# Weekly futures review — September 20, 2026

Research cutoff: September 20, 20:34:15 Toronto / September 21, 00:34:15 UTC. Investment horizon: September 13, 2026–January 13, 2027. Paper portfolio: $1,000,000 initial capital.

## Decision: one small gold starter enters the paper book

Enter **+1 December Micro Gold, MGCZ26**, in the simulated portfolio. A conditional plan was staged first; a genuinely later exact-contract last of **4,406.9 at 00:34:15Z** then qualified the **4,385–4,410** band. The deterministic one-adverse-tick convention produced a **4,407.0 paper fill**. Stop **4,265**, target **4,680**, and compulsory exit **October 30 at 20:00 UTC** remain fixed. Use a disclosed **$10,000 modeled margin reserve**. Do not widen the stop, add automatically or roll.

After entry, paper NAV is **$999,995.50**, cash **$999,996.50**, open P&L **−$1.00**, entry commission **$3.50**, modeled entry slippage **$1.00**, gross exposure **$44,069**, margin reserve **$10,000** and open risk **$1,783**. There are no pending orders. This is a low-confidence test of a small positive edge, not a claim that a generic safe haven must appreciate. No real brokerage order was placed.

**MESH27 is the best rejected current-entry challenger.** Its exact 7,817 observation is too expensive for the model's sensitivity. Retain only a finite research condition at or below 7,750 after September 23 U.S. flash PMI confirms expansion without faster output-price pressure. It is not ready for a paper order.

Resolve the weak incumbent ideas now: remove coffee, February crude, copper, yen, euro, wheat/soy, nickel, sugar, SOFR, gas and OAT/Bund from the actionable board. Keep their reasoning and reactivation standards as context; do not disguise retirement as an endless wait. December micro WTI retains one specific physical-evidence review, not a second starter.

## Regime and what is genuinely new

The central tension is resilient nominal demand versus restrictive policy and energy-sensitive inflation. The Fed raised its range to 3.75–4.00% on September 16. August nominal retail sales rose 1.2%, while manufacturing output fell 0.3%. CPI rose 0.4% monthly/3.4% yearly; core rose 0.3%/2.4%. One weak factory month is not a broad recession signal, and nominal sales are not proof of accelerating real consumption. [Federal Reserve decision](https://www.federalreserve.gov/newsevents/pressreleases/monetary20260916a.htm), [Census retail sales](https://www.census.gov/retail/marts/www/marts_current.pdf), [Fed industrial production](https://www.federalreserve.gov/releases/g17/Current/default.htm), [BLS inflation](https://www.bls.gov/news.release/cpi.nr0.htm)

The BOJ's 7–2 decision raised its guideline to 1.25%, effective September 24, but the yen weakened after the decision. The ECB deposit rate became 2.50% on September 16. Anticipated tightening by several central banks is not automatically fresh dollar or yen alpha. [BOJ decision](https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2026/k260918a.pdf), [Reuters yen reaction](https://www.reuters.com/world/asia-pacific/yen-weak-ahead-boj-decision-rate-hike-expected-2026-09-18/), [ECB rates](https://www.ecb.europa.eu/stats/policy_and_exchange_rates/key_ecb_interest_rates/html/index.en.html)

Saudi escalation sustains an oil supply tail, but reported sales, rerouting and uncertain repair duration prevent equating damaged capacity with net lost exports. U.S.–China talks on trade, AI and critical minerals supply an opposing risk-relief channel; no completed agreement is assumed. [Reuters September 20 oil](https://www.reuters.com/business/energy/oil-rises-after-houthi-attack-saudi-capital-2026-09-20/), [Reuters Saudi sales via Dawn](https://www.dawn.com/news/2030842), [Reuters talks](https://www.reuters.com/business/finance/us-treasurys-bessent-chinas-he-launch-talks-ai-trade-critical-minerals-2026-09-20/)

After re-examination, the common subjective regime tree remains persistent friction 40%, transport normalization 20%, renewed supply shock 25%, growth break 10%, and AI/liquidity disappointment 5%. These are judgmental framing weights, not calibrated odds or inputs mechanically reused for every trade. Trade-specific models below are new current-entry diagnostics. The original September 13 forecasts remain historical and are not scored before their January horizon.

