📅 2026-10-05 · COMEX精銅 · 單品種深度報告舊覆蓋策略

數據截止: 2026-10-02 · 結算口徑: 交易所最終日結算價

Copper (HG=F) Deep Dive | 2026-10-05 — Long into the 6.5 pivot shelf, 6.46 invalidation | 2026-10-05

Call: LONG copper (HG=F, execution month HGZ26) from the 6.5–6.55 pivot shelf, with a 6.46 invalidation on a settled basis. Three reasons: (1) the 5D -3.21% flush has carried price to the lower third of the 20-day 6.35–6.93 channel (pos 35%) while the last completed weekly bar (2026-09-28–2026-10-02) closed at 6.55, only 0.6% above the 6.51 weekly low, so the correction is mature rather than accelerating; (2) the physical tape is not confirming the selloff — COMEX registered stocks were unchanged at 471,400 short tons on 2026-10-01, SHFE warrants fell 721 MT (-6.7%) to 10,011 MT on 2026-09-30, and the only build is a modest +575 MT (+0.2%) in LME warehouse stock to 248,650 MT; (3) managed-money net length at 78,058 lots (netPct 25.92%, crowding 68.51) has already been trimmed by 4,464 lots w/w, so the crowded-long risk is being worked off into weakness rather than compounding. The copper/gold ratio at 1.57 sits in the 94th percentile of its 1-year range — pro-growth pricing is intact. Invalidation: a settle below 6.46 (S2).

1. Bottom Line & Directional Bias

Call: LONG copper (HG=F, execution month HGZ26). Invalidation: a daily settle below 6.46 (S2).

Three reasons. First, positioning has already de-risked into the move: managed-money net length fell 4,464 lots w/w to 78,058 as of 2026-09-29, with netPct at 25.92% and crowding at the 69th percentile of the 3-year window — elevated, but no longer at the 69.4 extreme printed on 2026-09-08. Weakness is being absorbed, not chased. Second, the physical balance is quiet-to-tight: COMEX registered stocks were flat d/d at 471,400 short tons on 2026-10-01, SHFE warrants dropped 721 MT (-6.7%) to 10,011 MT on 2026-09-30, and LME warehouse stock rose only 575 MT (+0.2%) to 248,650 MT on 2026-10-02, a 20-report change of +14,475 MT that is a rebuild from a low base rather than a surplus signal. Third, the technical setup is a pullback inside an uptrend: settle 6.55 sits at 35% of the 20-day 6.35–6.93 channel, with the 52-week high at 6.93 and the 52-week low at 4.83.

The risk to the call is macro, not micro: the 10-year yield at 5.277 and DXY at 101.86 are a persistent headwind, and a settle below 6.46 would signal that the 20-day channel floor at 6.35 is the real target.

2. Price Action & Technical Analysis

HG=F settled at 6.55 on 2026-10-02, +0.17% on the day (settle). The 5-day change is -3.21% and the 20-day change is -1.73% (settle) — the decline is front-loaded into the past week, not a slow grind. The 20-day channel runs 6.35–6.93, putting the settle at the 35rd percentile of that range. The 52-week range is 4.83–6.93, so price is roughly 76% of the way up the annual range but has surrendered the top of the recent channel.

Volatility is moderate and slightly elevated versus the recent regime: ATR14 is 0.13, or 1.99% of price (full daily range, not ±). RV20 is 26.5% annualized. Against that, the equity-vol complex is calm — VIX 15.31 (15th percentile, 1Y), GVZ 23.23 (15th percentile, 1Y), OVX 51 (49th percentile, 1Y) — so copper's own realized vol is running hot relative to cross-asset risk pricing. That asymmetry favors selling optionality only if the 6.46–6.5 shelf holds; it does not.

Pivots from the settle-based snapshot: P 6.57, R1 6.61, S1 6.5, R2 6.68, S2 6.46. The settle at 6.55 is just below the pivot, which is the classic location for a mean-reversion long against S1. In early Asian trade on 2026-10-05 (07:00), the last print was 6.58, +0.54% versus settle, with a session range of 6.57–6.6 (Asia) — price is already back above the pivot and probing R1. That is a constructive tell, but it is an unfinished bar and carries no settlement weight.

The last completed weekly bar (2026-09-28–2026-10-02) opened 6.77, high 6.77, low 6.51, closed 6.55, -3.21% w/w. The current week has no settled bar yet, so no weekly breakout or weekly reversal language applies. What the completed week does tell us is that the 6.51 low was bought and the close was 0.6% above it — a rejection wick rather than a breakdown close.

View: constructive above 6.5; the 6.57 pivot is the first gate, R1 6.61 the second. A settle below 6.46 voids the setup and opens 6.35.

3. Supply-Demand Balance & Fundamental Drivers

Inventories are the cleanest signal in this tape, and they are not bearish.

