📅 2026-10-02 · COMEX精銅 · 單品種深度報告

數據截止: 2026-10-01 · 結算口徑: 交易所最終日結算價

Copper (HG=F) Deep Dive | 2026-10-02 — Constructive Base, 6.48 Pivot Is the Line | 2026-10-02

Call: LONG copper (HG=F / HGZ26.CMX) on a 1–3 week horizon, with the trade invalidated on a settle below 6.43 (S2). Three reasons underpin the call. First, the pullback is a positioning flush, not a fundamental break: the 5D change is -3.71% (settle) while the 20D change is only -0.83% (settle), and price at 6.54 sits at the 33.1% position of the 20-day 6.35–6.93 channel — mid-range, not broken. Second, visible inventories are not building: LME warehouse stock fell 1,075 MT d/d to 249,400 MT and SHFE warrants fell 721 MT d/d to 10,011 MT, with COMEX registered up only 1,100 short tons to 471,400. Third, the copper/gold ratio at 1.58 sits in the 96th percentile of its one-year range, a pro-growth signal that argues against a sustained industrial-metals breakdown. The offsetting risk is crowding: managed-money net is 27.36% of open interest, a 69.05 three-year percentile, with the CTA proxy pinned at 98. That is why the stop is structural, not cosmetic.

1. Bottom Line & Directional Bias

Call: LONG copper (HG=F, reference contract HGZ26.CMX). Invalidation: a daily settle below 6.43 (S2).

Three reasons. First, the selloff is shallow in context. The 5D change is -3.71% (settle) but the 20D change is only -0.83% (settle), and at 6.54 the market sits at the 33.1% position of the 20-day 6.35–6.93 channel — a mid-range retracement, not a trend break. Second, the physical side is not confirming weakness: LME warehouse stock 249,400 MT (-1,075 MT d/d), SHFE warrants 10,011 MT (-721 MT d/d), COMEX registered 471,400 short tons (+1,100 d/d). Third, the copper/gold ratio at 1.58 sits in the 96th one-year percentile — a pro-growth configuration that historically does not coexist with a durable copper downtrend.

The main offset is positioning: managed-money net is 27.36% of open interest, a 69.05 three-year crowding percentile, and the CTA proxy reads 98. Crowded longs amplify downside on a macro shock, which is precisely why the invalidation is a hard settle level rather than a mental stop. Above 6.43, the base case is a grind back toward 6.7–6.93.

2. Price Action & Technical Analysis

Settle 6.54 [2026-10-01], -1.26% on the day. The 5D change is -3.71% and the 20D change -0.83%, both settle-based. The 20-day channel is 6.35–6.93, putting the settle at the 33.1% position — lower third of the range but comfortably above the floor. The 52-week range is 4.78–6.93, so the market is trading near the top of its annual envelope; the current move is a pullback within an uptrend, not a reversal of it.

ATR14 is 0.137, or 2.09% of price, expressed as a full daily range. RV20 is 26.8% annualized. Against that, the daily pivot set is P 6.56, R1 6.62, S1 6.48, R2 6.7, S2 6.43. The settle at 6.54 sits just below P, which is consistent with a market that lost momentum into the prior close but has not surrendered the S1 shelf.

In early Asian trade on 2026-10-02 (07:00), the last print is 6.57, +0.52% versus the settle, with an Asian session range of 6.58/6.57. That is a constructive open: price has reclaimed the pivot P 6.56 on the Asia bar. Note the distinction — the +0.52% is an Asia-session move, not a settle, and it does not by itself confirm anything. What it does do is keep the 6.48 S1 shelf intact and put R1 6.62 back in play as the first upside objective.

The last completed weekly bar (2026-09-21 to 2026-09-25) opened 6.72, high 6.93, low 6.7, closed 6.77, +1.11% w/w — a strong week that tagged the 52-week high. The current week (from 2026-09-28, four sessions) is unfinished and last prints 6.54, -3.37%; no weekly-close conclusion can be drawn from an open bar. The practical read: the prior completed week established 6.93 as the reference high, and this week is retracing that advance. Holding 6.43–6.48 keeps the retracement orderly.

View: constructive while above 6.48 (S1); a settle below 6.43 (S2) flips the structure to neutral-to-bearish and targets the 6.35 channel floor.

3. Supply-Demand Balance & Fundamental Drivers

Visible inventory data does not support the bearish price action. COMEX registered stocks were 471,400 short tons as of 2026-09-29, up 1,100 short tons d/d (+0.2%) — a marginal build, not a flood. LME warehouse stock was 249,400 MT as of 2026-09-30, down 1,075 MT d/d (-0.4%), though the 20-report change is +15,550 MT, meaning the LME has seen a net inflow over roughly the past month. SHFE warrants were 10,011 MT as of 2026-09-30, down 721 MT d/d (-6.7%), following a 2,693 MT draw the prior session to 10,732 MT. The SHFE warrant line is the tightest of the three: a 10,011 MT warrant book is small in absolute terms and the two-day draw is meaningful.

