XAUUSD signal: sell the $4,054 retest, target $3,995

Gold lost the October floor on Thursday and closed within $7 of the session low. Spot prints $4,028 at 20:00 UTC, down $97 from Wednesday's $4,125.50 on the Investing.com series — a 2.4% day that other feeds quote nearer 2.7% because they settle off a $4,140 reference. The high was $4,126.16 in the first hourly bar of the day; the low was $4,021.26 at 19:00 UTC. The $4,070 shelf that held since late September is gone, and so is the $4,105 low that Wednesday's plan was built around.

Start with the honest part. Wednesday's signal rested a sell limit at $4,144 and the tape never reached it — Thursday's high was $4,126.16 — so the desk's order sat unfilled while gold fell $100. The level was right and the location was wrong. An entry 0.2 ATR above spot looks disciplined on paper and trades only if the market obliges with a specific bounce. Today's entry sits 0.34 ATR above spot, inside the rule that an order must be within half an ATR of the market or be published as watch-only. That is the fix, not a new forecast.

What Thursday's tape did

All times UTC, hourly closes on Investing.com's XAU/USD series:

• 00:00–02:00: $4,113 to $4,126. The Asian session held the old floor.
• 03:00: the break — $4,116.24 down to $4,090.09, closing $4,092.49.
• 05:00: first test of the range bottom, low $4,066.86.
• 06:00–10:00: five hours of chop between $4,078 and $4,102. Buyers stepped in at the old low twice and failed to make a higher high either time.
• 11:00: $4,057.45. The floor was given up.
• 12:00: the claims print, a $34 spike to $4,091.74, straight back into the broken shelf, closing $4,088.09.
• 13:00: $4,054.25. One hour gave back the whole spike and $5 more.
• 16:00–19:00: $4,034.61, $4,048.43, $4,046.42, then $4,021.26 into the close.

The 12:00 bar is the session in miniature. A macro release lifted gold 34 dollars and the next hour took 35. When a market cannot hold a bid into news that should help it, the path of least resistance is down.

Why gold fell with a war on

This was not a quiet day for the world. Brent traded above $110 and WTI near $98 after a drone strike on Saudi Aramco's Shaybah field killed two people, with a separate attack on King Fahd International Airport in Dammam killing one (Commodity News). Gold normally bids on that. It fell 2.4% instead.

The mechanism is inflation, not fear. Oil above $110 is an inflation shock that keeps the Fed restrictive, and restrictive policy is the one thing gold cannot price around at 3.7% headline CPI (FXStreet). Initial claims came in at 194,000 for the week ending 17 October — a three-month low and below the 200,000 consensus — a labour market that gives the Fed no reason to soften (Kitco). The 10-year yield fell about 15 basis points to 4.96% and gold still could not hold, because the haven money went into Treasuries and the yen: EUR/USD held $1.1241 while USD/JPY dropped 0.7% to 157.10 (Reuters).

Fed pricing does the rest. Markets put roughly a one-in-ten chance on a hike at the 27–28 October meeting and near 80% on one in December. September PCE lands on 29 October. In that world a war premium in gold has to compete with a policy premium in cash, and for two months cash has been winning.

The trade

• Instrument: XAUUSD spot, MT5 gold
• Session: Friday 23 October, Asia through the New York close (00:00–21:00 UTC)
• Bias: bearish below $4,092 — the four-month range broke, and price sits under the 20-day and 50-day averages at $4,150 and $4,309
• Entry: $4,054 sell limit
• Stop loss: $4,113
• Take profit 1: $3,995 — 1.0R, close half
• Take profit 2: $3,936 — 2.0R, close the rest
• Risk-reward: 1:1 to TP1, 1:2 to TP2, $59 of risk per unit
• Do not take it if: gold closes an hourly bar above $4,092 before the order fills, or the fill arrives on a bar that closes back above $4,070
• Time stop: flat by 21:00 UTC Friday. No weekend carry into FOMC week.

Why these levels

Entry. $4,054 is the 50% retracement of the 13:00 high $4,088.13 to the 19:00 low $4,021.26, and it sits inside the 15:00 hour's breakdown block, $4,049.78 to $4,062.80. That is $25.50 above the 20:00 print, or 0.34 × ATR(14) — inside the half-ATR rule. The order sells the first bounce into the shelf that just failed; it does not chase the low. Every level here is mapped the same way as the desk's other session-level posts.

Stop. ATR(14) on daily bars is $75.87, so the 0.7 floor is $53.10 and $59 clears it at 0.78 × ATR. The stop sits $7.90 above Wednesday's sweep low of $4,105.10, which is 0.10 × ATR of buffer beyond the structural level on the Investing.com series. That buffer is not decoration: feed disagreement of $5 has decided this desk's verdicts before, and Kitco runs $5–8 below the aggregate feeds. Stop is nowhere near the last hourly bar's range of $4,028.13 to $4,029.50.

Targets. TP1 at $3,995 is 1.0R and lands on the $4,000 round number and the 29 July low of $3,996 (FXStreet). TP2 at $3,936 is 2.0R and sits $5 under the year-to-date low of $3,941. No two-to-three-R first target this time. A first target has to be reachable after the fill, or the plan is a lottery ticket with a stop attached.

What the R:R demands. Half off at 1R and a 2R runner pays +1.5R on a full winner against −1R on a loser, so the plan breaks even at 40% if both targets print. Scratch the runner at entry and the average winner is +0.5R, which needs 66.7%. Realistically it needs a TP1 hit rate near 50% with a third of those continuing to $3,936. If that sounds demanding, that is the correct reading: this is trend continuation after a $105 down day, into a round number, four sessions before an FOMC.

Invalidation and the bias flip

The short thesis dies on an hourly close above $4,092, the session's last lower high. Above it, the next resistance is $4,105, then $4,120, then the falling 20-day near $4,150 — which is where a failed bounce becomes a bigger trade rather than this one. A daily close back above $4,070 puts price inside the old range and voids the setup: cancel and wait for the next levels post.

Two macro switches matter. Brent back under $105 removes the inflation impulse that is holding gold down; a soft flash PMI at 13:45 UTC would remove the growth strength that supports hawkish pricing. Neither flips the trend alone. Both together would.

Execution notes

• The order is a resting limit, so it fills on the first touch of $4,054, not on a "rejection candle". If price trades through the zone, the desk is short.
• Retail spread on gold at the Asian open runs 20–35 cents. The $59 stop assumes a fill within a dollar of the level and no more than $2 of slippage on the stop.
• Unfilled by 13:30 UTC, cancel ahead of the flash PMI. A resting order into a growth print is a bet on the number, not on the structure.
• One order. If the setup invalidates, stand aside until the next session rather than moving the level to somewhere price already is.

Sources

• Reuters — gold drops as oil rally reinforces the higher-for-longer outlook
• Kitco — gold under pressure as jobless claims fall to a three-month low
• Kitco — gold pressured despite lower yields as Hormuz attacks lift crude
• FXStreet — gold slides as haven demand shifts to US Treasuries
• FXStreet — XAU/USD outlook: bears regain control and eye $4,000
• CNBC — gold hits a two-week low as the oil rally fuels inflation worries

Risk note. This is analysis, not financial advice. Leveraged gold trading can lose more than the deposit; size the position from the stop distance, not from the target. Levels are the desk's own read of the Investing.com XAU/USD series and may differ from your broker's feed by several dollars — the stop is the level that matters, and it is priced with that gap in mind.