Gold Is Down 23% From Its Record. Buying Opportunity or a Deeper Decline?

Research note: gold (GLD) and silver (SLV) daily closes, real yields and CFTC positioning through October 2, 2026. Frozen October 3, 2026. How we checked it · How we use AI.

• The summer recovery failed. Gold has given back 76% of a 17% rally from its July low.
• The recoveries had made progress by now. This one has not. 82 sessions after the first 20% day, gold is 4.1% below that day’s close. Every past recovery was above it at the same point.
• Speculators have started selling. Managed-money funds have cut their long positions by 17.6% since the August rally high.
• The case against another 2013. GLD’s share count has risen 3.0% since the first 20% day, against a 24.0% fall at the same point in 2013, and miners have outperformed gold.

Gold doubled and then some between 2023 and January 2026. Since the January 29, 2026 record, it has given back 23.3%, and silver, which ran even harder, is 48.2% below its own peak. The question for anyone holding either is whether this is a correction inside a long bull market or the start of a longer decline.

My read is that the next several months lean bearish while the longer record leans bullish. Gold’s history since 2004 says 20% declines are usually bought, and three of four were. But buyers had their chance this summer and could not hold it, the recoveries had all made progress by this point and this one has not, and the speculative buying that powered the August rally is now being sold. The confirmation I would want before expecting a 2013-style bear market, heavy selling by investors, has not appeared.

Here is the setup. Gold first closed 20% below its record on June 5, 2026 and fell as far as 26.4% below it, to a GLD close of $364.96 on July 16, 2026. From there it rallied 17.3% to August 25, 2026, closed above its 200-day average for 7 sessions between August 19 and August 27, and has since given back 76% of the rebound. It now sits 8.7% below the average. The lower pane is the share of gold futures open interest held net long by managed-money funds, the group of traders whose positioning swings the most.

Four times since 2004

GLD began trading in November 2004 and tracks the spot price of gold less a small fee. On its record, gold has closed 20% below a running high five times, counting this one.

Three of the four were over quickly. Gold was 15.5%, 18.0% and 15.8% higher a year after June 2006, August 2008 and September 2022. The exception was April 2013: a year later gold was 11.8% lower, and the decline ran to 45.6% from the 2011 peak before it ended in December 2015.

82 sessions in: where each decline stood

The three-out-of-four record counts from the first 20% day, but readers are deciding 82 trading sessions later. Measured at that same point, the episodes look less alike. Each column below is a separate piece of evidence: the price, the miners that mine the metal, and the investors who buy and redeem GLD shares.

Every eventual recovery was already above its first-20%-day close by this point. Only 2013 and today were below it. That is a useful distinction, and only one past failure supports it, so I read it as evidence and not as a probability.

The bearish case

The summer rally failed. It was the recovery case’s best test, and the chart at the top shows how it ended: a week and a half back above the 200-day average, then most of the gain gone. A market that cannot hold a rally of that size has more sellers than buyers at those prices, and that is the clearest reason I remain cautious.

Speculators are selling what they bought in August. Managed-money gross longs are 17.6% below their August 25, 2026 level. Short positions fell too, so this is long liquidation and not new short selling. It shows that selling has begun. It does not show how much is left.

That selling is still small next to what earlier episodes saw. Measured as a share of open interest, net exposure fell by at least half within six months of the first 20% day in 2006, 2008 and 2013; in 2022 funds were already net short at the first 20% day and went further short. Today net exposure has fallen from 34.4% to 29.6% of open interest, although the number of net long contracts has risen 7.3% as total open interest grew 24.7%.

The run was stretched. Gold rose 177% in the three years to January 2026, against 126% and 128% before the 2008 and 2011 peaks. Silver, up 387% to its own January peak, is now 48% below it.

Real yields are near their highest since 2008. The 10-year real yield, the Treasury yield less expected inflation, was 2.88% on October 1, 2026, a day after reaching 2.93%, the highest since November 2008. That is what an investor gives up each year to hold a metal that pays nothing.

I would not lean on real yields alone. In 2022 they had risen 2.10 points from the 2020 gold peak to the first 20% day, and gold was back at its record 1.5 years later. Rates matter for gold, but they did not decide any of these episodes by themselves.

The bullish case

The base rate. Three of four earlier 20% declines were recovered within 1.5 years of the first 20% day. This decline was also fast, 127 days from record to 20% down, like June 2006 and August 2008, both of which recovered within about a year.

GLD investors have added shares since June. This is the strongest objection to another 2013. In 2013 the decline came with heavy redemptions: GLD’s shares outstanding fell 24.0% in the 82 sessions after the first 20% day. This time they rose 3.0%. Shares move only when investors create or redeem them, so a falling price alone cannot shrink them. They are not a clean signal either: in 2022 they fell 5.4% over the same stretch, and gold recovered anyway.

Miners are holding up. Gold miners have gained 16.1% relative to gold since the first 20% day. In 2013 they lost 8.4% relative to gold over the same stretch, and in 2008 they lost 21.7% before a recovery, so miners are a separate confirmation measure and not a verdict. Today they point away from 2013.

A bull would add that the dollar, which often moves against gold, is almost unchanged over the past year (−0.2%).

Where I come out

The evidence supports further downside more clearly than a multiyear bear market. The failed rally, the lack of progress 82 sessions in, and the selling of August’s speculative positions point to a lower low, below GLD’s $364.96 close on July 16, 2026, before a durable recovery. Further liquidation could pressure prices, but the historical episodes do not establish that gold needs a positioning washout to recover.

What would make me expect a long bear market is the evidence 2013 had: investors redeeming ETF shares on a large scale and miners falling further than the metal. I treat both as confirmations to look for, not as rules drawn from one episode, and neither is present now.

What would change my read

Toward a recovery: a sustained recovery in price, even with speculators still heavily positioned. A close back above the 200-day average, now 9.5% above the price, that holds longer than August’s 7 sessions would do it. Net exposure falling below 17% of open interest while GLD holds above $364.96 would point the same way.

Toward 2013: a GLD close below $364.96, together with sustained redemptions shrinking GLD’s share count and miners falling behind gold.

How we checked it

• Gold and silver. GLD and SLV daily closes since launch (November 2004 and April 2006). Both track spot prices less a small annual fee, so their percent moves are the metals’ to within that fee. An episode starts at the first close 20% below the running record (30% for silver, which moves about twice as hard), and a new one needs a new record first.
• Real yield. The 10-year Treasury yield less the 10-year breakeven inflation rate, daily from January 2003 (FRED). It is the market’s 10-year real rate, close to the TIPS yield.
• Positioning. CFTC disaggregated Commitments of Traders for COMEX gold, managed-money net long as a share of total open interest, weekly from June 2006. “Within six months” is the lowest weekly reading in the 182 days after each first 20% day.
• Contract counts. Managed-money net long: 112,179 contracts at the first 20% day and 120,318 on September 29, 2026. Gross longs: 159,819 on August 25, 2026 and 131,711 on September 29, 2026; shorts 15,072 and 11,393.
• Miners and ETF flows. Miners are the GDX/GLD price ratio (GDX daily since May 2006). ETF flows are GLD’s shares outstanding, weekly from January 2012 (Massive reference data by date); they change only through creations and redemptions. Each comparison uses the latest weekly reading on or before the date.
• The summer rally. From the lowest close after the first 20% day to the highest close after that low; “given back” is the share of that gain lost by October 2, 2026.
• Small sample. Four completed episodes cannot establish a rule; they show what separated the endings that happened. The 1980 peak and the 1980s bear market are before these series begin.
• Frozen evidence. Every number on this page comes from the snapshot of October 3, 2026.