forecast for it, so the question is whether he confirms the hike path the market has priced or talks it down. We issue calls only on gold and EURUSD, which sit on the real-yield lane. Bitcoin is left out: BTC is range-bound in a tight box, and the stress is in bonds rather than in risk appetite, so no channel links Waller's words to BTC. Price feeds and technical levels both failed this time, so invalidations are set as % moves from the release-time price rather than at support/resistance.

The macro picture right now

The 10y real yield (TIPS) sits at 2.83%, z +3.03 — a one-year extreme, meaning a dovish surprise would hit a crowded pricing and could draw an outsized reaction. The curve prices +125bp of hikes through June 2027: that is the bar Waller must clear to surprise hawkishly, and it is high. The 10y at 5.255% and the 30y at 5.587% (highest since 2004) show the pressure is in bonds; MOVE at 101.8 with VIX at 16.04 confirms this is rate stress, not a flight from risk — which is why BTC stays out.

How the last print went

We have no ISM series this time, so the ±0.5 band around consensus is an estimate, not a calibrated range. On Waller's last two appearances: on Sep 16 we called gold lower and it fell 1.136% over 4 hours; we called EURUSD lower but it moved only −0.269% and was scored flat — the EUR call overreached. On Aug 19 both calls were right. The most useful clue on the lane: Williams just said one more hike may suffice, below what the market prices. If Waller echoes him, that is the dovish branch; where this could be wrong is that the 10y climbed from 5.05% to 5.25% right after Williams, so the market has not yet listened to Fed signals of restraint.

What it reaches, and how

We have not yet measured this event's impact on each asset, so the chains below rest on transmission channels, not on measurements. Gold: Waller's words → expected hike count → real yields → opportunity cost of holding gold; this path is open and running hard — gold futures just lost 3.98% to near $4,148 as yields surged. EURUSD: Waller → US short rates → US–euro rate differential → EUR; open, but the last reaction was small. The simultaneous ISM acts mainly through nominal rates, which makes Waller's share hard to isolate. BTC: no path — in a bond-stress regime without risk-off, yields cap the upside but do not push price.

Scenarios when the number lands

Three branches, only ONE happens. Each carries its own numeric threshold and an expected direction per asset.

1. Numbered threshold

   ISM Manufacturing PMI >= 55.3, or Waller endorses 2 or more further 25bp hikes before mid-2027

   If this branch plays out:▼gold▬BTC▼EUR20%likelihood

   The hawkish branch needs Waller to go beyond the +125bp already priced or a clearly hot ISM; with real yields already at an extreme, room for further rise is limited, so gold's drop would likely be shorter than on Sep 16.

2. Numbered threshold

   ISM Manufacturing PMI between 54.3 and 55.2, and Waller gives no hike count that departs from the priced path

   If this branch plays out:▼gold▬BTC▼EUR50%likelihood

   The inline branch adds no new information, leaving the running real-yield trend in charge; gold stays heavy and EUR stays pinned, though moves may be as small as last time.

3. Numbered threshold

   ISM Manufacturing PMI <= 54.2, or Waller says at most 1 more 25bp hike (in line with Williams)

   If this branch plays out:▲gold▬BTC▲EUR30%likelihood

   The dovish branch flips the running view: if Waller sides with Williams, the Fed–market gap is exposed, real yields could compress quickly from z +3, and gold would likely rebound more than EUR.

Where we commit — and what would prove us wrong

Unlike the branches above, these carry no “if”. They are the lines graded against real prices below, and they are never edited after the print.

• Gold (XAU/USD)BEARISHConviction 5/10

  Invalidated if: Gold rises more than 0.7% above its 14:00 UTC price on Oct 1 within 4 hours of the speech

• EUR/USDBEARISHConviction 4/10

  Invalidated if: EURUSD rises more than 0.3% above its 14:00 UTC price on Oct 1 within 4 hours of the speech

Whether Waller names a specific number of hikes — a count below what the market prices is the only thing strong enough to flip the bearish gold view.

Before — what we said

Neel Kashkari (Minneapolis Fed, a 2026 voter) speaks at 22:00 UTC. There is no consensus number; what gets released is tone, so the scenarios are scored on how many hikes he backs. This one matters because he is the first Fed voice after the soft Core PCE: his remarks show whether the Fed agrees with the market's lower odds of an October hike. We issue a call only for EUR/USD, the one asset the real-yield lane currently drives directly. We skip gold because it is trading on the dollar, and the gold price on the table is a futures quote nearly 10 hours old. We skip BTC because its path runs through the Nasdaq, and US equities are closed at 22:00 UTC. A neutral EUR/USD call does not flip the running bearish bias: that bias rests on OAT-Bund stress, which this speech does not touch.

