Keep the yen thesis. Preserve cash and the entry gates.

ATG daily monitor — September 8, 2026

Research cutoff: 2026-09-08, 20:50:01 EDT America/Toronto. Comparison: September 7 monitor and September 6 complete monthly review. Horizon unchanged: September 6, 2026–January 6, 2027.

Decision

Wait for China CPI/PPI at 21:30 EDT. No new entry, add, reversal or prospective paper intention qualifies. Retain the yen *research* thesis without adding; retain the eight conditional watches and four retirements. The simulated paper portfolio remains $1,000,000 cash, zero positions and zero pending intentions.

The official NBS calendar schedules September 9 inflation data at 09:30 China time, equivalent to September 8 at 21:30 Toronto. No distinct pre-event edge is established. Do not infer the outcome before release. NBS release calendar

Status is partial, not a new complete review: current exact public initial-margin estimates, some execution-certified notice/termination clocks, executable package depth and OAT/Bund hedge inputs remain unavailable. Several fresh official settlements have no observation clock. These limitations block exposure, but do not make an all-cash USD NAV uncertain. The September 7 research is the comparison baseline, alongside the September 6 full monthly review. This update is not a new full-universe scan.

What changed since the last monitor

• Yen appreciated a further 0.38%; Japan's GDP revision is now reviewed. Remaining value under unchanged September 6 assumptions fell to approximately 0.61% of current outright notional.

• Platinum/gold rose another 0.90%, chiefly because gold fell faster. The old model now leaves only about 0.11% gross remaining sensitivity before fees and stop-path effects. The new WPIC report and ordered close/retest sequence remain unverified.

• March wheat/soy has conflicting *time-specific* readings: the September 8 settlements put the ratio above the entry band; later asynchronous last trades put it inside. Wheat inspections and basis do not deliver the required physical confirmation.

• January–March gas compressed to $0.960, further below its entry threshold. ULSD/RBOB pulled back to $1.3613, improving price asymmetry but still offering a poor entry.

• Official VIX and Eurex December settlements were recovered. Better prices do not cure missing calendar edge or CTD/DV01 hedge definition. SOFR's actual expiry interpretation is corrected below.

These changes are market observations and research judgments, not paper performance.

All nine monitored theses

“Priced in” is subjective 0–5 research judgment, not a measured probability. “Hold” refers only to a surviving thesis. All watches are unactivated.

Every quote, observation time and source is listed in the appendix. Changes compare these specific observations, not exchange daily-change fields. A ratio change is not a linear package return.

1. Yen — thesis survives; do not chase

The official second estimate raised Japan's Q2 real GDP to +0.4% q/q and +1.4% annualized from +0.3%/+1.1%. Consumption was flat; domestic demand contributed −0.1 percentage point and net exports +0.5. The opened Reuters report cites +1.6% annualized consensus: this was an upward revision, not a consensus beat. The release occurred just after the previous cutoff but slightly more than 24 hours before this one, so it is since-last-monitor information, not fresh rolling-24-hour news. Cabinet Office release, Reuters GDP report

Entry/add remains 0.00635–0.00639 with intact policy and durable currency-demand evidence; the quote is above it. Targets 0.00670–0.00680 and research invalidation 0.00618 remain. The cited Reuters swap estimate already put a September BOJ hike near certainty; it is dated reporting, not a newly measured live curve. September 1 leveraged-fund net −102,188 cannot establish current short-covering. Expensive imported energy and stronger U.S. rates remain the main counterarguments. CFTC financial report

Remaining expected-value sensitivity is lower, score stays 4. China tonight and U.S. inflation are the immediate tests. A confirmed pullback with policy support could improve the decision; adverse policy divergence invalidates it. Review December 7–11 roll to March before December 14 last trade, charging actual carry and adjusting levels by the actual calendar gap.

2. Sugar — a fresh rejection pattern, not a retrospective fill

September 8's 19.39 high and 19.09 vendor settlement are consistent with a new rejection of the 19.20–19.40 band after the thesis began. They cannot prove an executable intraday failed-retest fill. The reported last was 19.08 at 12:59 EDT, below the band by review time. A prospective intention did not exist beforehand and cannot be backfilled. Exact SBH27 vendor quote

Require renewed tradable failed-retest evidence and exportable-supply/sugar-share confirmation. Targets remain 18.00 then 17.50; 20.15 is conditional invalidation after activation. Crowding alone is insufficient, and expensive oil can favor ethanol over sugar. The known August ISO outlook projects only a small next-season deficit and depends on Brazilian allocation; the retrieved primary UNICA report is stale, not new August/September confirmation. ISO August outlook, UNICA retrieved report The price-only fixed-scenario sensitivity improves slightly to 2.80%, while the adverse ethanol/supply tail argues against treating that as a better forecast. Score remains 2; high monitoring urgency, no chase. March covers January 6 without a roll. Physical delivery and exact lifecycle/safety data still require execution verification.