## Gold: why one micro earns a starter

The mechanism is investment demand and recovery after the policy shock, with a defined price ceiling. August gold ETFs added $18 billion and 121 tonnes, reaching 4,189 tonnes. Those figures were published September 9: they establish prior demand, not post-Fed inflows. Official ten-year real yields were 2.68% September 16 and 2.61% September 17, versus 2.60% September 11: still restrictive, but not an accelerating rise in the latest supplied series. [World Gold Council](https://www.gold.org/goldhub/research/gold-etfs-holdings-and-flows/2026/09), [Federal Reserve H.15](https://www.federalreserve.gov/releases/h15/)

September 15 managed-money gold positioning was 142,394 long less 9,278 short, net 133,116, down 1,856. This predates the Fed decision and is neither December-micro ownership nor evidence of an uncrowded trade. The risk of liquidation remains. [CFTC metals](https://www.cftc.gov/dea/futures/other_sf.htm)

The later exact MGCZ26 last is **4,406.9 at September 21 00:34:15Z**, after the conditional plan was recorded. The same response displayed bid 4,407.0, ask 4,407.1, volume 23,716 and open interest 70,498; only the raw last carried its own clock. The deterministic paper engine added one adverse 0.1 tick for a **4,407.0 fill**. Completed five-minute observations through 00:05Z sum 15,667 contracts and stayed above the 4,371.2 pre-entry support test; this is a conservative observed-trade lower bound, not a certification of every exchange print.

MGC represents ten troy ounces, with a 0.1 price tick worth $1. The 4,265 stop is below the retained exact-contract September 16 low of 4,273.1; it was not tightened to force the position size. One standard GC at equivalent geometry exceeds $18,000 buffered risk, so the exact micro matters. [COMEX Micro Gold rules](https://www.cmegroup.com/rulebook/COMEX/1a/120.pdf)

### Payoff and the assumptions that can break it

At the **maximum actual fill of 4,410**, notional is $44,100. Raw stop loss is $1,450; buffered planned risk is `145 × $10 × 1.25 + $7 commissions + $1 exit tick = $1,820.50`, or **0.18205% NAV**. Entry slippage is already embedded in the actual-fill assumption.

| Path through October 30 | Subjective probability | Exit | Net per micro |
| --- | ---: | ---: | ---: |
| Bull recovery, without an earlier stop | 30% | 4,680 | +$2,660 |
| Base recovery/time exit, without an earlier stop | 35% | 4,550 | +$1,360 |
| Ordinary stop before recovery | 30% | 4,265 | −$1,490 |
| Gap through stop | 5% | 3,950 | −$4,640 |

Each path deducts a **$40 scenario cost/stress reserve**: $7 commissions, $1 exit tick and $32 additional execution stress. This reserve is separate from the paper simulator's commission/slippage accounting, not an additional charge automatically booked.

Expected net is **+$595**, or 0.0595% NAV and 1.3492% entry notional—not a return on margin. At illustrative 4,409.2, the corresponding figures are +$603 expected net and $1,810.50 buffered risk.

Sensitivity is decisive:

- Moving ten probability points from bull recovery to ordinary stop leaves **+$180** at the 4,410 cap.
- Moving five points from bull recovery to the gap outcome leaves **+$230**.
- Applying both shifts gives **−$185**.
- Lowering both profitable exit prices by $100 gives **−$55**.
- An additional $50 cost reduces the baseline to **+$545**.

These assumptions support one micro, not a robust high-conviction allocation. The gap example stays below the $5,000 thesis limit, but a worse gap can exceed it. Discrete paper observations do not provide a continuously enforced stop.

The final **4,410** cap superseded the earlier 4,430 and 4,415 alternatives. The observed 4,406.9 last qualified before the September 21 noon Toronto analyst deadline. Reassess the filled position on September 25 and earlier if its thesis fails. The **4,265** stop, demand/yield invalidation, 4,680 target and October 30 time exit now govern it.