The macro picture right now

This release comes with no z-scores or rolling correlations in the positioning block, so we cannot say which variables sit at extremes; only prices are readable. The 10y yield is at 5.245%, a few basis points under the 5.28% cycle high after a 10.24% rise over the month. A hawkish tone therefore only needs to restore what was priced before PCE, while a dovish one has to fight a month of pricing. That is why the dovish branch carries the least weight but the largest potential move. Bonds still price roughly +125bp more by June 2027, more than the Fed is saying publicly. VIX at 15.8 shows a calm risk backdrop, so a hawkish shock would land on a complacent market. EUR/USD at 1.1361 sits just above the 5-touch 1.1359 support after losing 1.97% on the month; that is the first floor a hawkish tone would test. BTC at 84,540 is below the 85,094 level the running view treats as cleared, so its bullish trigger is not holding right now.

How the last print went

The fed funds upper bound (a proxy rebuilt from IORB) has just turned. After four straight cuts down to 3.75% in December 2025, the Fed hiked back to 4.00% on 17 September 2026. Clues to Kashkari's tone: August CPI at +0.40% m/m and payrolls at +162K give him reason to keep further hikes open, while the soft Core PCE is a reason to wait. Williams just said one more hike may be needed in late 2026, with 'no rush'. If Kashkari repeats that framing, the inline scenario plays out. Where this clue could be wrong: August CPI is older than Core PCE, and Kashkari may treat PCE as the fresher signal. We have scored two past speeches. On 19 August both calls were right: gold rose 0.729% and EUR/USD was flat at 0.168%. On 16 September gold fell as called (-1.136%), but EUR/USD fell only -0.269% and was scored flat. We had the direction right on EUR/USD but not the size, which is why we lean neutral this time.

What it reaches, and how

EUR/USD has a direct path: Kashkari's tone → expected number of hikes → 10y yield → the rate gap with the euro. The real-yield lane is the one driving EUR/USD (r=-0.37 with US10Y), so it is the asset most directly exposed. Gold has to take a detour. Its dominant lane is the dollar (r=-0.52 with DXY, level with VIX within error), and its direct link to US10Y is only -0.33, so the speech must move DXY before gold reacts. BTC goes the long way, through equities, because its dominant lane is the Nasdaq (r=+0.40). At 22:00 UTC US equities are closed, so that path is largely blocked until the next session. We have no measured release-day impact for this event family, so every sensitivity above rests on the transmission channels alone.

Scenarios when the number lands

Three branches, only ONE happens. Each carries its own numeric threshold and an expected direction per asset.

1. Price moved most like this branch 3 of 3 assets agreed

   Numbered threshold

   Kashkari backs a 25bp hike in October or says ≥ 2 more 25bp hikes are needed

   If this branch plays out:▼gold▼BTC▼EUR30%likelihood

   Kashkari treats the soft Core PCE as a single data point and leans on August CPI. The 10y yield goes back to test the cycle high, and the dollar recovers what it gave up after PCE. For EUR/USD, the next level below the 5-touch support is 1.1321.

2. Numbered threshold

   Kashkari leaves room for at most 1 more 25bp hike in 2026, with no commitment for October

   If this branch plays out:▬gold▬BTC▬EUR50%likelihood

   He repeats Williams's framing: room for one more hike, no rush, wait for ISM and payrolls. The yield lane gets nothing new. The next few hours mostly reflect thin early-Asia liquidity, and EUR/USD typically moves within about one H4 ATR (around 0.002).

3. Numbered threshold

   Kashkari says the 4.00% upper bound is restrictive enough, 0 further hikes

   If this branch plays out:▲gold▲BTC▲EUR20%likelihood

   He calls current policy restrictive enough, which runs against bond pricing. If it happens, this is the widest-moving branch: yields pull off the high and the dollar weakens. Gold gets lift through the DXY channel toward the 5-touch 4,278.1 resistance (futures scale). BTC lags because it has to wait for equities to open.

Where we commit — and what would prove us wrong

Unlike the branches above, these carry no “if”. They are the lines graded against real prices below, and they are never edited after the print.

