Why 194,000 Jobless Claims Outweighed a Wider War

By every headline measure, Thursday was a gold day. Houthi drones had hit Saudi Aramco's Shaybah oil field on Wednesday, killing two people, and Washington pulled some diplomats and their families out of its missions there. The New York Times reported that the US and Israel have drafted plans for renewed strikes on Iran after next month's midterms, short but intensive. Brent rose 3.1% to $112.17, its highest since May (Investing.com, Anadolu).

Gold fell 2.2% to $4,051.55 by 11:44 ET, kept going to a ten-week low — $4,034 on FXStreet's feed, $4,021 on the desk's own series — and closed near $4,028, with the high of $4,126 set in the first hours of Asian trade (CNBC, FXStreet).

The number that broke it was not in the Middle East. At 12:30 UTC the US reported 194,000 initial jobless claims for the week to 17 October against a 200,000 consensus, with the four-week average at 197,750 (FXStreet). Gold spiked to $4,091.74 on the release, then closed the 13:00 hour at $4,054.25 and made lower closes all afternoon.

On a day when the war widened, the labour market did more to gold than the war did.

The Fed has no labour-market reason to stop

The committee meets on 27-28 October, with the decision due at 18:00 GMT on the 28th, after September's first increase since 2023 (Investing.com). Markets price an 8% chance of a move next week and roughly 80% for December (FXStreet, CNBC), and Goldman Sachs expects one more quarter-point, taking the peak range to 4.00%-4.25% (CNBC).

A 194,000 print with a four-week average under 200,000 is not a labour market that forces a pause. It is one that lets Chair Kevin Warsh keep a hawkish tone at the press conference and leaves December on the table, and his remarks the previous Friday had already moved expectations little (Anadolu). A labour market this tight keeps the real cost of holding a metal that pays nothing rising, just as an energy shock gives the committee a second reason to stay restrictive (CNBC).

The Fed speaks on 28 October and core PCE lands on 29 October (FXStreet). There is no data escape hatch before the press conference.

The yield link broke

The cleanest evidence that this is a policy trade rather than a haven trade sits in the bond market. The 10-year Treasury yield fell 15 basis points to 4.963%, its lowest since 23 September — and gold fell with it, which is not how the textbook works (FXStreet).

Two things break that link. The haven money did not leave the system; it went into the instrument that pays the coupon, and a flight into Treasuries is not a bid into bullion when the Treasury pays 4.96% and the metal pays zero. And with Brent at $112 the inflation compensation inside that nominal yield is rising, so a 15bp fall in the headline rate is not a 15bp fall in the real rate. On Thursday the real yield barely moved while the nominal one dropped, and gold traded the real one.

It was not a dollar story either. The dollar index held 101.62, down 0.02% (FXStreet). Gold fell 2.3% against a flat dollar and lower nominal yields. Opportunity cost is the only explanation left.

Who is actually selling

September, the last time this configuration appeared, shows who did the selling. Gold fell more than 8% that month while global gold ETFs took in more than 70 tonnes — a pairing the World Gold Council says it had not observed before, with the price decline in the 93rd percentile of monthly falls and the inflow in the 77th percentile (WGC). The selling was in the futures book: COMEX managed money fell by 84 tonnes equivalent and the spreading category by 156 tonnes.

The owners were not selling. The leveraged paper was. The marginal price-setter trades the Federal Reserve's reaction function, not the map of the Gulf, so a widening war cannot lift the price while that book is short a December hike.

The sticky bid is somewhere else: central banks added 23 tonnes in September, the largest monthly addition since October 2023, as the desk wrote in Tuesday's article. Reserve managers do not care about 15 basis points on the 10-year: their buying sets a floor, not the price of any particular Thursday.

The complex confirmed it

The rest of the metals agreed, which separates a policy trade from a single-market accident. Silver fell 3.7% to $55.86, leaving the gold/silver ratio near 72.5. Platinum lost 2.6% to $1,586.22 and palladium 3.2% (Kitco/Reuters, Kitco). When the whole complex falls while crude rallies, the market is not buying insurance; it is repricing the discount rate on every non-yielding asset at once.

The levels this leaves

• $4,073 — the 14 October low, the floor that broke. It is now the first ceiling on any bounce.
• $4,092 — Thursday's last lower high. An hourly close above it ends the intraday downtrend.
• $4,000 — the round number and the next shelf, where the bear case has been pointing all week (FXStreet).
• $3,929 — the desk's measured move, arithmetic rather than prophecy: the October box ran $4,073.32 to $4,216.89, $143.57 wide, and a break of the floor projects the same distance below it.

Those four are how the desk grades itself, not a forecast to hide behind: every level published here gets scored on /track-record, this week's failures included.

What would change it

Base case, hawkish hold. Warsh keeps December live, core PCE stays firm on energy pass-through. Rallies fail at $4,073 and $4,000 gets retested, which is why the desk's session-levels work has been selling bounces, not buying dips.

Dovish surprise. Core PCE comes in soft, or Warsh signals December is data-dependent rather than pre-committed. A reclaim of $4,073 opens $4,126, then the desk's 20-day average near $4,150.

Labour crack. A four-week claims average above 250,000 is the desk's line, not a data point: the level at which the inflation problem becomes a growth problem and December gets priced out. Nothing in Thursday's print comes close.

Physical interruption. Crude above $120 with barrels actually lost through Hormuz is a growth shock, the one war outcome that helps gold because it forces the committee to weigh growth against inflation. The market would have to reprice the Fed first; a wider war alone has not been enough for two months, which is how last week's war-premium trade went wrong.

Takeaways

• Gold is trading the Fed's reaction function, not the Gulf. Until the labour market cracks, war headlines are a source of volatility, not a source of bid.
• Thursday's print removed the one argument for a pause that would have helped the metal, and the futures book sold it.
• Watch real yields, not nominal ones: the 15bp fall in the 10-year was eaten by rising inflation compensation at $112 Brent.
• The record on this range is public and unflattering — three stops, two first targets and one flat in the six graded signals, with the reasoning attached on /track-record.

Sources

• CNBC — gold at a two-week low
• Kitco/Reuters — gold hits a two-month low
• Kitco — gold and silver pressured
• FXStreet — gold slides as haven demand shifts
• Investing.com — Gulf tension lifts oil, sinks gold
• Anadolu — gold below $4,100 on oil and inflation
• World Gold Council — Weekly Markets, 5 October 2026

Risk note. This is analysis, not financial advice. Leveraged gold trading can lose more than the deposit; position size from the stop, not the target. Prices are the desk's read of the Investing.com XAU/USD series, which differs from other feeds by up to $13 in a session — the band is stated wherever they disagreed.