Gold sells the war: when geopolitical risk routes through real yields

Gold has dropped 4.76% on the session to 4,480 because the market decided the Iran headline was a Fed story, not a haven story. Oil up, breakevens up, nominals up 22 bp on the 10y, DXY up 1.1% — the entire chain prices in a hawkish reaction function before the metal gets a chance to do its usual job. The 270-point intraday range (4,725.80 → 4,455.00) is the kind of reversal that resets dealer gamma and the speculative book at the same time. From here the read is binary: either 4,455 holds as a weekly-pivot confluence and the 4h RSI of 28.3 produces the mean reversion it usually does, or 4,455 gives and we're going to look at the weekly S2 at 4,374 and the 200-day for the first time in months.

The session

Spot printed a session high at 4,725.80 in the Asia handover, then unwound the entire move into London — last 4,480.40, sitting between daily S1 (4,490.80) and daily S2 (4,475.30) with a session low at 4,455.00 that taps weekly S1 (4,465.00) on the nose. PAXG on Binance is +0.03% versus spot, so there's no crypto-side dislocation; this is a clean futures-led move with the physical wrapper just along for the ride.

The catalyst chain, in order: Trump's "two or three days" warning on Iran (per Trading Economics' read) put a fresh bid into Brent overnight via the Strait of Hormuz channel; that bid pushed breakevens higher; the curve responded by selling duration, with the 10y up 22 bp in 24 hours; DXY caught the safety bid that gold normally gets when the dollar isn't the cleaner expression of the trade. Net effect: real yields ripped, and real yields are still the only macro variable that consistently bosses this tape.

FXStreet led with "Gold hangs near late March lows as hawkish Fed bets and geopolitics underpin USD" — note the framing. Geopolitics is supporting the dollar, not gold. That's the inversion that broke the day.

Multi-timeframe read

The grid is decisively bearish but the lower frames are starting to disagree, which is what you'd expect at a 4h RSI of 28.

• 15m — RSI 51.6, MACD histogram +2.37 but rolling over. Price above EMA20, below EMA50 and EMA200. This is a dead-cat bounce signature: enough lift to look constructive intraday, not enough to flip the structure.
• 1h — RSI 39.4, MACD histogram +0.22 and turning up. Price below EMA20 and EMA50. The MACD turn is the first honest divergence on the board — it's how relief rallies start, but it's not a trend signal until the EMA20 reclaims.
• 4h — RSI 28.3. Deeply oversold. Below EMA20. There's no MACD reading in the snapshot, which is worth flagging — treat it as a data gap, not a confirmation.
• 1d — RSI 36.7, MACD histogram -14.15 and still falling. Below EMA20 and EMA50, but above the 200-day EMA. That last point is the trade — the multi-month uptrend is dented, not broken. The bears have not yet earned the right to talk about a regime change.

Where they agree: trend is down on every timeframe ≥ 1h. Where they disagree: 1h MACD is hinting at a bounce that the daily has not signed off on. That's the edge for the next 24h — fade either direction at the right level, but don't carry it.

Macro frame

DXY at 99.40 is doing the heavy lifting against gold. The 30-day daily-return correlation is -0.65, so the dollar move alone explains the bulk of the metal's pain in a statistical sense. But the dollar's own internals are starting to stretch: 4h RSI 69.7, 1d RSI 59.8, MACD positive but flat. The dollar isn't overbought on the daily yet, but the 4h is close, and that's where the gold bounce — if it comes — will be born.

10y at 4.67% with +22 bp in a single session is the bigger tell. That's a duration-led move, not a Fed-funds move, which means the bond market is repricing term premium and inflation risk in tandem rather than just pulling forward a hike. Real yields are almost certainly up materially — I don't have the breakeven print in the snapshot so I won't put a number on it, but the direction is unambiguous and that's the variable gold actually trades against.

On non-US central-bank colour: Kitco's front page returned no article text on this fetch, so I have nothing fresh on official-sector buying or PBoC behaviour — treat that as a data gap rather than a quiet tape. The COT print from 2026-05-12 still shows managed money net long +171,622 against commercials net -210,258 with OI at 376,496 — that's a stretched spec long going into the rout, and unwind risk from here is asymmetric. Friday's COT will be the one to read.

