4,000 has flipped from support to resistance. Below here the road has three steps, and 3,903 is the hard one.
The short version: spot is 3,984.2 (snapshot 07:16 Beijing time, 12 November 2026, Tonghuashun), under 4,000. As long as it cannot close back above 4,000, the primary path is a test of 3,960. If 3,960 gives way, the double-top measured target at 3,903 comes into range.
What this decline actually broke
Spot gold: open 4,038.78, high 4,066.81, low 3,982.43, close 3,987.24, down 1.35% (Tonghuashun, 12 November 07:16). Range 84.38.
December COMEX: settled 4,000.50, down 0.74% (Kabutan citing MINKABU, 12 November 03:58), settlement print 3,992.00 (Dow Jones, 11 November 14:04 ET).
Three things broke:
First, the 10 November low at 4,013.20. That was the floor of the previous leg down. Losing it invalidates the last line of longs.
Second, the 6 November low at 4,026.038. This is not an ordinary low. It is the neckline of the double top. A neckline is the confirmation line; a close beneath it turns the pattern from possible into valid.
Third, the 4,000 round number, and it broke on a close. A wick through a round number and a close below it are different events. A wick is a probe. A close is turnover.
Stack the three and every long built above 4,000 is now underwater.
The double top cashes in today
I ran the double-top math on 11 November. Today the numbers have moved, so here it is again.
The two peaks: 4,148.60 on 9 November, 4,146.696 on 10 November. Take the higher one, 4,148.60.
Height: 4,148.60 minus 4,026.038 equals 122.56.
Measured target: 4,026.038 minus 122.56 equals 3,903.48.
The logic is plain: the height of the pattern, projected down from the neckline. It is not a prophecy. It is a scale for "if this holds, roughly where does it go".
Yesterday that scale was 120 dollars away, abstract. Today it is 80 dollars away, a route.
And 3,903.48 is not isolated. The 20-day Bollinger lower band sits near 3,900 (Huitong, 11 November 00:34), and the lower edge of Wang Yang's 3,900 to 4,200 range is 3,900 (Sina Finance, 11 November 09:02). Three sources on one price is what makes a level key rather than a guess.
Three scenarios, each with its own kill switch
A single call is lazy, because the market does not follow scripts and people panic when it does not. So: three paths, each with the condition that proves it wrong.
Scenario A: rebound repair, roughly 30% in my book.
Trigger: claims at 21:30 come in materially weak (prior 189k, FRED, week ending 31 October), or the dollar index slips back below 102.5.
Path: 4,050 first (StockWire X, 12 November 00:00), then 4,066.81, and 4,080 on a strong move (4-hour Kijun, Traders Union, 12 November 01:08).
Kill switch: the rebound stalls at 4,050 on shrinking volume and rolls over through 3,982.43. That means it was short covering, not buying.
Scenario B: grind lower into 3,960, roughly 45%, the primary path.
Trigger: claims stay below 190k and the dollar holds above 102.9.
Path: 3,960, the divide named by both Myfxbook and StockWire (11 November 19:13 and 12 November 00:00), then 3,919, the 4 November low, then 3,903.48.
Kill switch: two consecutive closes above 3,960 and a reclaim of 4,000. That would mark the move as a shakeout.
Scenario C: break 3,900 and accelerate, roughly 25%.
Trigger: a close below 3,900 with the dollar through 103.2.
Path: 3,860, the next-leg trigger WGC flagged, then 3,800, Myfxbook's extension target (11 November 19:13).
Kill switch: 3,900 broken and reclaimed the same session, leaving a long lower wick. That is a false break, usually with central bank or physical buying underneath. WGC's Q3 figure of 321 tonnes, up 42% year over year, is a real bid.
Where to stand
To be precise about terminology: the following are reference levels, not key levels. A single source gets called a reference; two or more converging sources get called key.
- Short side: do not chase at the current price. Sell a rebound into 4,050 to 4,066, stop above 4,080, first target 3,982, second 3,960. Entering at 4,058 puts risk at 22 dollars against 98 dollars of reward, roughly 4.4 to 1. The value of that trade is the waiting. Chasing shorts at 3,984 is the worst risk-reward on the board.
- Long side: a small probe near 3,960, stop below 3,945, target 3,998. Risk 15 against 38, about 2.5 to 1. This is a counter-trend trade, one attempt only, no averaging down.
- No-trade zone: 3,984 to 4,000. Resistance above, support below, and the middle is the hardest place to be.
Both sides of the information
Bullish: central bank buying at 321 tonnes in Q3, up 42% (WGC); UBS holding a year-end 2026 target of 4,600 (Huanqiu, 11 November 10:23); physical bids in the 3,800 to 3,900 zone.
Bearish: full December hike pricing with 50bp bets live (Chinanews, 12 November 07:04); the dollar from 102.15 to 102.908 in two days (Xinhua Finance, 12 November 06:10); the 10-year at 5%; and equities pulling capital (Nasdaq +1.6%, Wind).
The bearish side is the accelerator right now. The bullish side is the brake. Do not guess when the brake gets pressed while the accelerator is down.
Tonight's claims print at 21:30 is the first read on how hard that pedal is being pressed.
If you still hold longs entered above 4,000, do you add at 3,960 or wait for 3,903?
Data collected 07:16 to 07:20 Beijing time, 12 November 2026. Sources: Tonghuashun, Wind, Kabutan, Dow Jones Market Data, FRED, StockWire X, Traders Union, Myfxbook, WGC, Huitong, Sina Finance, Huanqiu, Chinanews. Structural projection, not investment advice, no return guarantee.
Related
Check out more trading insights in the Trading Journal.