COMEX registered stocks: 471,400 short tons on 2026-10-01, d/d 0 (+0.0%). Flat is the message — the exchange deliverable pool is not being rebuilt into the price decline. LME warehouse stock: 248,650 MT on 2026-10-02, d/d +575 MT (+0.2%), 20-report change +14,475 MT. The 20-report rebuild is real but small in absolute terms and has been decelerating: 249,400 MT on 2026-09-30, 248,075 MT on 2026-10-01, 248,650 MT on 2026-10-02 — a two-day net draw of 750 MT followed by a 575 MT add. This is churn, not accumulation. SHFE warrant: 10,011 MT on 2026-09-30, d/d -721 MT (-6.7%). A 6.7% single-day draw in Shanghai warrants is a meaningful tightening signal for the Asian physical market and is the single most bullish line in the inventory block.

The term structure is in contango: M1-M2 -0.0315 (-0.48%), roll yield -5.77%, slope 0.0337. Contango here reflects carry — the 10-year at 5.277 makes storage finance expensive — and is a roll cost for longs, not a price cap. It does, however, argue against a violent prompt squeeze: there is no backwardation demanding immediate physical delivery.

Macro transmission is through the dollar and rates. DXY 101.86 (-0.07%, 2026-10-04) and ^TNX 5.277 (+0.76%, 2026-10-02). A 5.28% 10-year is a genuine headwind for a carry-heavy industrial metal, and the dollar at 101.9 is mid-range rather than weak. The offset is the copper/gold ratio at 1.57 (ratio×1000), 94th percentile on 1Y and 52nd on 3Y — the market is still pricing copper as a growth asset relative to gold, and a ratio in the top 6% of its one-year distribution is not what a hard-landing tape looks like.

The one headline in the 48-hour window is a mining-sector ranking piece (2026-10-04), which is thematic rather than flow-relevant.

View: physical tightness in Shanghai plus flat COMEX and churning LME stocks argue the selloff is positioning-driven, not fundamental. Bullish, contingent on 6.46 holding.

4. Positioning & Fund Flows

CFTC managed-money positioning (as-of dates per row):

2026-09-29: OI 301,201, long 92,094, short 14,036, net 78,058, Δ -4,464
2026-09-22: OI 301,657, net 82,522, Δ +17,416
2026-09-15: OI 289,463, net 65,106, Δ -17,048
2026-09-08: OI 297,491, net 82,154, Δ +9,272

The four-week pattern is violent two-way churn around a 65k–83k net-length band, not a one-way liquidation. The most recent week trimmed 4,464 lots while price fell 3.21% over the 5D window — the reduction is smaller than the price damage, which means the selloff was driven more by long liquidation of weaker hands and short-term momentum than by a wholesale institutional exit.

Crowding: netPct 25.92%, crowding 68.51 (3-year percentile), down from 27.36%/69.05 on 2026-09-22 and 27.62%/69.39 on 2026-09-08. This is elevated but not extreme — the |z|>1.5 extreme flag is not triggered, and the percentile is drifting lower. Calling this trade “crowded” would overstate it; the correct read is a well-populated long that has begun to lighten. The CTA trend proxy is steady at 62 across all four weeks, and hedge pressure sits at 54.82%, down from 58.27% on 2026-09-15 — commercial hedging appetite is fading, which typically accompanies a market that has already discounted bad news.

Implied versus realized: copper has no IV line in this feed, but the cross-asset read is that VIX (15.31, 15th pct), GVZ (23.23, 15th pct) and OVX (51, 49th pct) are all calm while copper RV20 is 26.5%. Optionality across macro assets is cheap relative to copper's own realized movement.

View: positioning is a headwind that is actively deflating, not a reason to sell. Neutral-to-supportive for the long.

5. Cross-Asset Relative Value

Copper/gold ratio: 1.57 (ratio×1000), 1Y percentile 93.65%, 3Y percentile 51.98%. The one-year percentile is the operative number: copper has been strongly favored over gold across the past twelve months, and the ratio remains in the top 6% of that window. The three-year percentile at 51.98% shows this is a one-year phenomenon, not a structural regime — the ratio is mid-range against a longer history. The practical read: the pro-growth leg of the metals complex is still being paid, and copper is not the funding short in a metals pair.

Against rates and the dollar, copper's problem is carry. ^TNX 5.277 (+0.76%) and DXY 101.86 (-0.07%) define a regime where holding a contangoed industrial metal costs money (roll yield -5.77%) and the dollar offers no tailwind. That combination caps upside velocity but does not reverse direction as long as the growth ratio holds.

Risk metrics (retrospective): DD52w 13.49%, DD20d 7.03%, Sharpe30 0.0038, VaR95 -2.97%. A 30-day Sharpe of essentially zero confirms a market that has gone nowhere net over the past month — consistent with the 20D change of -1.73% and the churn in the COT data. The 20-day drawdown of 7.03% is the current pain trade; the 52-week drawdown of 13.49% is well short of a bear-market signature.

View: copper is the preferred expression of reflation versus gold, but the rate/dollar backdrop argues for buying weakness rather than chasing strength.

6. Historical & Seasonal Patterns

Seasonality for the same calendar start, next 20 sessions, last 15 years: mean +1.92%, median +1.89%, up 10 of 15 years. Best case 2011 +15.53%, worst case 2012 -7.4%.

