XAUUSD signal: buy the $4,045 dip, out before CPI

The market is shut until 23:00 UTC. Friday's last prints put spot at $4,066.63 on the Investing daily bar and $4,070.59 at the GoldPrice close, so the weekend reference is a band of $4,066-4,074 (Investing, GoldPrice). That is +4.93% on the week from $3,875.44 and $80 under Thursday's $4,147.02 high (Kitco's weekly recap).

The plan bids $4,045, risks $60 to $3,985, targets $4,105 and $4,165, and is flat at 13:15 UTC Tuesday, fifteen minutes before CPI.

Friday was a payroll hole with a $37 repair

Gold opened at $4,087.48, printed $4,106.94 inside the first hour - the day's high was $4,107.50 - and then sold for seven hours into the print. The 01:00 UTC bar ran $4,091.51 down to $4,056.59.

At 13:30 the payrolls number turned a grind into a hole. The five-minute bar opened at $4,060.22 and printed $4,029.07, a $31.15 move inside five minutes, and CNBC's feed records the low at $4,033.29 - a $4.22 disagreement between two feeds, which is why every level here is named to a feed (CNBC).

The repair matters more than the hole. The 14:00 bar held $4,030.18, the 15:30 bar held $4,044.50, the 18:00 bar ran to $4,077.58, and the session closed at $4,066.63 - $37.56 above the low. Two tests of the same shelf, seventy minutes apart, and both were bought.

The floor is three sessions' lows inside $9

The shelf this plan is entered above is not Friday's low alone. The 22 October low was $4,020.99, the 28 October low $4,024.36, and Friday printed $4,029.07 and $4,029.95 - three sessions, four tests, inside $9. The last time gold lost that shelf, on 22 October, it did not stop until $3,870.17, which is where this month's rally began.

Then the weekend handed the dip a buyer. The People's Bank of China added 830,000 ounces, about 25.8 tonnes, in October - its 24th consecutive month of purchases and, at roughly 129 tonnes year to date, its heaviest annual run since it restarted buying (Bloomberg). October's addition was the largest single month since September 2023, taking reserves to 78.3 million ounces, about 2,435 tonnes (China Daily).

Read that with the calendar: gold fell almost 7% in October and the official sector bought more that month than in any month since 2023 - a buyer whose size rises as the price falls, sitting directly under the level this plan is entered above. It does not make $4,000 a floor; it means a sweep through $4,030 has to find someone else willing to sell.

Kitco's weekly poll had Wall Street split evenly across higher, lower and sideways, with retail leaning long - a split tape is a range, traded at its edges.

The ceiling is the dollar, at a two-week high

Friday's dollar move was the mirror of gold's. The index closed at 102.30, up 0.7% on the day to its highest since 29 October (CNBC, Portfolio Terminal).

Yields went with it: the 10-year finished the week at 4.86% after 4.94% the week before, the 2-year at 4.51%, and the 10-year TIPS real yield at 2.63% (Metal Pilot's weekly table). A quarter-point hike in December is still priced at 74.8% on CME's FedWatch and 77% in Metal Pilot's table - the debate is only how far.

The flows say where the money went: GLD shed 0.57 tonnes on the week to 1,040.51 tonnes while SLV added 70.1, and silver closed at $51.94, +5.72% on the week, leaving the gold-silver ratio at 78.4 (GoldPrice).

So the week's +4.9% is not a rates trade. It is the fiscal trade: Democrats took both chambers, gridlock was read as less borrowing, and gold was bought alongside a firmer dollar and rising hike odds. That combination is real, but it is unstable - it reverses the moment an inflation print puts the real yield back in charge.

Tuesday's CPI is the clock

October CPI lands at 13:30 UTC on Tuesday 10 November (08:30 ET), with September's 3.7% headline and 2.6% core as the base (BLS). PPI follows on Friday, claims on Thursday, so the week has one dominant print and three footnotes.

The desk is not trading that print. A hot core does three things at once - lifts the real yield, firms the dollar, prices a December hike closer to certain - and the 14 October CPI took $97 out of gold between its high and its low. A soft core does the reverse and re-opens $4,147 and the $4,177-4,199 shelf gold fell out of on 22 October.

That binary is why this plan has a clock rather than a price target.