3. Wheat/soy — physical confirmation remains absent

USDA-reproduced March settlements were 761.50/1,338.50, ratio 1.137841, above the 1.115–1.135 band. Later CME marks were 756.50/1,336.25, ratio 1.132273, but the legs were 404.285 seconds apart. Neither reading substitutes for the required post-WASDE evidence.

September 8 inspections showed wheat 342,733 tonnes versus revised 431,977, down 20.7%; marketing-year shipments 5.130 million tonnes were 27.7% below the comparable year-earlier figure. Soybean inspections rose to 422,016 from 284,051. Gulf SRW basis was unchanged at 25–55 over December; soybean basis unchanged at 102–103 over November. Thus basis confirmation is not strengthening, rather than merely unobserved. Soybean good/excellent condition was 58%, unchanged weekly versus 64% a year earlier; spring-wheat harvest was 86% versus 83% average. USDA inspections, USDA daily grain/basis report, USDA Crop Progress

Wait for September 11 WASDE and delivered-demand/basis improvement; demote if two subsequent export cycles fail. Targets 1.18–1.20, conditional stop 1.08. Larger exportable wheat supply, normalized shipping or soybean-specific crop damage can defeat the relative thesis. The 2:1 contract count is not equal beta; ratio-stop crossing does not define dollar loss without leg-price scenarios. March legs avoid a horizon roll. Expected-value sensitivity fell to ~1.94%; score stays 2.5.

4. Platinum/gold — stronger ratio, much less remaining value

Both prices declined versus the monitor; gold declined more. Five 50-ounce platinum contracts against one 100-ounce gold contract now have $904,050 gross notional. The ratio rose to 0.418145, but the 65.623-second skew makes the ratio indicative and the quote is not a close.

WPIC's current page still points to the Q1 report and identifies September 9 as the forthcoming Q2 release; its exact time remains unverified. The old 297 koz deficit and 1,747 koz stock forecast are not new bullish evidence. Lower demand and higher recycling remain counterweights. WPIC current page, Q1 report

Require supportive new report, then close above 0.414, then defended 0.410–0.414 retest. Targets 0.445–0.460; conditional stop 0.390. Failure of the report plus failed breakout should remove the thesis, not defer it another quarter. Score rises to 4; old-model sensitivity falls to ~0.11%. China inflation tonight and WPIC make urgency high, but no entry justified. Gold can outperform in growth/liquidity stress; near-dollar matching is not volatility neutrality. Plan a mid-December platinum roll before the carried December 31 first-notice date, with exact dates and actual April carry reverified.

5. SOFR — steeper curve is not confirmation

The March implied rate is 4.210%, December 4.065%; 100×(95.935−95.790)=14.5 bp. The slope widened 1 bp, contrary to the required flattening confirmation. It gives a hypothetical flattener more room, but no improved policy evidence. Keep the old event setup cancelled; entry band and numerical target remain incomplete. Conditional invalidation is 16 bp. China, PPI and CPI are the next tests.

Important contract correction: SR3Z26 names the quarter beginning December 16, 2026, with last-trade date March 16, 2027; SR3H27 begins March 17 and last trades June 15. Neither expires by January 6, although the December leg's accrual changes during the horizon. The verified current rule implies a 0.005 outright execution tick for both at this date, not the generic feed's 0.0025 lowLimit. Each leg has $25/bp sensitivity; quarterly averaging and repo basis prevent a pure meeting-probability interpretation. Score 2; no numerical return refresh. CME SOFR mechanics

6. Natural gas — lower premium, higher adverse tail

The January premium is $0.140 below the $1.10 gate. Latest U.S. storage is 3,214 Bcf, 5.2% above the five-year average, but this is August 28 data released September 3. National surplus does not establish regional or January deliverability. Fresh German winter projections point to LNG competition and conditional cold-weather risk; the ministry's counterview expects imports and storage to cover demand. Neither is proof of a realized U.S. shortage. EIA storage, INES scenario reporting, German ministry counterview

Wait; conditional stop $1.35, no validated numerical target. A defensible target and adverse leg-level scenario remain unavailable. Score 3→3.5; prospective value worsens as the entry premium compresses while winter risk rises. STEO and Thursday storage can change the assessment. Review December 18–23 exit/roll before January gas's December 29 last-trade date; carry is not free profit.