• EUR/USDSIDEWAYSConviction 4/10

  Invalidated if: EUR/USD (spot) closes an H4 candle below 1.1321 or above 1.1380 within 8 hours of the speech

Whether Kashkari treats one soft Core PCE print as enough to shelve October. That single line decides whether the 10y returns to its cycle high or confirms a pause.

Before — what we said

The Fed announces at 18:00 UTC on Sept 16, press conference at 18:30 — consensus looks for the upper bound at 4.00% from 3.75%, the first HIKE after six straight cuts and Chair Warsh's first meeting. The 25bp itself leaves almost no room for surprise; the entire dispersion sits in the Economic Projections and the tone of the press conference, so the tree below branches on the dot plot rather than on the rate. We deliberately leave Bitcoin out of the calls: FOMC days widen BTC's range only 1.17x a normal day (n=20) — barely a difference — and BTC's correlation with the real-rate lane sits under the noise floor, so a BTC line here is a coin toss dressed as information. The standing bearish view on BTC holds; it simply does not come from this meeting.

The macro picture right now

The 2-year at 4.63% (z +2.69) against an upper bound still at 3.75%: the market is not pricing one hike, it is pricing a cycle — so tonight's hawkishness is measured against that number, not against 4.00%. The 10-year real yield at 2.60% (z +2.57, +0.18 over five sessions) is the binding constraint on gold and it is still climbing. The 10-year breakeven sits unchanged at 2.37% even with WTI at 99.485 (+22.06% m/m): the oil shock is flowing into freight and pump prices, not into breakevens, which gives the Fed grounds to call it a supply shock. MOVE at 83.9 (z +1.22, +10.8 over five sessions) shows rate hedges were bid up into the meeting, while VIX 17.81 and HY spreads at 2.65% (z -1.27) say this is not credit stress but an AI equity de-rating (Nasdaq -3.06% m/m). The most important break: broad USD sits at z -1.32 despite yields at extremes — the rates-to-FX channel is faulty, so a hawkish dot plot need not deliver the textbook dollar bid. Finally, gold positioning at 56.4% OI (z +1.49) is a crowded long while EUR at -4.52% OI (z -1.23) is already short: gold's left tail is the fatter one, the euro's right tail is the twitchier one.

How the last print went

The policy series ran one way across six level changes: 5.00% (Sept 2024) → 4.75% → 4.50% → 4.25% → 4.00% → 3.75% (Dec 11, 2025), each step exactly 25bp, then no mark at all for nine months. A 4.00% print would be the first reversal in the series, and since the Fed has never changed the level by more than 25bp anywhere in this range, the 50bp branch has to stay light. The nearest clues lean hawkish: August 2026 CPI m/m printed 0.40% (roughly 4.9% annualized) and non-farm payrolls came in at +162K — not an economy asking to be rescued; the lane also carries the 10-year clearing 5% for the first time since 2023 and real yields at their highest since 2008. Where that clue can fail: much of the 0.40% is energy with Brent near $100, and consumer confidence at 47.8 alongside household inflation expectations of 4.6% is the classic argument for one hike and a pause. At the Aug 19, 2026 FOMC we scored 2/2 (gold bullish +0.729%, eurusd neutral +0.168%), but that was an unchanged meeting — one correct read is not enough to raise confidence into a meeting that carries an SEP.

What it reaches, and how

The event runs on the real-rate lane, and the three assets do not stand at equal distance from it. EURUSD is the only straight line: us10y r=-0.26 ±0.12 is the single axis above the noise floor across all three, and FOMC days widen EURUSD's range to 1.69x a normal day (n=22) — the highest of the group. The specific chain: dot plot → US 2-year yield → Fed-ECB policy gap → EURUSD, on top of the near-100bp French-German risk premium the euro carries alone. Gold has to go around: us10y r=-0.20 ±0.12 is under the noise floor, so we have NOT measured a direct link, and gold's dominant lane is the dollar (dxy r=-0.57). The real path is dot plot → real yields → USD → gold; the 1.38x FOMC-day range (n=22) shows it does run, but the middle leg is faulty with broad USD at a weak extreme — which is why gold gets lower confidence than EURUSD. One regime warning: gold-vix r=-0.38 is the wrong sign versus theory, gold has been falling exactly when risk rises, so do not expect it to act as shelter if the press conference sells equities. BTC is the blocked path: us10y r=-0.14 is below the floor, its dominant lane is also dxy (r=-0.44), so it shares the faulty middle leg with no event-day amplification to compensate.