Netting these: the LME has rebuilt some cushion over the past month, COMEX is broadly flat, and China's exchange deliverable stock is draining. That is a mixed but not bearish inventory picture, and it is inconsistent with the -3.71% 5D move being fundamentally driven.

On the supply side, the news flow carries a reminder of how thin the mine-supply buffer is. A fatal accident was reported at Codelco's Radomiro Tomic mine (2026-10-01), one of the largest copper operations globally; incidents of this type carry operational and regulatory risk. Separately, Generation Mining began construction at the Marathon copper-palladium project in Ontario (2026-10-01) — a positive medium-term supply addition, but one that is years from first concentrate and irrelevant to the next two quarters. The balance of headline risk in the near term skews toward supply disruption rather than supply surplus.

Macro transmission runs through the dollar and rates. DXY is 102.04, +0.58% on 2026-10-01, and the US 10-year yield is 5.237, -1.06%. A firmer dollar is the single cleanest explanation for the prior-session copper decline, and it is a headwind that can persist. But the copper/gold ratio at 1.58 (96th one-year percentile) says the market is still pricing industrial demand as intact relative to the defensive metal — a configuration that typically requires a genuine growth scare, not a dollar bounce, to break.

View: inventories and mine-supply headlines are net supportive; the dollar is the live risk. Bias stays long while LME stock does not accelerate higher and SHFE warrants keep draining.

4. Positioning & Fund Flows

CFTC managed-money positioning is the clearest two-sided risk in this setup. As of 2026-09-22, open interest was 301,657, longs 96,421, shorts 13,899, net 82,522, a weekly change of +17,416. That is a large one-week build in net length. The prior week (2026-09-15) showed net 65,106, a change of -17,048, so the market swung from a heavy reduction to a heavy addition in a single reporting week.

Crowding metrics: netPct 27.36% of open interest as of 2026-09-22, a 69.05 three-year percentile. That is elevated but not extreme — it is not above the 90th percentile, and the |z|>1.5 extreme flag is not triggered. The CTA trend-following proxy reads 98, essentially fully long, and has been pinned there for four consecutive weeks (2026-09-01 through 2026-09-22). The hedge ratio is 56.61%.

The interpretation matters. A CTA proxy at 98 with net length at the 69th percentile means the systematic community is already positioned for higher prices. That is fuel for a squeeze if the market breaks higher, but it is also a source of mechanical selling if the trend signal flips — and the -3.71% 5D move is exactly the kind of price action that starts to erode trend signals. The +17,416 weekly net build on 2026-09-22 was followed by a price decline into 2026-10-01, so the most recent cohort of longs is underwater. That is a vulnerability, not a thesis.

On volatility: RV20 is 26.8%. The cross-asset implied-vol backdrop is subdued — ^VIX 16.39 (35th one-year percentile), ^GVZ 23.32 (16th percentile), ^OVX 51.69 (51st percentile). With copper realized vol at 26.8% and broad-market implied vol cheap, optionality is not obviously expensive, which favors expressing the long via defined-risk structures rather than outright leverage.

View: positioning is a headwind to near-term upside but not a reason to be short. Crowding at the 69th percentile leaves room before it becomes a contrarian sell signal.

5. Cross-Asset Relative Value

The key ratio in the feed is copper/gold at 1.58, in the 96th percentile of its one-year range and the 53rd percentile of its three-year range. The one-year reading is the important one: copper has been strong relative to gold over the past twelve months, and the current level is near the top of that band. A rising copper/gold ratio is a pro-growth signal; a ratio at the 96th one-year percentile says the market has been pricing reflation, not recession.

The three-year percentile at 53.44% is the nuance. Over a longer window, copper/gold is mid-range — the past year's strength has recovered ground lost earlier, rather than establishing a multi-year extreme. That argues against calling the ratio stretched, and it argues against treating the current copper pullback as the start of a regime change in the growth complex.

Cross-checking with the dollar and rates: DXY 102.04 (+0.58%) and ^TNX 5.237 (-1.06%). A 5.24% ten-year yield is a restrictive backdrop for industrial metals on the margin, but the fact that copper/gold is holding near its one-year high despite that yield level is a sign of underlying demand resilience. If the ratio were breaking down while yields sat at 5.24%, the bear case would be far stronger.

View: copper/gold at the 96th one-year percentile is a tailwind for the long bias and a reason to fade the severity of the current drawdown. Watch the ratio as the cleanest cross-asset confirmation — a decisive break lower would undercut the long thesis before price does.

