Gold is supposed to go up when the world falls apart. That's the pitch, the premise, the entire reason a trillion dollars sits in the metal.
So explain this: there's an active military conflict in the Persian Gulf, the Strait of Hormuz is disrupted, oil is above $100, and GLD has just posted its worst week since 1983. Down 18.5% from its January peak of $496 to $404. Seven consecutive losing sessions. The worst monthly performance since October 2008.
If you own gold right now, you're being told two contradictory things. The macro story says this should be your moment. The price chart says something is deeply wrong.
We wanted to know which one is right. So we did something most analysts won't — we tested the actual data.
What everyone is saying
The conventional narrative has crystallised quickly. Gold's selloff is being blamed on three converging forces: a stronger US dollar (the Dollar Index hit 100.50 in early March), hawkish Fed policy (rates held steady, rate cuts pushed to September at the earliest), and forced liquidation of leveraged positions as margin calls cascade through the futures market.
The paradox — a hot war driving gold down — resolves when you understand the mechanism. Oil above $100 feeds inflation expectations, which forces central banks to stay restrictive, which pushes real yields higher, which makes non-yielding assets like gold relatively more expensive to hold. The dollar strengthens as global capital seeks safety in US Treasuries rather than bullion. Leveraged gold longs get squeezed. The selling feeds on itself.
JP Morgan is maintaining their $6,300 year-end target. Deutsche Bank stands at $6,000. Neither has flinched. Both view this as a tactical correction inside a structural bull market — the same structural forces (central bank accumulation, de-dollarisation, US fiscal deficits) that drove a 65% rally through 2025 haven't changed.
There's also an interesting divergence between paper and physical markets. Futures collapsed, but physical gold premiums have stayed elevated. Demand from institutional buyers, jewellers, and individual stackers hasn't budged. The physical market — where actual metal changes hands — is telling a different story than the screen.
All of which is useful context. None of which tells you whether to buy the dip.
What the data actually shows
We track 118 assets across every major asset class, with 97 indicators calculated daily and five years of history. For each indicator, we compute a rolling percentile — where today's reading sits relative to the past year. This creates four dimensions: Extension (how stretched price is from its moving averages), Momentum (rate-of-change energy), Flow (buying versus selling pressure), and Volatility (how calm or chaotic the environment is).
Here's where GLD sits right now:
| Dimension | Percentile | What it means |
|---|---|---|
| Extension | 0th | Price is more compressed relative to its moving averages than at any point in the past year |
| Momentum | 0th | Rate of change indicators are at their absolute floor |
| Flow | 1.6th | Selling pressure is near the most extreme we've measured |
| Volatility | 89.8th | The trading environment is extremely chaotic |
Extension and momentum both at zero. That's not "low." That's the floor. GLD has literally never been this compressed across these dimensions in the past twelve months. In a dataset spanning 1,232 trading days with full indicator coverage, this is one of the most extreme readings we've recorded.
The traditional composite scores — the kind you'd see on most analysis platforms — would tell you GLD scores 6 out of 100 on Trend and 17 out of 100 on Flow. "Weak." That's the entire insight. Weak.
We wanted to know: does "weak" mean anything? And more importantly, does "weak" always look the same?
The test
We pulled every day since November 2021 where GLD's extension and momentum percentiles were both below the 20th — 94 observations. Then we measured what happened 21 and 63 trading days later.
The short-term numbers are striking. When both extension and momentum drop below the 20th percentile, GLD was positive 83.3% of the time at 21 days, with a median return of 2.5%. At extreme levels (both below the 5th percentile), the 21-day hit rate jumps to 94.1% across 20 observations. Gold bounces from these levels. Almost always.
But here's where it gets interesting. The 63-day outcomes tell a completely different story depending on which episode you look at — and the traditional scores can't tell the episodes apart.
Three episodes, three outcomes, same "weak" score
We identified every time GLD hit a multi-dimension extreme (extension and momentum both below the 5th percentile) since 2021. There are four distinct episodes. The first three have resolved. The fourth is happening right now.
June 2021: GLD at $165. Extension at the 0th percentile, momentum at the 0.5th. The composite scores said "weak" — trend score 11, flow score 33. What happened? A short bounce of about 2.3% over 21 days, then a decline of 1% over 63 days. Dead cat.
July 2022: GLD at $161. Extension at the 1.3rd percentile, momentum at the 2nd. Composites again: trend score 14, flow score 19. Essentially the same reading as 2021. What happened? Bounce of 3.6% at 21 days, then -2.2% at 63 days. Another dead cat.
October 2023: GLD at $169. Extension at the 3.4th percentile, momentum at the 2.8th. Composites: trend score 8, flow score 21. If anything, the composites looked worse than the prior episodes. What happened? A 9% gain at 21 days and 12% at 63 days. This was the start of gold's massive bull run that would take it to $496.