7. ULSD/RBOB — improved price, inadequate compensation

The retreat to $1.3613 increases distance to the $1.45 target to $0.0887, but distance to conditional $1.03 stop is $0.3313. The resulting ~0.268:1 distance ratio is better than ~0.034:1 yesterday, not attractive enough to override the $1.12–$1.18 entry band. It is not a target-hit probability or expected return.

Fresh industry reporting emphasizes winter diesel constraints, while demand destruction is the counterargument. EIA's old August 28 distillate stock reading remained 14% below average despite 98% refinery utilization; four-week supplied volumes were down 6% year-on-year. These are dated inventories, not new Tuesday releases. Diesel industry reporting, EIA petroleum summary

Score falls 4.5→4; entry value improves but no activation. Observed session volumes were 65/61; they do not establish executable package depth. A gasoline-specific rebound, refinery recovery or shipping relief can hurt the spread. Review before Christmas; both January legs last trade December 31 and cannot span the horizon unchanged.

8. VIX — small calendar move, no demonstrated edge

Official September 8 October/December settlements rose by 0.2636/0.2762 versus September 4, leaving the spread only 0.0126 lower at −0.8492. It remains outside −1.20 to −1.05. Broad volatility repricing can raise both legs without rewarding this calendar. Score 3.5, expected return still unquantified. Cboe settlements

A specific event-relative edge and complete target remain unestablished; conditional stop −2.00. Date-only marks cannot execute. The exchange specifies a 0.05 outright tick and a 0.01 native spread tick. October 21 and December 16 expiries require exit/roll before January 6; carry and leg slippage can dominate a small edge. Cboe specifications

9. OAT/Bund — useful data recovery, not an executable hedge

The official exact-December settlements improve source quality. Their differences from the prior vendor closes (+0.23/+0.26) are changes between unlike benchmarks, not official daily returns. No OAT/Bund ratio is calculated.

Exact September 8 volume/open interest was 217,488/783,792 for OAT and 1,186,196/1,947,237 for Bund. This demonstrates activity, not hedge validity. CTD, conversion factors, futures-adjusted DV01, repo/basis and integer counts remain unknown; equal EUR 100,000 face values are not equal duration. Target, stop and expected return remain unestimable, score null. Eurex OAT statistics, Bund statistics

Legacy September last trading has passed with no paper exposure. December last trade is December 8 at 12:30 CET; December 4 is an earlier proposed exit date. ECB guidance could matter, but cannot substitute for CTD/DV01 verification, fresh execution FX and current margins. Eurex specifications

Retired ideas — all four reviewed, none reinstated

• Crude normalization, CLV26−CLZ26: $94.50/$88.06, indicative $6.44; +$0.15 versus September 7 and +$0.42 versus September 6. It is $0.19 above the old $6.25 invalidation, not an executed stop. Fresh Gulf transport/refining attacks and low reported vessel crossings do not establish normalization. Alternative routes and demand destruction remain counterarguments, but a renewed short has worse evidence. October last trade September 22 also prevents an unchanged four-month expression. CME crude, September 8 escalation reporting, Shipping report

• Dow/Nasdaq rotation, YMU26/NQU26: 52,795/29,526.75; indicative ratio 1.788040, −0.55% versus the monitor and about −1.08% versus September 6's 1.8075. Relative confirmation remains weaker; no automatic reversal. September contracts last trade September 18. Ratio movement is not a sized dollar return. CME Dow, CME Nasdaq

• Cocoa long, CCZ26: September 8 vendor settlement $5,913/tonne versus the $6,188 September 4 benchmark carried into the prior monitor: −$275, or −4.44%. Reported last was $5,927 at 13:29 EDT. Price decline alone does not reinstate the long; rain-related counterrisk persists and no fresh primary quantified inventory/supply release was verified. Exact vendor page

• CAD/EUR, prior 6CZ26/6EZ26: no fresh exact-contract pair recovered. Retain the September 6 quoted 0.72585/1.16545 as historical context only; no new move, hedge or return is asserted. Keep retired; no replacement or reverse signal. Historical source endpoints: CAD, EUR

Returns: preserve the old forecast date; separate mark sensitivity

The September 6 tree remains 40% persistent friction, 20% transport normalization, 25% renewed supply shock, 10% growth break and 5% AI/liquidity disappointment. These are subjective scenario weights, not measured win probabilities or allocations. No daily re-estimation is claimed.