The $10,000 margin amount is a conservative **paper reserve, not verified exchange initial margin**. AMP displays root MGC maintenance of $2,426, without an effective date; the reserve is approximately 4.12 times that broker figure. Separately, the retained same-day CME December maintenance reference is $2,213. These different bases are not interchangeable. Carried December lifecycle dates are first position November 27, first notice November 30 and last trade December 29; conservative date-start bounds are management conventions, not exact exchange event clocks. The October 30 exit precedes all of them. After exit, assume USD cash earning 0% through January 13; no free roll/carry return is added. [AMP margin table](https://www.ampfutures.com/trading-info/margins), [CME gold lifecycle](https://www.cmegroup.com/markets/metals/precious/e-micro-gold.calendar.html), [CME margins](https://www.cmegroup.com/markets/metals/precious/e-micro-gold.margins.html)

## Best rejected challenger: March micro S&P

MESH27's exact last was **7,817 at September 21 00:03:31Z**. At $5 per point, its notional is $39,085. January futures outcomes of **8,600/8,000/7,000**, weighted **25%/50%/25%**, yield net **+$3,905.50 / +$905.50 / −$4,094.50** after $9.50 two-sided friction. Expectation is **+$405.50**, but moving ten points from base to bear makes it **−$94.50**. Reducing every terminal level by 100 points has the same effect. These terminal scenarios are not calibrated stop-before-target probabilities.

A 7,400 fundamental review stop gives $2,615.75 buffered risk at 7,817, slightly above the usual starter band but below the thesis ceiling. The primary rejection is fragile economics, not contract size. The proposed 7,400 level is a risk tolerance, not an independently sourced technical-support claim.

At **7,750**, the same model improves to **+$740.50**, remains **+$240.50** after the probability stress, and uses **$2,197** buffered risk. Retain this exact finite condition: September 23 U.S. flash composite PMI at least 50, no acceleration in its reported output-price measure, then a fresh MESH27 price **≤7,750 and >7,400**. Review by **September 23, 20:00 Toronto**: re-underwrite, materially revise or drop. A contraction-driven selloff does not qualify. Target 8,600; 8,000 is an interim review, not a partial exit of one indivisible contract; exit by January 13. [PMI date calendar](https://www.investing.com/economic-calendar/s-p-global-composite-pmi-1492)

It is **not ready for an order**. Returned volume 78 and open interest 1,882 lack independent session/vintage clocks; the observed volume is too small for the proposed one-contract simulation. Dated margin/reserve and complete lifecycle evidence remain necessary. Native March last trade is March 19, 2027, beyond the horizon; no December roll is assumed. [CME micro-index specifications](https://www.cmegroup.com/education/courses/micro-e-mini-futures/micro-e-mini-futures-products-overview)

## Oil: correct the arithmetic, then judge the physical uncertainty

The September 16 EIA report shows crude stocks down 0.6 million barrels to 423.4 million, but total commercial petroleum up 2.6 million and four-week supplied products down 0.6% yearly. Distillate stocks remained 13% below their five-year norm. Product tightness and demand adjustment coexist; this is not a uniform shortage. [EIA petroleum](https://www.eia.gov/petroleum/supply/weekly/pdf/highlights.pdf)

The latest supplied MCLZ26 last is 91.90 at 00:03:28Z; displayed ask 91.97 plus a tick gives illustrative 91.98. A supportive **25/35/30/10** path tree at **100/96.4/88.4/82**, less $20 per micro, gives **+$128**. A conservative **20/30/40/10** tree at **100/95/88.4/82** gives **−$12**. That economically plausible normalization case is the reason to wait for better physical evidence.

The alternative **30/35/30/5** tree has mean exit 94.36 and correctly gives **+$218 per micro**, with **+$102** after shifting ten points bull to ordinary stop. The preliminary comparison understated these by $120 per contract. Correcting that error matters: the case for preferring gold is stronger support for its limited recovery test and less dependence on unmeasured export persistence, **not** a false claim that oil fails that particular sensitivity.

