XAUUSD signal: sell the $4,004 retest, $3,934 target

Gold climbs $4,067, gives all of it back, closes under $4,000

Wednesday gave the bulls one clean chance and they spent it by 15:00 UTC. Spot gold climbed to $4,066.77 in the European morning and finished the New York session at $3,983.55, down 1.2% (CNBC). December futures settled $3,992.00, $28.70 lower at -0.71%, which puts $113, or 2.75%, gone in two sessions — the largest two-day decline since 30 October (Dow Jones via Morningstar).

The desk's own series printed the same shape: the 00:00 UTC bar opened $4,033.95, the high came at $4,058.60 in the 05:00 hour, the low at $3,979.75 in the 19:00 hour, and the last print at 20:00 UTC was $3,981.25. Feeds disagree by up to $8 on the high, so treat the session as a $4,058-4,067 high into a $3,980-3,995 close rather than a tick.

Two days, one trade

Tuesday's CPI was the cause; Wednesday only removed the last excuse to buy.

• December is priced for a hike. Fed funds futures put 87% on a 25bp increase at the 8-9 December meeting and 13% on 50bp (FXStreet). Governor Waller had said before the print that a 0.3% core reading would make him consider a half-point; core printed 0.3% for the third month running (CNBC).
• The short end moved hardest. The 2-year rose 17bp to 4.74% and the 10-year 11bp to 4.99% on Tuesday (Saxo). The cash Treasury market was shut for Veterans Day, so Wednesday had no yield session to trade — and bullion fell anyway, which is the tell that this is positioning, not a data reaction.
• The dollar is at a 19-month high. DXY +0.63% at 102.85, its best since April 2025 (CNBC).
• The daily structure is a staircase down. A lower low for a third straight session, lower highs at $4,133.99 (9 November) then $4,066.77 (11 November), and price still $194 below the 50-day at $4,177 and $217 below the 100-day at $4,200 (FXStreet).

That is a trend, and the desk does not fade it. Every losing signal in October was a short into a rising market or a long into a falling one; the direction here is settled by structure, so the only question is where to sell, not whether.

Why a bounce-sell and not a chase

Selling $3,981 now is selling into support after a $154 fall from Monday's $4,133.99 high. The cupboard below is not empty: $3,973-3,975 (the 2-3 November tops that broke), $3,952, the 4 November low at $3,919 by FXStreet's count — $3,929 on the desk's series, so call it a $3,919-3,929 band — and $3,879-3,890 from 30 October.

What the desk wants is the level that failed today. The shelf ran $3,994-4,002 in the 15:00-16:00 UTC hours before the last leg to $3,979.75, the round number $4,000 sits inside it, and FXStreet names $4,000 as first resistance with the front month settling at $3,992. A retest is a 0.6% bounce, and Asia has not been opening on its lows: 11 November opened at $4,033.95 and rallied $25 to $4,058.60, and 10 November opened at $4,124.60 near its high. The order goes in at $4,004, two dollars above the shelf's top edge, so a wick through $4,000 fills it instead of missing by a dollar.

The failure mode this desk keeps repeating is a sell limit parked above spot that never fills while the move happens without it. So the clause is explicit below: if the session opens below $3,950 the shelf has gone without a retest, TP1 would already be $16 from the open, and the plan is void — not chased lower.

The plan

• Bias: short while $4,066 holds
• Instrument: XAUUSD spot
• Session: Asia and London, 00:00-13:15 UTC on 12 November
• Entry: $4,004 sell limit
• Stop loss: $4,074
• Take profit 1: $3,934 — 1.0R, take 50% off
• Take profit 2: $3,864 — 2.0R
• Risk-reward: 1:1 to TP1, 1:2 to TP2, blended 1.5R if both fill
• Do not take it if: the 00:00 UTC open is above $4,064 (the stop then sits inside the opening hour's range) or below $3,950 (the shelf is gone), or the limit has not filled by 11:00 UTC
• Time stop: flat at 13:15 UTC, before the New York equity open

Risk is $70, which is 0.98 x ATR(14) of $71.39 measured on the desk's PAXG daily bars. The reference series has changed: Investing.com now answers this host with a Cloudflare challenge, so the arithmetic is PAXG-based — but at Investing's last reading of $93.62 the same stop is 0.75 x ATR, so the plan clears the 0.7 minimum on either series. The stop sits 0.1 x ATR ($7.14) above today's $4,066.77 high, the print that must hold for the thesis to be wrong, and it is a mile outside the last hourly bar's $3,981-3,982 range.

Entry is $23 from spot, inside the half-ATR leash that keeps a resting order in the market rather than on a poster. TP1 at $3,934 sits just above the 4 November low band rather than inside it, so the target is a test of that low, not a demand that it breaks. TP2 at $3,864 needs the 30 October low to give way; it is the runner, not the plan.

The arithmetic you have to accept

A blended 1.5R breaks even at a 40% strike rate. The honest caveat sits inside that number: if TP1 fills and the runner is scratched at entry, the winner pays +0.5R against a -1R loss, and that book needs 66.7%. The desk's measured average win is +0.82R, which needs 49.1%. Take the 50% at TP1 into dollars, and keep the remainder only while $3,980 keeps rejecting — a runner that survives is the only thing that drags the average win toward 1.5R.

Sizing follows from the stop, and this one is honest about its floor. At 0.5% of equity a $10,000 account risks $50, which against a $70 stop is 0.007 lots — under the 0.01 minimum, so that account cannot take this trade inside its own rule and should sit it out. A $100,000 account risks $500 and carries 0.07 lots. The position-size calculator does the same arithmetic for any balance.

Yesterday's plan is worth recording honestly: the $4,033 sell limit filled at the 00:00 UTC open and the plan was flattened at 13:15 around $4,030 — a scratch. TP1 at $3,967 never came closer than $12.75 of the session low. That is the same mechanical failure as the October run in a milder form: a first target parked in noise instead of at a level.

What makes this wrong

• A reclaim of $4,066.77. That breaks the lower-high sequence and stops the trade out at the same time.
• DXY rolling over from 102.85. Dollar weakness is the one input that could turn a hot-CPI tape back into a buying opportunity; central bank demand gave gold a floor through the third quarter, as StoneX's Bob Haberkorn argued (Reuters via Mining.com).
• A dovish Fed speaker before Friday. Waller and Bowman are arguing for more tightening, which is the crowded side; one governor pushing back on December is a $30 bounce.

Signals are graded on the public ledger at /track-record whether they win or lose, and the setup logic lives on the /signals board and under /topics/real-yields. The CPI print that started this move is written up in October CPI runs hot.

Not financial advice. Levels are for the 00:00-13:15 UTC session on 12 November 2026 and expire at the time stop; a fill outside the stated open conditions voids the plan.

Sources: CNBC gold, Dow Jones via Morningstar, CNBC dollar, Saxo market quick take, FXStreet levels, Reuters via Mining.com.