ATGGLOBAL MARKETS
RESEARCH
Research archive

Gold edge fails the stress test; exit the micro starter

MGCZ26 recovered to 4,399.4 after the North American session, but the remaining six-week edge is too dependent on a favorable drift assumption. Exit the one-contract simulated starter rather than carry a fragile point-estimate edge. Copper remains the strongest challenger, but its 6.8990 close is a chase and three micros now exceed the normal starter band.

Dated research, not live quotes. MGCZ26 is an exact-contract post-reopen IBKR observation at 18:08:34 Toronto, genuinely later than the simulated exit decision. MHGZ26, MESZ26, ZTZ26 and MCLZ26 are exact-contract 16:59:xx pre-maintenance observations; ZCZ26 is its 14:19:59 grain close. Those closing observations are research marks, not post-reopen executable quotes. Adjacent gold maturities have contemporaneous indicative bid/ask but sparse or stale last trades. Research views and scenario estimates are separate from simulated positions and performance.

Gold edge fails the stress test; micro starter exited

Research cutoff: September 22, 2026, 6:08:40 p.m. Toronto. The gold execution observation is a post-reopen exact-contract trade. Copper, equity-index, rates and crude observations are exact-contract pre-maintenance closes; corn is its afternoon grain close. This is a simulated futures portfolio, not a real brokerage account or personalized investment advice.

Decision first

Exit the one-contract MGCZ26 paper position; add zero and do not reverse. December Micro Gold traded at 4,399.4 at 6:08:34 p.m. Toronto, genuinely after the simulated exit decision. The deterministic paper engine applied one adverse tick and closed the position at 4,399.3.

The lifecycle result is a $77 gross loss and $84 net loss after $7 total commissions. Modeled slippage is $2 and is embedded in the fills rather than deducted again. The paper portfolio is now flat: cash and NAV $999,916; realized net P&L −$84; no positions, margin reserve, open risk or pending orders.

This is a conviction exit, not a stop. Structural ETF and official demand remain supportive, but the tactical edge through October 30 is too sensitive to a favorable drift assumption. One contract cannot be trimmed fractionally.

Why the gold trade no longer earns its risk

The exact post-reopen MGCZ26 observation was 4,399.4, with a 4,399.5/4,399.8 displayed bid/ask, new-session volume 1,242 and open interest 75,879. Bid and ask did not carry independent exchange timestamps. COMEX Micro Gold is 10 troy ounces and a $0.10/oz tick is $1 per contract. CME Micro Gold rules

The curve still shows carry rather than scarcity. Contemporaneous indicative bid/ask midpoints were about 4,362.95 for October, 4,396.60 for December and 4,433.90 for February. That puts February roughly $37.30/oz above December. The adjacent contracts were thin and some last trades were stale, so these are indicative curve observations, not executable calendar-spread quotes.

Long-term demand has not disappeared:

• World Gold Council data show August ETF inflows of $18 billion and holdings up 121 tonnes to 4,189 tonnes; its September 21 monitor reported positive weekly buying across regions. These are lagged flows, not proof of buying today. WGC August ETF report, WGC September 21 monitor

• China reported 20.2 tonnes of official purchases in August, but Shanghai Gold Exchange withdrawals were only 62 tonnes, down 22% month over month and 27% year over year. India’s domestic discount had widened to $78/oz by September 11. Official and physical channels are sending different signals. WGC China update, WGC India update

• CFTC September 15 data show standard-gold managed money net long 133,116 contracts, down only 1,856 on the week. This remains a crowded long market, not a washed-out base. The report is all-expiry positioning, not MGCZ26 positioning. CFTC

Policy became less forgiving during the day. Richmond Fed President Barkin said inflation risks exceed employment risks and noted that more than 60% of PCE categories were rising above 3% year over year. Boston Fed President Collins backed the prior week’s 25 bp hike, while New York Fed remarks on ample reserves were implementation updates rather than dovish guidance. Barkin, Williams, Jefferson, Perli

Priced-in: 3.5/5. Investment and official-demand support is well recognized; restrictive policy, contango and weak Asian physical demand are only partly discounted.