At 91.98/88.40, buffered risk is $455.50 per micro or $1,366.50 for three. Numerical capacity alone does not authorize any quantity. Later completed-bar volume 1,077 conflicts with the earlier untimed snapshot's 1,246; Friday daily versus final five-minute closes also differ. No reliable weekend return or execution-qualified session total is inferred.

Reassess new Saudi loading/repair evidence September 21 morning and after September 23 EIA. Require credible net losses persisting into December despite rerouting, robust current-price expectation and intact 90.60 pre-entry support. Otherwise demote. Any later qualified December micro plan should exit well before its November 19 termination; October 30 is the tactical reference. No order tonight. [CME Micro WTI](https://www.cmegroup.com/trading/energy/files/micro-wti-crude-oil-futures-fact-card.pdf)

February WTI is demoted separately. At MCLG27 illustrative 86.05, a 25/35/30/10 tree at 104/88/76/70 gives only **+$35** after $20, turning **−$245** after ten points bull→stop. One micro's $1,264.25 buffered risk fits, but winter-duration evidence does not. Its returned volume 51 is also thin and unclocked. The old $82–84 band is no standing order.

## Resolution of every other incumbent

| Thesis | Economic decision and evidence | What could justify a genuinely new idea |
| --- | --- | --- |
| Coffee short | **Avoid/demote.** Certification improved, but 268.80 offers poor short geometry and one standard contract exceeds the risk limit. | A new favorable failed rally, continuing certified growth and a verified smaller policy-sized expression. |
| Copper short | **Avoid/demote.** 6.8185 standard / 6.8145 own micro is not a failed retest. Micro size fits, but observed volume is thin and the tariff/physical reversal is unproven. | Actual refined-copper policy clarity, shrinking regional premium/loosening deliverable stocks and a prospective rejection. |
| Wheat/soy | **Retire actionable setup.** Official 2W/S is 1.098157; wheat stocks rose while projected U.S. soy exports strengthened. | Delivered-wheat basis/export improvement and new explicit leg-level dollar-loss scenarios. |
| Yen | **Avoid/demote old squeeze.** The expected BOJ hike did not produce durable currency demand; lagged funds were already net long. | A fresh post-September 24 demand thesis and verified March price/carry/lifecycle. |
| Euro short | **Avoid/demote chase.** March micro sizing works, but current-price EV is negative. | A new rally with independently improved economics, not automatic reuse of the old band. |
| SOFR flattener | **Avoid/context only.** The new 25.5-bp slope is explainable by restraint; one pair fits risk but has negative modeled EV. | Material activity/policy evidence that changes the payoff distribution. |
| OAT/Bund | **Avoid/context only.** No DV01-matched package is defined; raw prices are not sovereign yield spreads. | CTD, conversion factors, repo/basis, DV01 and integer hedge counts. |
| Gas | **Demote.** Indicative January–March premium 0.932 has compressed; national surplus does not remove regional winter convexity. | A newly enlarged premium and stronger regional salt/LNG/weather evidence, with a horizon-safe plan. |
| Sugar | **Avoid; stay suspended.** Liquidation is observed, but 18.28 leaves little downside reward against squeeze risk. | A new favorable rally/rejection plus actual exportable-supply improvement. |
| Nickel | **Avoid/context only.** Ore restart is not verified class-I surplus; exact instrument/access data remain missing. | Quantified Weda Bay volumes/grades and verified deliverable-metal conversion and instrument. |

Important factual detail must survive the demotions:

**Coffee:** StoneX's September 18 report records **12,990 bags graded, 10,528 added to certified stocks, total 258,415**. It is wrong to say conversion remains wholly unobserved; the official ICE daily total was not independently recaptured. At 268.80, targets 255/240 leave 13.80/28.80 cents versus 36.20 to 305. A diagnostic short tree of 30/30/30/10 at 240/265/305/330, less $100, gives **−$2,800 per standard**. Buffered standard stop risk is **$16,968.75 before costs**, and still $9,375 at a 285 entry. No verified smaller same-underlying future was found. [StoneX September 18](https://www.stonex.com/en/insights/daily-coffee-report-9-17-26-2026-09-18/), [ICE Coffee C](https://www.ice.com/products/15/Coffee-C-Futures)