Scenarios when the number lands

Three branches, only ONE happens. Each carries its own numeric threshold and an expected direction per asset.

1. Price moved most like this branch 3 of 3 assets agreed

   Numbered threshold

   A 25bp hike to 4.00% BUT a 2026 year-end dot-plot median at 4.50% or higher (at least two more hikes), or a straight 50bp move to 4.25%.

   If this branch plays out:▼gold▼BTC▼EUR32%likelihood

   This branch validates the exact path the 2-year is demanding and extends the tail of the cycle, pushing 10-year real yields above current levels and pressing directly on a crowded long in gold. It is also the only branch strong enough to pull broad USD out of its weak extreme, because it forces a repricing of the whole cycle rather than just the next meeting.

2. Numbered threshold

   A 25bp hike to 4.00% with the 2026 year-end dot-plot median at 4.25% — exactly one more hike, then stop.

   If this branch plays out:▼gold▬BTC▼EUR48%likelihood

   The paradox of the central branch: it is still SOFTER than the path a 4.63% 2-year is pricing, so the first move at the front end could be a modest easing. But it removes neither the real-rate constraint sitting on gold nor the widening Fed-ECB policy gap — both legs of the bearish view survive, only with a narrower travel range.

3. Numbered threshold

   Hold at 3.75%, OR hike 25bp but with the 2026 year-end dot-plot median stopping at 4.00% and the press conference framing the oil shock as temporary.

   If this branch plays out:▲gold▲BTC▲EUR20%likelihood

   This branch unwinds close to 90bp of expected tightening embedded in the 2-year, pulls real yields off their highest since 2008 and repairs the faulty dollar leg. It is the only one of the three that flips all three standing bearish views, and the already-short EUR positioning makes the euro leg the sharpest snap.

Where we commit — and what would prove us wrong

Unlike the branches above, these carry no “if”. They are the lines graded against real prices below, and they are never edited after the print.

• EUR/USDBEARISHConviction 6/10

  Invalidated if: Wrong if EURUSD clears 1.1553 (3-touch resistance, 1.13 H4 ATR away) and closes the first post-press-conference H4 candle above it.

• Gold (XAU/USD)BEARISHConviction 5/10

  Invalidated if: Wrong if gold trades above 4,300.6 (4-touch resistance, 1.03 H4 ATR away) and holds it through the H4 candles after the press conference.

The first number worth reading is not the policy rate but where the 2026 year-end dot-plot median lands relative to 4.63% — the 2-year is pricing a full cycle, while tonight the Fed is only being asked to confirm a single step.

Before — what we said

The FOMC minutes land at 18:00 UTC on 19 August. There is no consensus estimate because this is a text, not a number; what the market reads is a headcount — how many participants treat a return to hikes as a live discussion. That question already carries a price: the 2y yield sits above the current target-range ceiling while the Fed has stood still since December 2025. Bitcoin is left out of our calls: minutes days deliver a 3.68% range against 3.21% on ordinary days (1.15x), and BTC's correlation with the 10y at r=-0.16 is below the noise floor — we have no measured channel to name a direction with. Gold keeps a mild upward bias even though the neutral branch is the heaviest, because of the transmission asymmetry set out below, not because we expect a soft text.

The macro picture right now

The 2y at 4.17% (z +1.55) against a 3.75% policy ceiling is precisely the amount the market has prepaid for the hawkish case, so a text that merely fails to feed that faction has to bleed some of it back. The 10y real yield at 2.41% is a cycle high (z +2.22) while 10y inflation expectations sit at just 2.28% (z -0.57): the squeeze comes from term premium and fiscal worry, not inflation — which makes the balance-sheet passages of the minutes weigh as much as the rate passages. Cutting the other way, DXY at 99.64 marks an eighth week below 100: the yields-to-USD link is broken, which is why a hawkish tone need not become a stronger dollar. The risk cushion is thin — VIX at 14.25 and HY spreads at 2.67% (z -1.25) say nobody bought protection into the event, so any surprise has to build its defence on the spot. Positioning is skewed at both ends: gold specs at 54.4% of OI (z +1.24) are the side that gets squeezed on a hard tone, while EUR is already short at -7.48% of OI (z -1.64, down another 5.44 points in five sessions), leaving the downside branch of EURUSD with less fuel than the upside one.