Two scenarios

Conviction numbers below are qualitative — they reflect my honest read of where the tape is leaning. They are not back-tested probabilities.

Buy setup

• Trigger: 1h close back above 4,521.70 (daily pivot) on rising volume, ideally with DXY rejecting 99.50.
• Invalidation: 1h close below 4,455.00 (session low / weekly S1 zone).
• Target: 4,615.10 (weekly pivot), with 4,705.90 (weekly R1) as the stretch.
• Conviction: 40%.
• Rationale: 4h RSI 28 with the 1h MACD histogram turning up is the textbook setup for a mean-reversion squeeze, and weekly S1 at 4,465 plus the session low at 4,455 is the only confluence on the board worth defending. Stretched managed-money longs have a habit of producing one violent dead-cat before they actually capitulate. This is a tactical bounce, not a trend re-entry.

Sell setup

• Trigger: 1h close below 4,455.00, with DXY holding above 99.30.
• Invalidation: 1h close back above 4,521.70 (daily pivot).
• Target: 4,444.40 (daily S3) first, 4,374.20 (weekly S2) as the measured move.
• Conviction: 55%.
• Rationale: Every timeframe ≥ 1h is below its key EMAs, the daily MACD histogram is still expanding to the downside at -14.15, and the macro story (rising real yields, firm dollar, hawkish repricing) hasn't exhausted. Weekly S1 at 4,465 is the last natural buyer above the 200-day; through it, the next real bid is the weekly S2 zone, with the 200-day EMA below that as the line that defines whether this is a correction or a regime change.

Levels worth marking

Working from the snapshot's daily and weekly pivots, and tagging confluences.

• 4,725.80 — session high; must-recover for any "this was a flush" narrative.
• 4,705.90 — weekly R1; first real upside target.
• 4,615.10 — weekly pivot; the line that separates "bounce" from "regime back on."
• 4,537.20 / 4,521.70 — daily R1 / daily pivot; nearest overhead cluster.
• 4,490.80 — daily S1; minor.
• 4,475.30 — daily S2.
• 4,465.00 / 4,455.00 — weekly S1 + session low confluence. This is the level.
• 4,444.40 — daily S3; first downside target through the confluence.
• 4,374.20 — weekly S2; the measured-move target for the sell case.
• 4,363.73 — 200-day SMA per FXStreet; this is where the longer-term uptrend gets tested.
• 4,224.10 — weekly S3; tail risk only.

The 4,455–4,465 zone is doing too much work to ignore. If it goes on a 1h close, the next defended level is roughly 80 points lower.

Calendar / catalysts

Forexfactory's calendar fetch returned 403, so I don't have a clean event list — treat the below as the public schedule everyone already knows about, not a calibrated read.

• FOMC minutes — flagged by multiple desks as the near-term focus. Will be parsed line-by-line for how the committee is framing an oil-driven inflation impulse against growth softening. Any "patient" language is gold-positive; any "vigilant" language extends the sell.
• US flash PMIs — services in particular. A hot services print stacks onto the hawkish trade.
• Iran headline risk — the "two or three days" comment puts a binary catalyst on the tape. A de-escalation headline sends oil and yields lower simultaneously and gold bounces hard; an actual strike does the opposite of what most retail expects (yields up, dollar up, gold initially down on the real-yield channel before any haven bid can establish).
• Friday COT — the 2026-05-12 print is now stale. If managed money was still adding into the high, the unwind from here is the story.

Sources cited

• https://www.fxstreet.com/markets/commodities/metals/gold — analyst framing and the 200-day SMA reference at 4,363.73.
• https://tradingeconomics.com/commodity/gold — Iran "two or three days" quote and the Strait of Hormuz / oil chain.
• https://www.investing.com/commodities/gold — technical-summary read (Strong Sell on daily/5h, Sell on weekly/hourly) and the intraday range.
• Reuters commodities, Kitco front page, and Forexfactory calendar were attempted; Reuters and Forexfactory failed to return content, and Kitco returned only a header. Not cited.

(not financial advice)