The distribution is positively skewed at the tails but the central tendency is the point: a 10-of-15 hit rate with a median of +1.89% is a modest, repeatable early-October tailwind. The sample is small (15 observations) and the 2011 outlier flatters the mean, so the median is the more honest anchor. Applying a +1.89% median move to the 6.55 settle implies a 20-session reference near 6.67 — just above R2 at 6.68.

This is context, not a thesis. It aligns with the technical read (price at 35% of the 20-day channel, above the 6.5 shelf) and with the positioning read (net length already trimmed), which is why it earns a place in the report rather than being dismissed.

View: seasonality is a mild tailwind that supports holding a long through mid-October; it does not justify adding size on its own.

7. Scenario Analysis (Base / Bull / Bear)

Base case — 50%: grind higher from the pivot shelf. Trigger: a daily settle back above P 6.57, ideally with the Asia strength (6.58, +0.54% vs settle) confirmed at the COMEX close. Path: 6.57 → R1 6.61 → R2 6.68, with the 20-day channel top at 6.93 as the outer bound over a multi-week horizon. Action: hold the long, add on a confirmed settle above 6.61, trail the stop to breakeven once 6.68 trades. This is the path consistent with the section 1 call.

Bull case — 30%: short-covering squeeze through R2. Trigger: a settle above 6.68 combined with a further draw in SHFE warrants or a surprise in the ISM Services PMI print (forecast 54, prior 55.4, surprise threshold ±1.4) that weakens the dollar. Path: 6.68 → 6.8 → retest of the 6.93 52-week high. Action: add on the 6.68 settle, target 6.9, move the stop to 6.6. The fuel for this is the 78,058-lot net long plus a CTA proxy pinned at 62 — a break of R2 would force trend followers to re-engage.

Bear case — 20%: settle below 6.46 invalidates. Trigger: a daily settle under S2 6.46, most plausibly on a hot ISM Services print or a hawkish FOMC minutes read (BJT 10-08 02:00) that pushes the 10-year further above 5.277 and lifts DXY off 101.86. Path: 6.46 → 6.35 (20-day channel floor) → 6.2. Action: exit the long on the settle, stand aside; do not attempt to catch the 6.35 print without a reversal bar. A break of 6.35 would put the 52-week low at 4.83 in play over a longer horizon, but that is not this week's trade.

Probabilities sum to 100%. The base case is the section 1 call.

8. Trading Strategies & Risk Management

Strategy 1 — Core long from the pivot shelf (conviction 7/10). Entry 6.55 (at the settle, or better on any dip toward S1 6.5). Stop 6.44 — beyond S2 6.46 and roughly one ATR14 (0.13) below entry. Target 6.68 (R2), with a secondary objective at 6.9. Horizon 1–5 days for the first target, up to 20 sessions for the extension. Size: half of normal risk budget at entry, adding the second half only on a daily settle above 6.61 (R1). Rationale: price is at 35% of the 20-day channel, positioning has already been trimmed 4,464 lots, and Shanghai warrants drew 6.7%.

Strategy 2 — Momentum add on the R2 break (conviction 6/10). Entry on a daily settle above 6.68. Stop 6.55 (back at the original entry, protecting the position). Target 6.9, with 6.93 (52-week high) as the stretch. Horizon 5–10 sessions. Size: one-third of normal risk budget, layered on top of Strategy 1 only. This is the bull-case expression and should be abandoned if the settle fails to hold above 6.68 within two sessions.

Risk management: total exposure across both strategies should not exceed 1.5x normal risk budget given the 5.277 10-year yield and the contango roll cost of -5.77%. The 6.46 settle is the single hard invalidation for the entire thesis — no averaging down below it.

9. This Week's Data Calendar

BJT 10-05 22:00 | ET 10-05 10:00 — ISM Services PMI (SEP), forecast 54, prior 55.4, surprise if outside F±1.4 (USD/HIGH). The week's key macro input for copper via DXY.
BJT 10-08 02:00 | ET 10-07 14:00 — FOMC Meeting Minutes (USD/HIGH). Direct read-through to the 10-year at 5.277 and the dollar at 101.86.
BJT 10-07 22:30 | ET 10-07 10:30 — EIA Crude and Gasoline Stocks (USD/MEDIUM); API Crude BJT 10-07 04:30 | ET 10-06 16:30. Energy-complex context only.

This report is generated automatically from public quantitative and macro data for research and market tracking only. It does not constitute investment advice or a recommendation to trade.

單品種深度報告深度分析

COMEX黃金GC=F 2026-10-05

Gold (GC=F) Deep Dive | 2026-10-05 — Selloff Exhausted at 20D Floor, Tactical Long | 2026-10-05

Copper (HG=F) Deep Dive | 2026-10-05 — Long into the 6.5 pivot shelf, 6.46 invalidation | 2026-10-05

今日風險地圖

單品種深度報告 深度分析

往期研報封存

⟠ 本月策略回測表現

單品種深度報告深度分析