XAUUSD signal: sell the $3,945 retest, target $3,886

Gold rejected $3,970 for the third time in two sessions and gave back the whole of Monday's rebound. The desk sells the $3,945 shelf toward $3,886, stop above $4,000.

Levels in this article
  • $3,920.63
  • $3,895.59
  • $3,969.96
  • $3,971.15
  • $3,895.48
  • $3,926.70
  • $3,921
  • $3,970.45

Trade plan at a glance

Entry
$3,945 sell limit
Stop loss
$4,004
Take profit 1
$3,886 — 1.0R, close half
Take profit 2
$3,827 — 2.0R, close the rest
Risk-reward
1:1 to TP1, 1:2 to TP2
Do not take it if
spot is above $3,966 when you read this, or the reopen fills you above $3,950 — the arithmetic below assumes the $3,945 fill — or gold opens above $3,979…

Primary levels from this article — any conditional alternative is laid out in the body below. Sizing comes from the stop distance; confirm on your own feed before acting, spreads and gaps are real.

Tuesday undid Monday. Spot gold was $3,920.63 at the 20:00 UTC cut, −0.68% on the session, after a day that ran from a $3,895.59 low at 09:00 to a $3,969.96 high at 13:00 and then surrendered the whole $50. Gate's feed put the extremes at $3,971.15 and $3,895.48 — agreement to $1.19 on the high and eleven cents on the low (Gate) — and Kitco had spot at $3,926.70 at 19:05 (Kitco). The band at the cut is $3,921–3,927.

The ceiling has now been tested three times

The top of this range is not a guess. Gold printed $3,970.45 on Monday afternoon, $3,970.31 in Monday's 16:00 UTC bar, and $3,969.96 in Tuesday's 13:00 bar — three highs inside 49 cents of each other across two sessions, with no hourly or daily close above any of them.

Nor is it a random level. The 23 and 26 October session lows were $3,975.64 and $3,978.15: the floor of the range that held for a week before Friday's liquidation broke it. The old floor is now the ceiling, and price cannot even reach its underside — sellers sit five to eight dollars below it and take the bid every time. Saxo saw the same thing, gold "encountered renewed selling near USD 3,970, a former support level now acting as resistance" (Saxo).

JOLTS took the bid away in one hour

The 15:00 UTC release decided the afternoon. September job openings came in at 7.523 million against a 7.15 million consensus — a 333,000 beat on a prior month itself revised up to 7.190 million from 7.079 million, with the openings rate rising to 4.5% from 4.3% (Kitco, Investing.com). September factory orders added +1.8% against +1.2% expected. Gold went from $3,945.46 at the top of that hour to $3,923.14 at the bottom of it and never recovered the shelf.

The dollar index rose 0.19% to 101.86 and the 10-year yield added five basis points to 4.96% (Reuters via Euronext, Trading Economics). The 10-year real yield was already 2.62% on Monday, from 2.60% (FXEmpire). Real yields are what matter for a metal that pays nothing, and they have not fallen on any of the three days gold has tried to rally.

Rate pricing is where the sources disagree, so here is the range rather than a false point estimate: Barchart quotes 79% for a 25bp hike at the 8-9 December FOMC (Barchart); CNBC quotes 73%, down from 76% earlier in the day (CNBC). Roughly three in four. A 1.8% bounce does not change that arithmetic, and a 0.68% give-back does not either.

What Wednesday actually has to trade

  • Midterm results, counted through the Asian session. Prediction markets favour the Democrats taking both chambers, which is close to priced; a Republican sweep is the surprise that lifts yields and the dollar (FXStreet).
  • ADP employment and ISM services at 15:00 UTC — the last labour read before payrolls, with October manufacturing already at 55.5 for a tenth straight month of expansion.
  • Non-farm payrolls, Friday 13:30 UTC. Everything this week is a rehearsal for it.

Yesterday's order filled, and is not a win yet

Sunday's plan rested a $3,966 sell limit with a $4,034 stop (that plan). Tuesday's 13:00 bar ran to $3,969.96, so the order filled. It is not a win: TP1 at $3,898 printed at 08:00 UTC, five hours before the fill, and the lowest print after the fill was $3,919.35 in the 19:00 bar. At the 20:00 cut it marks +0.67R, with its time stop arriving tonight. The grading pass books it at 21:15 UTC and this article does not pre-write the verdict — a target printed before the entry is not a target hit, and the track record is the only place a verdict belongs.

What today changed is the shape of the box. Lower high, $3,969.96 against Monday's $3,970.45; higher low, $3,895.59 against Monday's $3,875.21; a third session compressed between $3,870 and $3,970, with the daily closes converging — $3,875.44 Friday, $3,947.67 Monday, $3,920.67 today. Compression under a broken floor in a market that lost 6.8% in October is a coil, and the coil's upper edge is the side with the resistance. Every signal on this board is a bet on which edge gives first.