6. Historical & Seasonal Patterns

Seasonality for the same calendar start over the next 20 sessions, measured across the last 15 years: mean +1.42%, median +1.88%, and up in 10 of 15 years. The best instance was 2011 at +15.57%; the worst was 2012 at -6.62%.

The distribution is favorable but the sample is small and the tails are wide. A 10-of-15 hit rate is roughly a 67% historical frequency, and the median (+1.88%) exceeds the mean (+1.42%), which indicates the average is dragged down by the -6.62% 2012 outlier rather than by a cluster of moderate losses. In other words, the typical October-to-early-November window is modestly positive, with occasional violent downside.

Applied to the current setup: the seasonal window is a mild tailwind that aligns with the long bias, but it is context, not a trigger. The 2012 worst case (-6.62%) is a useful reminder of what a macro-driven break looks like — and it is the scenario the 6.43 invalidation level is designed to contain. Seasonality supports holding a long through the window; it does not justify adding size.

View: seasonal bias is modestly positive and consistent with the long call, but the wide left tail reinforces disciplined stops over conviction sizing.

7. Scenario Analysis (Base / Bull / Bear)

Base case — 50%: range repair toward 6.7. Trigger: the Asia-session reclaim of pivot P 6.56 holds through the US session, and the 2026-10-02 payroll print does not deliver a large upside surprise on Average Hourly Earnings. Path: price grinds from 6.54 through R1 6.62 toward R2 6.7, with the 20-day channel top at 6.93 as the outer bound. Action: hold the long, trail the stop up behind 6.48 (S1) once 6.62 trades. This case agrees with the section 1 call.

Bull case — 30%: breakout retest of 6.93. Trigger: a soft dollar reaction to the payroll data (DXY back below 102) plus continued SHFE warrant draws below 10,000 MT. Path: 6.7 clears, R2 gives way, and the market retests the 52-week high at 6.93 tagged in the completed week of 2026-09-21 to 2026-09-25. Action: add on a settle above 6.7, target 6.9–6.93, and tighten risk to breakeven. The CTA proxy at 98 means a break higher would force systematic re-buying, which is why this path can move fast.

Bear case — 20%: settle below 6.43. Trigger: a hot payroll print (Non-Farm Employment Change above the 89K forecast plus the 73K surprise threshold, or Average Hourly Earnings above 0.4%) driving DXY higher, combined with LME stock extending its 20-report build of +15,550 MT. Path: 6.48 (S1) fails, 6.43 (S2) settles through, and the market targets the 20-day channel floor at 6.35. Action: exit the long on the settle, stand aside, and re-engage only on a reclaim of 6.48. A break of 6.35 would open the 52-week structure below.

Probability-weighted, the distribution is skewed to the upside: 80% of outcomes sit at or above the current settle, with the base case targeting 6.7.

8. Trading Strategies & Risk Management

Strategy 1 — Core long (HGZ26.CMX). Entry 6.54–6.57 (settle 6.54; Asia 6.57). Stop 6.42, below S2 6.43 and roughly one ATR14 (0.137) beneath entry. Target 6.7 (R2), secondary 6.9. Horizon 1–3 weeks. Conviction 6/10. Size: half of normal risk budget, given the 69.05 crowding percentile and the CTA proxy at 98 — the position is being added into a market where systematic length is already maximal.

Strategy 2 — Pullback add. Entry 6.48–6.5 on a retest of S1 that holds on a closing basis. Stop 6.42. Target 6.7. Horizon 1–2 weeks. Conviction 5/10. Size: quarter of normal risk budget, only if Strategy 1 is already working. This is a scale-in, not a separate thesis.

Risk management: the single largest event risk in the window is the 2026-10-02 payroll cluster (Average Hourly Earnings, Non-Farm Employment Change, Unemployment Rate, all at BJT 20:30 / ET 08:30). Consider halving exposure into that print or expressing the position with defined-risk options, given RV20 at 26.8% against cheap broad-market implied vol. Do not add on the day of the FOMC minutes (BJT 10-08 02:00 / ET 10-07 14:00).

9. This Week's Data Calendar

BJT 10-02 20:30 | ET 10-02 08:30 — USD Average Hourly Earnings m/m (F 0.3%, P 0.3%), Non-Farm Employment Change (F 89K, P 162K), Unemployment Rate (F 4.1%, P 4.1%). BJT 10-05 22:00 | ET 10-05 10:00 — ISM Services PMI SEP (F 54, P 55.4). BJT 10-08 02:00 | ET 10-07 14:00 — FOMC Minutes. The payroll cluster is the dominant near-term risk for the dollar and therefore for copper.

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 2026-10-02

Gold (GC=F) Deep Dive | 2026-10-02 — Bearish Below Pivot 4198, GCZ26 | 2026-10-02

Copper (HG=F) Deep Dive | 2026-10-02 — Constructive Base, 6.48 Pivot Is the Line | 2026-10-02

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