Three episodes. The composite scores — 11/33, 14/19, 8/21 — are essentially indistinguishable. They all say "weak." But the outcomes ranged from -2.2% to +12%.
The dimension everyone is missing
When we dug into what actually separated the dead-cat bounces from the genuine reversal, two factors emerged that the composite scores completely ignore.
Volatility regime. The dead-cat episodes (June 2021 and July 2022) both had volatility percentiles above 72 — an elevated, chaotic environment where forced selling and margin calls dominate price action. The genuine reversal (October 2023) had volatility at the 24th percentile — a quiet, compressed environment where selling had exhausted itself. This is the difference between a leveraged liquidation event (volatile, mechanical, temporary) and genuine capitulation (quiet, exhausted, potentially structural).
Flow depth. October 2023 saw flow collapse to the 3.6th percentile — a near-total washout. The dead-cat episodes had flow at a less extreme 14th to 21st percentile — selling pressure was elevated but not exhausted. When everything is weak but flow hasn't fully capitulated, the bounce tends to be shallow. When flow is completely washed out, the recovery can be real.
These are dimensions the composites don't capture. The composite flow score saw all four episodes as roughly the same (19 to 33 out of 100). The percentile framework sees that October 2023's flow at the 3.6th percentile is a fundamentally different condition than July 2022's flow at the 15th.
Where gold stands right now
Today's GLD sits at $404 with this profile:
- Extension: 0th percentile (matches the dead-cat episodes)
- Momentum: 0th percentile (matches the dead-cat episodes)
- Flow: 1.6th percentile (matches — even exceeds — the October 2023 reversal)
- Volatility: 89.8th percentile (matches the dead-cat episodes)
- CMF: +0.01 (essentially neutral, unlike the deeply negative CMF in October 2023)
- Risk profile score: 81.6 out of 100 (the highest of all four episodes by a wide margin)
The data is genuinely split. The flow depth — at 1.6, the lowest of any episode — is the signal that preceded gold's best recovery. But the volatility character — at nearly the 90th percentile — matches the episodes that produced dead-cat bounces. The macro context (forced liquidation driven by dollar strength and rising yields) aligns with the volatile-selloff pattern rather than the quiet-capitulation pattern.
There is one anomaly worth noting. The risk profile score at 81.6 is unusually high — far above the 50 to 56 range of the prior episodes. This measures risk-adjusted return quality (Sortino ratio, drawdown characteristics, upside/downside capture). It suggests that gold's underlying risk-return profile remains strong even as everything else has collapsed. That wasn't the case in the prior dead-cat episodes.
The structural question
Beyond the technicals, there's a larger question worth sitting with. Gold rallied 65% through 2025 on a clear thesis: central banks diversifying away from dollars, US fiscal deficits growing unsustainably, geopolitical fragmentation increasing demand for non-sovereign stores of value.
Has any of that changed? The Iran conflict has actually accelerated most of these trends. The dollar's current strength is a tactical response to energy shock, not a structural reversal of the de-dollarisation trend. Central bank gold buying hasn't slowed. US debt dynamics haven't improved.
What's changed is the short-term plumbing of the market. Leveraged positions are being unwound. Margin calls are forcing mechanical selling. The paper market has temporarily overwhelmed the physical market.
This is what makes the volatility percentile so informative. When gold's selloff is driven by leveraged liquidation (high vol), the recovery tends to be a bounce within a continuing correction. When it's driven by genuine loss of interest (low vol), the recovery tends to mark the beginning of a new leg up. The market needs to finish the liquidation before the structural thesis can reassert itself.
What to watch
If you're trying to figure out whether this is June 2021 (dead cat, more pain ahead) or October 2023 (genuine reversal, massive opportunity), there are two things to monitor.
Volatility. The single most important signal would be gold's volatility percentile falling from the current 90th toward the 30th to 40th range over the next two to three weeks. That would suggest the forced selling has run its course and the environment is shifting from mechanical liquidation to genuine price discovery. If volatility stays elevated, the historical pattern says the bounce will be short-lived.
Flow recovery. The current flow reading of 1.6 is deeply washed out. If flow begins to recover (climbing back above the 20th percentile) while volatility is declining, that would be the combination that preceded October 2023's genuine reversal. If flow stays pinned near zero while volatility remains high, the selling isn't done.
The 21-day bounce is the high-probability outcome — gold has recovered from these percentile extremes in the short term 83% to 94% of the time. The question isn't whether gold will bounce. It's whether the bounce marks the start of something real, or the pause before the next leg down.
Right now, the data says watch and wait. The answer is in the volatility.
Data as of March 23, 2026. Analysis based on GLD (SPDR Gold Shares) daily indicator data from November 2021 to present. Percentile calculations use a rolling 365-day window. This is not investment advice — it is a description of how the current conditions compare to historical episodes, based on our indicator framework. Past patterns do not guarantee future outcomes.