Sunday review — start with a set of seemingly contradictory data points. On Friday (October 23), spot gold settled near $3,995.6/oz, down roughly 0.7% on the day — the first close below the $4,000 handle since July 29, with COMEX futures losing 4.05% on the week, the worst since June. Yet that same week delivered two headline-grabbing “buy-side” stories: the People’s Bank of China added 23 tonnes of gold in September — its largest monthly purchase since October 2023 — extending its buying streak to 23 months and lifting official reserves to 2,410 tonnes. And World Gold Council data shows global gold ETFs absorbed a record $31 billion of net inflows in Q3, pushing total holdings to an all-time high of 4,256 tonnes. Central banks are stacking, ETFs are absorbing — and price is collapsing. What is going on?
1. Dissecting the Order Flow: Allocator Buying vs. Trader Selling
To resolve the paradox, you first have to split “buying” into two entirely different species:
- Allocator flows (central banks + ETFs): the slow variable. Reserve diversification proceeds on quarterly and yearly horizons, and long-only ETF allocation rebalances monthly at best. These flows are stable and predictable, but small as a share of daily turnover — they do not accelerate because of one bad week, and they cannot double their bid on a single session.
- Trader flows (CTAs, leveraged longs, liquidation selling): the fast variable. This selloff has two engines. First, oil-driven rate expectations — Brent rose 6.24% on the week to $116.72, reinforcing sticky inflation and repricing a December hike (CME FedWatch now shows the probability of a 25bp December move surging from ~30% toward 80% — see our near-dove, far-hawk rate framework), which sharply raises the opportunity cost of holding zero-yield gold. Second, a liquidity squeeze triggered by the tech rout — as David Meger of High Ridge Futures put it, energy-driven inflation fears are the core of the selling, and margin calls on the equity side are forcing investors to sell gold indiscriminately to raise cash.
The key conclusion: when the intensity of fast selling exceeds the absorption rate of slow buying, price falls — even if the total tonnage of buying sets records. Allocators support the “floor,” not the “price.” This extends the ETF allocator-absorption divergence we tracked on October 20: smart money is catching, but it is in no hurry.
2. The Weekend’s Third Signal: Physical Buyers Flood the Market
Beyond allocators and traders, this weekend added a physical-market signal worth logging. On Saturday, October 24, Shenzhen’s Shuibei gold market was packed — shoppers joked there were “more people buying gold than buying groceries.” The data: after spot gold broke $4,000, major Chinese jewelry brands cut retail prices by another 6-12 yuan/gram overnight, with Lao Miao quoting 1,196 yuan/gram, Chow Sang Sang 1,200, and Chow Tai Fook 1,202. Retail gold jewelry is down roughly 100 yuan/gram in a month — and nearly 500 yuan below the year’s peak of 1,700.
Behaviorally, “demand created by the decline” is textbook price elasticity — wedding-driven necessity buyers and value-oriented savers are deploying purchasing power en masse below $4,000. Physical demand is also a slow variable: it cannot stop a fall, but it keeps thickening the cushion under price as the decline extends. Morgan Stanley’s head of metals strategy, Amy Gower, stated this week that gold enjoys strong bottom support at $4,000 — a call that rests precisely on this three-layer structural bid of central banks, ETFs, and physical consumers.
3. The Quant View: Three Signals to Track in a Mismatch Market
As a quant trader, I do not predict bottoms — I track signals. Three indicators belong on the dashboard right now:
- Oversold readings. The daily RSI has fallen toward 28, into technical oversold territory, with the stochastic pinned at low levels — the technical conditions for a short-term bounce are accumulating. But oversold is never, by itself, a reason to buy; it is only a necessary condition for downside momentum to decay.
- Level-based absorption tests. First support sits at $3,970 (Friday’s low zone), with deeper support at $3,900. Overhead, first resistance is $4,050-$4,070, then $4,100-$4,120. If $3,970 holds early next week with long lower wicks, the “allocator absorption” hypothesis gets tape confirmation; a clean break through $3,900 says trader selling is not yet exhausted.
- The event catalyst. Next Wednesday’s FOMC decision (early October 29, Beijing time) plus the September PCE print will referee this mismatch. A hot core PCE hardens the “oil → inflation → hikes” chain and sets up a second leg down; a soft print unwinds December hike pricing and lets the oversold bounce run. The full three-node calendar and position rules are in Saturday’s super-week playbook and will not be repeated here.
A pragmatic base case, echoing Green Dahua Futures chief expert Wang Jun: with high oil prices, December hike expectations, and global deleveraging all in place, gold likely stays under pressure through late October and early November, with Q4 shaping up as a wide range between a floor and a ceiling — a core band of $4,000-$4,400. Under that framework, my own response is mechanical: my monitoring stack keeps running on my ECMarkets ECN account — in exactly this kind of wide, whippy range, spread stability and limit-order fill quality determine the friction cost of mean-reversion strategies, and execution edge gets multiplied by volatility. That is also why my copy-trading stack has long been bound to ECMarkets as the execution venue — a wide, whippy range is the ultimate stress test of a broker’s infrastructure.
💡 Key Takeaways
- The data puzzle: Gold confirmed a weekly close below $4,000 (settling near $3,995.6, -4.05% w/w), even as central banks bought for a 23rd straight month and ETFs took in a record $31B in Q3 — buying and price have badly diverged.
- The mismatch: Central banks and ETFs are slow allocator flows; CTAs and margin-call liquidations are fast trader flows — allocators support the floor, not the price, and fast selling can temporarily overwhelm slow absorption.
- The physical signal: Shuibei was packed this weekend; retail jewelry is down ~100 yuan/gram in a month and nearly 500 below the year’s peak — price elasticity is releasing a third layer of structural demand.
- Three signals to track: daily RSI near 28 (momentum decay ≠ a buy signal); absorption tests at $3,970/$3,900; and next week’s FOMC + PCE as the referee.
- Base case: Pressure persists into early November; a $4,000-$4,400 wide range is the working Q4 framework — track signals, do not predict bottoms.
About the Author: Dongyi, professional quantitative trader focused on XAUUSD gold trading strategies, running DongyiTrade (dongyitrade.com), sharing quantitative trading strategies and in-depth market analysis.
Contact: WeChat: DongyiTrade | Telegram: @DongyiTrade
