For three weeks, gold had one line that kept holding. On 30 October it found buyers near $3,870. On 12 November, after a hot inflation print and a jump in Treasury yields, it dipped to $3,852 — and bounced again. Two tests, two rebounds. The kind of zone that ends up on every chart labelled "support".
On Wednesday it gave way. Gold slipped through $3,850 in Asian trading, kept falling into the London session and traded down to roughly $3,745–3,770 — the lowest level since late September 2025, and about a third below the January record near $5,600.
This post is not a call on where gold goes next. It's about what happened to that floor, and why *how* it broke matters more than the fact that it broke.
## How we got here
**Week of 9 November.** Gold opened the week firm, rebounding to around $4,140 on cheaper oil. Then Tuesday's October CPI landed: headline +0.4% m/m, core +0.3% against 0.2% expected — the third straight 0.3% core print. That same morning Fed Governor Waller had said a core reading of 0.3% or higher could justify a half-point hike in December. Gold lost 2.6% on the day.
On Thursday the 10-year Treasury yield pushed towards 5.10% and Brent rose about 3% as hopes for a Middle East settlement faded. Gold fell another 2.8% and printed the $3,852 low. Friday's softer PPI gave it a small bounce, but the week still closed down 4.56% near $3,884 — the worst week since early June.
**This week.** Monday and Tuesday held in the $3,860s. On Wednesday Waller went further: on CNBC he described inflation as worsening and spreading, and argued for a 50 bp hike at the December meeting. The dollar index climbed to its highest level since March 2025. That was the push the floor couldn't absorb.
## A floor is not a promise
Support is simply the place where buyers showed up last time. The useful question is not *where* they showed up, but *why*.
On 30 October and 12 November, gold was falling because the market was repricing the Fed. Both bounces were bets that the repricing had gone far enough: a 25 bp December hike was priced in, Chair Warsh was refusing to give forward guidance, and nothing new seemed likely to arrive before the meeting.
Wednesday changed the question. It's no longer *whether* the Fed hikes in December — that has been close to fully priced since the CPI report. It's *how much*, and *how many more times*. Futures were already leaning towards another hike in January. When the reason a level held disappears, the level stops meaning anything. The line is still on the chart, but nobody is defending it anymore.
That's the distinction I keep coming back to on this blog: a level that **breaks** versus a level that is **taken**. A spike through support in the seconds after a headline is a break — liquidity vanishes, price travels, and very often it comes straight back. A level is taken when the reason behind it has gone and price stays on the other side.
## Three checks before calling it "taken"
**1. A close, not a wick.** An intraday slide in London is not the same thing as a daily close below $3,850. Where New York finishes the session matters more than the low it printed in the morning.
**2. The retest from below.** If $3,850–3,870 has really flipped, the next rally should stall there. Price coming back up into the old floor, failing, and rolling over is the confirmation. A rally that reclaims the zone, closes above it and then holds above $3,900 would say the opposite — that Wednesday was a flush, not a new regime.
**3. Who's on the other side.** The structural bid hasn't gone away. According to the World Gold Council, central banks bought about 321 tonnes in Q3 — up 42% year on year — and 675 tonnes year to date. That demand is slow and price-sensitive. It doesn't stop a repricing, but it tends to appear on dips. Western ETF investors, on the other hand, have been sellers as real yields rise. If a floor forms again, it will be because official buying absorbs that selling, not because a line on a chart says so.
And one more thing that matters here: **time**. This break was not one candle. It was a grind across the Asian and London sessions, after two days of pressing against the level. A grind like that looks more like "taken" than a news spike does — but it still has to pass the first two checks.
## The macro backdrop in one paragraph
Gold doesn't trade inflation; it trades real yields and the dollar. With the 10-year yield around 5.1% and the dollar index at multi-month highs, every hot data point raises the cost of holding an asset that pays nothing. Price is below its 20-, 50- and 200-day averages. Until yields or the dollar turn, rallies in this market are more likely to be corrections inside a downtrend than the start of a new leg up.
## Levels I'm watching
- **$4,140–4,150** — the early-November high, where the last rebound died
- **$4,000** — the psychological pivot, lost on 11 November
- **$3,900–3,935** — where Friday's PPI bounce ran out of steam
- **$3,850–3,870** — the broken floor; now the retest zone
- **$3,745–3,770** — Wednesday's low area
- **$3,700** — the next round number, with no recent trading history around it
## What comes next
The FOMC minutes from the October meeting are due this evening. That was the meeting where the Fed held at 3.75–4.00% with two dissents in favour of a hike, so the interesting part is how many other participants leaned the same way, and whether anyone discussed the size of the next move.
After that the calendar thins out until the 8–9 December FOMC. One detail is easy to miss: the November CPI report is released on 10 December — the day *after* the decision. The Fed will decide on data the market already has. That makes Fed speakers over the next three weeks unusually important, and it makes this exactly the kind of environment where headline spikes are frequent and most of them are not breakouts.
So no forecast here, only a description of the environment: trending, headline-driven, with a strong dollar. In that kind of market a broken level is more likely to be retested than instantly reclaimed — but "more likely" isn't a trade until price confirms it.
If you'd like to see how a rules-based breakout approach behaves in conditions like these, the full verified statistics are on my profile. Have a different read on the $3,850 level? Leave a comment — I answer everyone.
---
*Past performance is not indicative of future results. Trading gold and other leveraged instruments involves substantial risk and is not suitable for every investor. Results achieved when copying a signal may differ from the provider's results due to spreads, slippage and execution.*