These are remaining-value sensitivities, not refreshed forecasts, historical performance, paper P&L, return on margin or measured target-hit probabilities. The original forecast dates remain September 6. Old roll assumptions are deliberately held constant; actual roll curves may differ. Commissions, slippage and path-dependent stops are excluded, so net strategy expectancy is unavailable.

On the original modeled set, sugar had highest raw upside (13.39% relief scenario), and sugar/yen effectively tied for best weighted return given uncertainty. Sugar's preferred 19.30 reference, 17.50 target and 20.15 stop give about 2.1:1 conditional distance asymmetry; no fill at that reference occurred. Today's price-only ranking cannot overrule physical, catalyst or execution gates. Sugar short and platinum/gold both lose in the supply-shock scenario; yen can also initially suffer from an energy-import shock. No empirical correlation matrix was estimated from asynchronous snapshots.

Funded paper portfolio — first prospective review

The simulated portfolio began September 8 at 18:45:13.597 UTC. Its inception and capital are not backdated or reset. There were no open positions or pending instructions, and no fills occurred. No historic “Hold” was converted into holdings. Research conditions are not executed paper trades.

No new simulated trade is proposed: every price/physical/calendar gate above remains binding; SOFR, gas and VIX lack complete terms, bonds lack a valid hedge, and current public initial-margin estimates were not verified for any prospective entry. Known multipliers/ticks are not complete audited entry specifications. Some notice/safety clocks, fresh FX for EUR settlement and exact executable-package liquidity remain missing. No margin is invented or substituted with broker data.

At $1m NAV, the portfolio remains entirely in cash with no deployed risk or margin. The 5:1 metals package must not be presumed low risk just because initial dollars are close.

Commission assumption is $3.50 per contract per side; adverse slippage one tradable outright tick per side, embedded in future fills, not double-charged. No costs are incurred without fills. Cash and benchmark earn 0%; no annualized return, Sharpe or hit rate is inferred. Stops are checked only at intermittent observations, not standing exchange orders, and gaps can exceed budgets.

Next 72 hours and urgency

All displayed times are Toronto/EDT. Passed event times do not prove outcomes; Japan GDP, inspections and Crop Progress were actually read. All forward events below remain unreviewed as outcomes.

Priority: (1) wait for tonight's China release; (2) verify WPIC before interpreting any platinum close/retest; (3) reassess the energy outlook and storage/product releases without chasing; (4) use Friday inflation, export sales and WASDE for the yen/SOFR/wheat gates; (5) preserve all mechanical roll checks. The horizon has not rolled forward one day.

Quote and provenance appendix

New CME numbers are ten-minute-delayed exact-contract last trades, not inferred September 8 settlements. September 7 had no CME-group settlements, so prior holiday marks were not false zero changes. The feed's Globex tradeDate may read September 9 while observation time is still September 8 in Toronto. Grain apostrophe quotes are converted from eighths of a cent: 756'4 = 756.50 and 1336'2 = 1,336.25. CME holiday notice

The CME rows retain their quote and retrieval timestamps. Date-only source observations stay date-only; retrieval time, feed refresh metadata and Eurex's date selector are not substituted for trade or settlement times. The supplemental USDA March settlements are date-only and are not used as executable quotes.

Derived ordered-leg definitions: 2W/S = 2×wheat/soy; PL/G = platinum/gold; SOFR slope = 100×(December price−March price); gas = January−March; products = HO−RB; VIX = October−December. Each asynchronous combination is explicitly indicative; matching delayed timestamps alone do not establish depth. No spread highs/lows, stop hits, fills or empirical correlations were reconstructed from leg extremes.

Fresh strict-24-hour energy reporting was distinguished from older EIA inventory vintages and the slightly older Japan GDP release. Reported ship counts are not complete export-volume estimates; executive shortage views and German winter scenarios are not audited realized losses. No real account data, brokerage access or real orders were used.