How the last print went

The policy-rate series (a proxy rebuilt from IORB plus 10bp, not the official target range) runs one way: 5.00% to 4.75% to 4.50% through 2024, then 4.25% to 4.00% to 3.75% at the September, October and December 2025 meetings. Six cuts in the same direction, and no new level marked after 11 December 2025 — eight months on hold, with no turn on record to use as a template. The clues for this print conflict: July CPI at 0.07% m/m and non-farm payrolls at -23K are evidence against a hike faction, but the minutes describe a meeting held before all of that data was on the table, so they can read more hawkish than the current data flow — and that is exactly where this inference can fail. On the same lane, the 30y auction cleared at the highest yield since 2001 and Wells Fargo flipped to calling a 25bp hike this year, while soft retail sales and consumer confidence pulled hike odds back down. No prior edition of this event has been scored by us.

What it reaches, and how

The minutes run on the real-yield lane, and that is not the lane currently steering gold. Gold against the 10y sits at r=-0.24 ±0.12, below the noise floor: there is no straight line. The real route detours — tone of minutes to the 2y and real yields, to DXY (r=-0.53 with gold, and the dominant lane), to gold. That detour is blocked in the middle, since the dollar has ignored yields for eight weeks, so hawkish content loses amplitude on the way to gold while dovish content passes intact; that asymmetry is the whole reason the mild upward bias survives. Gold's minutes-day range of 2.04% against 1.46% normally (1.39x) is enough to make the event worth reading. EURUSD is the cleanest path: r=-0.34 against the 10y is above the noise floor, giving a direct rate-differential leg, plus a second leg through risk appetite (vix r=-0.46, the dominant lane) — and 1.66x the normal range is the highest ratio of the three. BTC has no measurable leg on this lane.

Scenarios when the number lands

Three branches, only ONE happens. Each carries its own numeric threshold and an expected direction per asset.

1. Numbered threshold

   Minutes record 'several' or more participants viewing a 25bp hike as possibly appropriate, with the 2y yield at 4.25% or higher (+8bp) within 4 hours of release

   If this branch plays out:▼gold▼BTC▼EUR27%likelihood

   The minutes reveal a group genuinely debating a return to tightening. Transmission runs through the 2y first and only then tries to drag DXY along — and the DXY segment is where it usually jams, so gold's decline can be shallower than the yield move implies. EURUSD takes the full hit via its direct leg, but the crowded EUR short caps how far the branch travels and leaves it prone to a snap-back.

2. Numbered threshold

   No distinct pro-hike faction, two-sided data-dependent tone; the 2y yield stays within 4.13%–4.21% and DXY holds below 100.00 four hours after release

   If this branch plays out:▬gold▬BTC▬EUR51%likelihood

   The minutes hold a data-dependent tone, weighing inflation and labour risks without leaning either way. The prepaid tightening premium is neither topped up nor refunded, EURUSD stays inside its heavily tested 16-pip band, gold reverts to its non-rates story, and attention shifts to the next payrolls and CPI prints.

3. Price moved most like this branch 3 of 3 assets agreed

   Numbered threshold

   Minutes frame labour-market downside as the dominant risk with no hike language; 2y yield at 4.09% or below (-8bp) and DXY at 99.30 or below within 4 hours

   If this branch plays out:▲gold▲BTC▲EUR22%likelihood

   The minutes show most participants more worried about the labour market than inflation. The tightening premium sitting in the 2y gets refunded, DXY drops further below 100 and gold captures the full move because this is the direction the channel does not block; EURUSD is pushed by its rate leg and its risk-appetite leg at the same time.

Where we commit — and what would prove us wrong

Unlike the branches above, these carry no “if”. They are the lines graded against real prices below, and they are never edited after the print.

• Gold (XAU/USD)BULLISHConviction 5/10

  Invalidated if: Gold trades below 4,369.1 (the 6-touch support) within 4 hours of the release

• EUR/USDSIDEWAYSConviction 6/10

  Invalidated if: EURUSD prints outside 1.1549–1.1594 — half an H4 ATR beyond the 1.1557 floor and the 1.1586 cap (8 touches) — within 4 hours of the release

The gap between the 2y yield and the current target-range ceiling: if the minutes cannot feed the hawkish faction, that gap is the first thing the market has to hand back — before DXY or gold have time to respond.