Market Personality · Report
HG — Copper
HG is in an elevated-stress regime right now — vol is running hot and/or tail-shock days are piling up. Position sizing matters more here than strategy selection. We built this read from 37 years of HG daily data.
Archetype: Violent Range
Archetype verdict
Copper is a Violent Range market right now — the archetype label is how the full 37-year read of its personality rolls up into one tag. Big swings without direction. Mean-reverters only, tight stops, small size.
Three readings drive that call (all scored against HG’s own history, not against other markets):
- Volatility: Stressed — 92nd percentile — elevated
- Trend strength: Mean-Reverting — 23rd percentile — medium-term persistence runs low (shorter memory)
- Tail risk: Clean — 18th percentile on kurtosis — no fat-tail shocks piling up
- Risk first: Vol or tails are elevated — trim position size and raise stops. The realized-volatility distribution has fat right tails right now; sizing matters.
- System selection: Prefer strategies that explicitly size on realized vol (vol-targeting, risk-parity weighting). Static-size systems historically take outsized drawdowns here.
- If you’re sizing for risk: Vol is in the 92nd percentile of HG’s own history — a touch warm. Consider trimming positions 10-15% from your baseline until vol normalizes.
DNA radar
This radar chart is HG’s fingerprint today. Each of the six spokes measures one trait of the market — how strongly it trends, how volatile it is, how fat its tails are, and so on — ranked against HG’s own 37-year history. A big spike on one spoke means HG is unusually strong on that trait right now. The overall shape is the market’s personality at a glance.
5-year trajectory
Today’s radar is only one frame. This strip shows four frames — 5 years back, 2 years back, 1 year back, and today — so you can see HG’s personality evolve over time. All four use the same yardstick (HG’s full history), so any change in the shape is a real change in the market, not a change in how we’re measuring it. The label under each frame is the archetype we called at that time.
How to read it: a growing polygon means the market is heating up on those traits. A shrinking polygon means it’s cooling off. Big label changes between frames (e.g. Healthy Trend → Full Crisis) are the real regime shifts worth paying attention to.
Current-state scorecard
Here’s the full read — all 16 traits we track, measured over four lookbacks (1 year, 2 year, 5 year, lifetime). Each number is a percentile: how today’s reading ranks against this market’s own history over that window. BIG 6 tags the six traits that drive the radar chart above. Use the arrow column to see whether each trait is rising, flat, or falling vs its 2-year distribution.
← swipe horizontally on mobile to see all columns.
Trend column: ↑↑ strongly rising vs 2-year distribution, ↑ rising, → flat, ↓ falling, ↓↓ strongly falling. Thresholds are 0.5 and 2.0 standard deviations. Percentile ranks use the full lifetime distribution — a 70th-percentile reading means today is higher than 70% of all observed days. The 1-year column uses only ~252 days and should be read as a short-term signal, not a precise rank.
Macro lens
Macro lens
Separate from the per-market pair signals above, these are findings where HG‘s personality archetype carries information about an external macro regime — an inflation print, a yield-curve inversion, a drought. These are empirical “when this market’s personality looks like X, that external macro regime tends to follow” findings — context signals, not trade recommendations.
When copper settles into a Working Range archetype, a seasonal dry anomaly in the Atacama Desert mining region follows within six months 45% of the time — more than double the 22% baseline.
Chile produces roughly 25% of world copper, and Atacama mining is water-constrained — droughts directly cut smelter output. Copper’s Working-Range archetype leads confirmed Atacama deficits at 2.11× lift, 50% of all active-deficit months show HG in Working Range (2.94× its base rate). Supply-side early-warning for the copper complex.
When copper enters a Healthy Trend archetype, a seasonal dry anomaly in the North China Plain follows within six months 53% of the time.
Copper is a broad China-activity proxy through construction and industrial demand. NCP water stress (China’s biggest grain-producing region) often coincides with broader economic stress that shows up in the copper tape. Context signal for China-exposed commodity positioning.
When copper settles into a Working Range archetype, U.S. industrial production contracts YoY within six months 59% of the time.
Dr. Copper lives up to the nickname: working-range (sideways) copper is historically a reliable precursor to industrial-production contraction. This complements the fundamental HG × China credit + Chinese NCP signals — copper reads both U.S. and Chinese industrial cycles.
When copper enters a Hidden-Risk Range archetype, CFTC copper speculative long positioning hits a 5-year extreme within six months 56% of the time.
Hidden-Risk Range on HG — calm prices with latent downside — is historically the strongest precursor to crowded-long copper positioning. Full Crisis shows 2.88× conditional lift during active extremes. Position-cycle read for macro / CTA allocators.
Past co-movement does not guarantee future results. Macro-lens entries are context signals — not trade recommendations.