**EUR positioning:** September 15 leveraged longs **103,260** minus shorts **131,416** equal **−28,156**. Weekly changes were **+8,452 longs and +3,323 shorts**, so the net short **narrowed by 5,129**, from −33,285—not fresh net short accumulation. Existing short exposure still matters, but the flow interpretation must be accurate. Yen leveraged funds were net long 23,170 before the BOJ. [CFTC financial futures](https://www.cftc.gov/dea/futures/financial_lf.htm)

**Euro economics:** M6EH27's own March row was 1.1561 at September 20 23:45Z, not the December standard quote. Short outcomes 1.1250/1.1500/1.2000 weighted 25/45/30 yield **−$42.625 per micro** after $9.50 friction. Three micros risk $2,086.31. Feasible sizing therefore does not rescue the trade. [Exact March row in micro-euro contract table](https://www.wsj.com/market-data/quotes/futures/M6EZ26)

**SOFR:** SR3Z26 95.680 at 00:02:29Z and SR3H27 95.425 at 00:03:20Z imply **100 × (95.680−95.425) = 25.5 bp** higher later-quarter rate. The 51-second-skew lasts are indicative, not executable bid/asks or meeting probabilities. For +H27/−Z26, outcomes 0/25/65 bp weighted 25/50/25 give **−$145.25** after $64 friction; buffered risk to 60 bp is $1,142.13. [CME SOFR](https://www.cmegroup.com/markets/interest-rates/stirs/three-month-sofr.contractSpecs.html)

**Physical balances:** September WASDE raised world wheat ending stocks 3.0 million tonnes to 276.3 million and cut U.S. soybean ending stocks 10 million bushels to 310 million through stronger exports. September 18 settlements were wheat 730.00 and soy 1,329.50. September 17 export-sales outcomes remain unreviewed, not zero. Gas stocks of 3,298 Bcf were 118 above the five-year norm, but South Central salt was 222 Bcf, 8.3% below its norm. These facts oppose simplistic wheat-demand or aggregate-gas-surplus claims. [USDA WASDE](https://www.usda.gov/oce/commodity/wasde/wasde0926.pdf), [USDA settlements](https://www.ams.usda.gov/mnreports/ams_3192.pdf), [EIA gas](https://ir.eia.gov/ngs/ngs.html)

Sugar gross longs fell 10,397 and shorts rose 2,857, lowering net length 13,254 to 225,430. That meets one old condition, not the payoff or supply test. Nickel's source explicitly cannot reliably estimate remaining-2026 Weda Bay ore volumes/grades. [CFTC agriculture](https://www.cftc.gov/dea/futures/ag_sf.htm), [Eramet](https://www.eramet.com/en/news/pt-weda-bay-nickel-restarts-its-mining-operations/)

## Other challengers and global coverage

March corn is the strongest new agricultural alternative, but stays off the actionable board. WASDE lowered yield to 178.5 bushels/acre and ending stocks by 86 million bushels; March nevertheless carries 14 cents over December and delivered basis has not broadly accelerated. At a 541.75 hypothetical slipped settlement reference, 25/40/30/5 outcomes 600/560/500/470 give **+$255.50** after $32, turning **−$244.50** after ten points bull→stop. Buffered standard risk is $2,628.88. The own MZCH27 quote is **545.5**, not 541.5, and returned volume is only eight. Reassess Crop Progress and delivered demand, not an automatic order. [USDA WASDE](https://www.usda.gov/oce/commodity/wasde/wasde0926.pdf), [USDA cash/futures](https://www.ams.usda.gov/mnreports/ams_3192.pdf)

Short MNQH27 is rejected: observed investment resilience conflicts with the anti-AI thesis, and one micro at 30,407 against a 32,750 adverse level needs $5,865.50 buffered risk. U.S. duration has conflicting inflation and growth channels; December ZNZ26 short risk to 110 is $5,135.91 and would require an early exit or separately underwritten roll. A micro yield future is not a scaled deliverable-note substitute.