Where the levels actually are

  • $3,979 — this plan's invalidation: 0.1 × ATR above the highest of the three rejection prints, and above the 26 October low of $3,978.15. Two hourly closes above it and the broken floor has been reclaimed.
  • $3,975.64 / $3,978.15 — the 23 and 26 October session lows. The old range floor, now the ceiling.
  • $3,969.96 – $3,971.15 — the three rejection highs, Investing.com and Gate respectively.
  • $3,945 — the shelf. The 15:00 bar's high was $3,945.46, the 14:00 bar closed $3,944.54: the last hour before JOLTS selling took control. This is the entry.
  • $3,920.63 — the 20:00 UTC cut.
  • $3,895.59 — Tuesday's low, the 09:00 bar; the 08:00 bar printed $3,896.42. A double floor inside one session.
  • $3,875.21 / $3,870.17 — Monday's and Friday's lows: the one-year low and the 50% Fibonacci extension of the January-June-August cycle (Forex.com).
  • $3,827 — 2R, and below the double bottom.

The plan

  • Instrument: XAUUSD spot, MT5 gold
  • Session: Wednesday 4 November — Tuesday 23:00 UTC reopen (07:00 SGT) to the 20:00 UTC cut (04:00 SGT Thursday)
  • Bias: lower while $3,971 caps it; two hourly closes above $3,979 end the sequence
  • Entry: $3,945 sell limit
  • Stop loss: $4,004
  • Take profit 1: $3,886 — 1.0R, close half
  • Take profit 2: $3,827 — 2.0R, close the rest
  • Risk-reward: 1:1 to TP1, 1:2 to TP2
  • Do not take it if: spot is above $3,966 when you read this, or the reopen fills you above $3,950 — the arithmetic below assumes the $3,945 fill — or gold opens above $3,979, which is the invalidation and not the entry.
  • Time stop: flat or unfilled by 20:00 UTC Wednesday

The arithmetic, so it can be checked

  • Stop distance $59 = 0.785 × ATR(14) $75.20, on fourteen completed daily bars from 12 October to 2 November. The 0.7 floor is $52.64; the stop clears it by $6.36. 30 October's $156.96 range is still inside the fourteen-bar window, which is why ATR is this wide at all; the stop widened with it rather than staying at a number that looked tidy.
  • $4,004 sits $32.85 above Tuesday's $3,971.15 high — 0.44 × ATR — and above the $4,000 round number. A textbook 0.1 × ATR buffer would put the stop at $3,979, only $34 from the entry: 0.45 × ATR, inside the noise band where $5.60 and $0.29 have decided this desk's stops before. Widening to $4,004 costs reward multiple and buys a plan one sweep cannot end.
  • Entry is $24.37 above the cut — 0.324 × ATR, inside the half-ATR rule ($37.60) and 46 cents under the 15:00 bar's high. The last two opens did the work this order needs: Monday's 00:00 bar ran $3,898.36 to $3,903.71 and Tuesday's ran $3,948.53 to $3,964.46 before both sessions sold off. A bounce into $3,945 is the normal shape of the Asian session here, not an exotic level.
  • Break-even win rate: 67% if the runner is scratched at breakeven after TP1, 50% at a 50/50 runner, 40% if TP2 prints every time. The simple version — whole position out at $3,886, no runner — needs 50%, the lowest bar this desk has published all month.
  • TP1 is $10.79 above Monday's $3,875.21 low, so the first target prints on a retest of the double bottom without the low having to break. TP2 at $3,827 does need the low to break, which is why it is the runner and not the thesis.

Execution: the spread runs $0.30-0.50 at the 23:00 reopen and widens for the first fifteen minutes, so a $59 stop is not $59 of realised risk on a thin fill. Investing.com is the reference feed — a $5 disagreement between feeds on the stop is an execution error, not a market outcome, and $4,004 is deliberately placed where no plausible feed prints it.

ISM services at 15:00 UTC is inside this trade's window. A hot print supports the plan; a soft one spikes gold, which is why the invalidation is a structural break at $3,979 rather than a guess. If the order has not filled by 14:45 UTC, cancel it and let the print pass — a resting limit through a data release is a coin toss at worse odds than the plan was priced at.

Sources: Kitco — JOLTS, CNBC — gold and the midterms, Saxo — market quick take, Barchart — dollar and rate odds, FXStreet — gold below $3,950, Gate — session extremes, Trading Economics — 10-year yield, Forex.com — the $3,870 pivot. Levels and ATR are the desk's own calculations from Investing.com daily and hourly bars.

This is technical analysis of a live market, not financial advice. Gold is moving $75 a day and a $59 stop is a real loss if it is hit — size the position off the stop distance, not the other way round. Levels are levels until the tape says otherwise.

Trading these levels? XM — MT5, gold, demo first paid link