37-year history
Three ways to look at HG’s full 37 years of personality. First chart: every Big-6 trait over time, with the “normal range” shaded in cream and the big macro events (crashes, policy pivots, crises) marked. Second chart: the whole 37 years compressed into one map so you can see how the market has moved through regime-space. Third chart: a month-by-month color calendar of which archetype HG was in — scroll it like a weather map of the market.
Coverage: 1989-12-01 to 2026-11-05 · 9,275 daily observations.
Feature history
One panel per Big-6 trait, over all 37 years. Hover any point for the exact reading. Drag the bar at the bottom to zoom into a specific era (e.g. 2008, 2020). The cream band is the p10-p90 envelope — the middle 80% of all-time readings. Treat it as a “normal zone” with a caveat: markets with fat tails routinely visit the extremes, so a value outside the band isn’t anomalous, just historically unusual. Triangles along the top mark major market events (structural breaks — crashes, policy turns, war); hover to see each one.
Personality map
This map shows where HG has lived over 37 years. The shaded region is a density map — darker patches are where the market has spent most of its time. The labeled circles (1990s, 2000s, 2010s, 2020s) are decade medians, color-matched to time (navy = early, gold = recent). The gold diamond is today, with the exact reading. Left-to-right is how strongly the market was trending (right = more trend, left = chop). Up-down is how volatile it was vs HG’s own history (0 = quiet, 100 = extreme). Read it as: is today close to where the market usually lives, or at the edge of unfamiliar territory?
Regime calendar
Every month of HG’s life, labeled as one of 12 personality types (combinations of volatility, trend, and tail risk). Each color is one archetype — the legend below the chart ranks archetypes by how much of the history the market has spent in each. Hover any band to see the exact month and archetype. Triangles are the same macro-event annotations from the first chart.
Methodology
The 16 traits. We group them into four families (click any underlined term for a quick definition):
- Trend traits (does the market keep going in one direction?): Hurst exponent (R/S estimator), variance ratios, ADF p-value, OU half-life, lag-1 autocorrelation.
- Volatility traits (how big are the daily swings, and are they sticky?): annualized vol (std dev of daily log returns, rolling 252-day window, scaled by √252), vol-of-vol, GARCH(1,1) persistence.
- Shape traits (are the tails fat? which side are shocks on?): skewness, excess kurtosis (Fisher), leverage correlation.
- Complexity traits (is the price path noisy or clean? do cycles repeat?): sample entropy, spectral entropy, fractal dimension, rolling max drawdown.
How we pick the archetype. Three pieces — volatility level, trend strength, tail risk — combine into a 12-box grid of personality types. We use HG’s own history (not a cross-market ranking), so a quiet market and a wild market can both be scored fairly against themselves. The classifier reads exactly three of the sixteen traits (Hurst percentile, annualized-vol percentile, kurtosis percentile) through hand-set thresholds. The other 13 traits inform the scorecard but not the archetype label.
How to read a percentile. “70th percentile vol” means today’s vol is higher than 70% of HG’s own history — NOT higher than 70% of other markets. Every number on this page is scored against HG itself. Cross-market comparisons live on the personality clusters page.
Caveats worth knowing.
- Percentile series look ahead. Readings on the feature-history chart are ranked against the market’s full lifetime, including data after the x-axis date. A 2008 reading at the 90th percentile is high relative to what we now know of history, not relative to what was known in 2008. This is a snapshot characterization, not a point-in-time rank series.
- The archetype classifier reads only 3 of 16 traits. Features like skewness, entropy and the complexity family feed the scorecard but not the label.
- Hurst is a structural measure, not a daily-trend signal. Hurst > 0.5 indicates long-memory in the 252-day rolling return series — useful for characterizing regime. It does NOT reliably predict positive next-day return autocorrelation. Our in-house empirical test on 37 years × 18 markets found no such link. Don’t read a “trending” Hurst label as a day-trading signal.
- Rule-based classifier. Thresholds are hand-set. The classifier does not learn from alert performance (that lives in TA Profile).
- Short-window instability. 1-year percentiles use ~252 days; on recently-listed markets (e.g. BTC) they are noisier than lifetime stats. Read the Lifetime column when confidence matters.
- Rolling windows lag sharp breaks. Expect feature readings to lag sharp regime breaks by a few bars.
- Event markers are hand-curated. The triangles on the feature history and regime calendar are macro-event annotations, not the output of a change-point detection algorithm. Alignment between an event and a visible shift on the chart is for reference, not a statistical claim.
Full plain-English definitions for every feature in the Feature glossary.
Related
- Market Personality hub — back to the market index.
- Personality clusters — both the structural map and today’s state map across all live markets.
- Feature glossary — plain-English definitions for all 16 features.
- Live risk-state detector (Phase 4 continuous stress score) — coming soon.