The plan

• Bias: higher while $4,030 holds; a daily close under $4,000 ends the sequence
• Entry: $4,045 buy limit
• Stop loss: $3,985
• Take profit 1: $4,105 (close half)
• Take profit 2: $4,165
• Risk-reward: 1:1 to TP1 and 1:2 to TP2 - $60 at risk, $60 and $120 paid
• Time stop: flat at 13:15 UTC Tuesday 10 November, no position through 13:30
• Do not take it if: gold opens under $4,030, or spot is above $4,090 at the 23:00 UTC reopen
• Instrument: XAUUSD spot
• Session: Monday 9 November 2026 SGT, 04:00 SGT to 13:15 UTC Tuesday

The arithmetic, from the desk's own daily bars. ATR(14) is $83.57 across the fourteen completed sessions to 6 November, with the range floor at $37.90 and the ceiling at $156.96. The stop at $3,985 is therefore 0.718 × ATR from the fill - $60 of risk - and the entry sits 0.30 × ATR ($25) under the closing band. The stop is $36, or 0.43 × ATR, beneath the $4,020.99 shelf, so a sweep of the obvious stops under $4,030 does not reach it. TP1 at $4,105 is $2.50 under Friday's high, so the first target does not need a new high to pay. TP2 at $4,165 is $12.59 under the 15 October high, the first shelf above the range.

Break-even on that geometry is a 40% strike rate if both targets are reached - average win 1.5R against a 1R loss - and 66.7% if the runner is scratched at entry and every winner pays 0.5R. At a 45% hit rate the expectancy is +0.125R. This plan needs four wins in ten, and it cannot get there by moving the stop closer.

Why this entry, and not a deeper one. Friday's plan is the cautionary tale. The $4,080 buy limit filled at 01:00 UTC when that bar ran $4,091.51 to $4,056.59, and the order was flattened at the 13:15 time stop at $4,057.59 for -0.39R: the payroll hole that would have decided it in the buyer's favour arrived fifteen minutes after the exit. A fill is not a trade; the exit did the damage.

Today's entry sits at a price that printed twice on Friday - the 15:30 bar's low was $4,044.50, the 16:00 bar opened at $4,054.79. Thursday opened at its low ($4,059.93 against a $4,057.82 low) and never came back; Friday opened near its high and sold $78. The last two sessions behaved in opposite ways, so no "the market always dips" rule applies - which is why the entry is $25 away rather than $50. If it does not fill, that is a 0R and the ledger will say so.

What makes it wrong

• A gap through the shelf. If gold opens Monday under $4,030 and holds there, the three-session floor has failed, and the order is cancelled rather than filled into a break. The stop at $3,985 is insurance, not a plan for that tape.
• A gap above the plan. If spot is above $4,090 at the 23:00 UTC reopen, the entry is more than 0.5 × ATR away and the order is withdrawn - chasing a vertical open with a $60 stop breaks the risk geometry that makes this plan break even at 40%.
• A close under $4,000. That is the invalidation of the bias, not just of the order: the 22 October break started from exactly this shelf.
• A quiet Monday. The likeliest failure is not a loss but another 0R - a session that marks time between $4,050 and $4,090 and never gives the fill. The desk has booked four of those this month, and the ledger counts them as the nothing they are.

The structure read sits on the signals board, the previous plan it replaces is here, the official-sector bid runs through the central banks hub, and every verdict - including the no-fills - is on the track record.

Sources

• Investing.com - gold daily, hourly and five-minute (ATR, session range, the 13:30 five-minute hole, hourly shape)
• GoldPrice - 6 November close (close, week, silver, gold-silver ratio)
• CNBC - gold drops 1% as payrolls lift the dollar (intraday low, dollar, FedWatch)
• Kitco - weekly recap and survey (weekly high, split Wall Street vote)
• Bloomberg - China extends gold-buying streak to two years (830,000 oz, 129 t year to date)
• China Daily - biggest monthly addition since 2023 (78.3 m oz, reserve share)
• Portfolio Terminal - DXY 6 November close (102.30)
• Metal Pilot - weekly market table (yields, TIPS real yield, ETF flows, Fed pricing)
• BLS - CPI release schedule (10 November, 08:30 ET)

Not financial advice. Levels are the desk's own read from the feeds named above; a $60 stop on an $84 ATR instrument is a wide stop, and position size must be cut to fit it rather than the stop moved to fit the size.