Japan's NKDZ26 65,240 and Australia's APZ26 8,735 have real Sunday clocks but no independently superior price/hedge thesis. Hong Kong's HSIZ26 24,764 remains a Friday last; European index close fields lack observation clocks. China trade talks are not an agreement. Canada/Latin American FX carry and oil sensitivity do not establish a verified deferred relative-value package. India/South Asia exact deferred pricing and hedge inputs were not recovered. These are honest coverage limits, not proof that local opportunities do not exist. The review is a substantive liquid global screen, not an exhaustive quantification of every exchange contract.

Silver adds industrial sensitivity to precious-metal exposure. Platinum's latest 265-koz 2026 surplus forecast argues against reviving the old deficit pair, while only 3.4 months of above-ground demand cover argues against a casual surplus short. Diesel's relative physical case is stronger than undifferentiated oil, but the January HOF27–RBF27 indication is already $1.6129/gallon with seven-minute skew; a 10-cent adverse move costs $4,200 per standard pair. No new hedge or artificially tight stop is approved. [WPIC quarterly](https://platinuminvestment.com/supply-and-demand/platinum-quarterly), [HOF27](https://www.wsj.com/market-data/quotes/futures/HOF27), [RBF27](https://www.wsj.com/market-data/quotes/futures/RBF27)

The seven already retired ideas—crude normalization, Dow/Nasdaq rotation, cocoa, CAD/EUR, platinum/gold, ULSD/RBOB and October/December VIX—remain retired. No current superior case, risk-matched package or valid lifecycle justifies reinstatement. January VIX at 19.75 has no positive gross expectation under the supplied 30/17.5/14, 25/50/25 diagnostic and is negative after friction; no equity holding currently needs that hedge.

## Next decisions and weekly accountability

Within the next 72 hours: monitor the live gold stop/thesis without widening it; Crop Progress September 21 at 16:00 Toronto; daily Saudi export/repair and coffee certification evidence; Jefferson's listed September 22 speech; September 23 flash PMIs, EIA petroleum at 10:30 Toronto, and the 20:00 Toronto MESH27 decision. The PMI date is supported by a vendor calendar, not a verified primary release clock. Fed speech times should be rechecked because the calendar footer is stale. September 24 U.S. gas storage is outside the immediate 72-hour window. The BOJ's September 24 Japanese effective date can overlap that UTC window; its exact effective clock is not supplied. [USDA calendar](https://www.nass.usda.gov/Publications/Calendar/reports_by_date.php), [Fed September calendar](https://www.federalreserve.gov/newsevents/2026-september.htm), [EIA petroleum schedule](https://www.eia.gov/petroleum/supply/weekly/schedule.php), [BOJ holidays](https://www.boj.or.jp/en/about/outline/holi.htm)

The weekly lesson is specific: standard-contract granularity was too restrictive for gold, but finding micros does not make every weak thesis investable. Coffee's decline while the book stayed flat is a missed direction, not earned profit; its certification evidence nevertheless deserved correction. Copper's rise shows why a price entering a fade zone is not itself a rejection. Yen's anticipated policy story did not produce the required demand. Those reviews end in decisions, not reset deadlines.

Gold, oil, copper, sugar and risk assets share exposure to rates, shipping and nominal demand; they are not independent diversifiers. The filled MGC starter has 4.4069% gross exposure, a 1% modeled margin reserve and 0.1783% buffered risk. The rest remains cash unless a new idea independently earns risk. Idle capacity is not an investment thesis.

Source observation times, publication times and retrieval times remain separate. Native IBKR responses support the exact numerical market observations; provider documentation is attribution, not a public quote page. Friday softs, carried bonds and unresolved yen normalization have not been relabeled live. All scenario probabilities, targets and stress reserves are analyst judgments. The paper fill follows the documented scheduled-observation convention; it is not a real order, continuous market feed or guaranteed stop.