Stop-aware re-underwriting

About 38 calendar days remained to the October 30 tactical exit. From the 4,396.4 decision mark, the 4,265 stop was 2.99% lower and the 4,680 target 6.45% higher. Reward/risk looked attractive, but barrier geometry made the stop far more likely to be reached first.

The scenario probabilities below are judgmental and model-assisted, not market-implied. A diagnostic used 21.956% observed annual volatility, hourly monitoring and an explicitly favorable 10% annual price drift. It omits jumps and stop gaps.

PathWeightExitIncremental net P&L from 4,396.4, less $8
Bull: ETF/official demand overwhelms yields32%4,680 target+$2,828
Base: neither barrier hits8%4,475 on October 30+$778
Bear: hawkish repricing or liquidation60%4,265 stop−$1,322
Weighted100%—+$173.60

That point estimate is only 0.0174% of NAV. Moving just five percentage points from target to stop changes expected value to −$33.90; a $30/oz stop gap also removes the edge. At zero assumed drift the diagnostic is approximately flat before meaningful gap risk. The starter still fit the risk budget, but the quality of edge no longer did.

Every active watch and challenger

No fresh candidate qualified for an entry. The best remaining idea is conditional Micro Copper, but it is not a standing order.

Exact contractSeptember 22 observationFour-month decisionFeasible implementation / decisive trigger
MHGZ26 Micro Copper6.8990, 4:59:53 p.m.Wait; do not chase. Physical tightness and momentum remain supportive, but current payoff is near 1:1.Three micros only after a new 6.78–6.80 test and reclaim; 6.60 stop, 7.20 target. At 6.80 buffered risk is about $1,894.50; at 6.899 it rises to $2,822.63, above the normal 0.25% starter band.
MESZ26 Micro S&P7,827.75, 4:59:59 p.m.Wait. The 7,810–7,820 band was not touched after the 9:31 cutoff; post-cutoff low was 7,822.50.Three micros only after a new pullback/reclaim; 7,710 stop and 8,040 target. Reassess after September 23 PMI.
ZTZ26 2-year note101.8828125, 4:59:57 p.m.Avoid / retire.The session high was 101.9609375, so the required 102 touch and rejection never occurred. The finite September 22 setup expired.
ZCZ26 corn536.25¢, 2:19:59 p.m.Avoid / demoted.The session low was 535.50¢, never reaching the finite 525¢ gate. Re-underwrite only after September 24 export sales or genuinely new crop evidence.
MCLZ26 Micro WTI$87.36, 4:59:59 p.m.Avoid.The old long failed. Do not chase a short into backwardation and headline gaps ahead of September 23 EIA.
MESH27 Micro S&PMorning 7,922.25, only 803 displayed volumeAvoid / retired.The March expression remains too thin; use MESZ26 if a new setup qualifies.

Exact closing-price references: Micro Copper, Micro S&P, 2-year note, corn, Micro WTI. CME confirms Micro Copper’s 2,500-pound multiplier and $1.25 tick, and Micro S&P’s $5 multiplier. CME Micro Copper, CME Micro S&P

The cross-asset scan also reviewed MNQZ26, MNGX26 and M6EZ26. Nasdaq and natural gas showed the strongest momentum, but both were extended and lacked a superior verified entry. November gas is preferable to near-expiry October, yet Thursday’s storage report is the proper reset. The euro micro remained inside its session range ahead of PMIs. None justified staging an order.

What changed after the morning cutoff

• Micro Copper rose 0.715% from the morning mark, but its post-cutoff low was 6.8230; the 6.78–6.80 trigger did not occur. Daily lows must not be used to backfill a trade.

• Micro S&P slipped 0.163% from the morning mark. Its full-session 7,810.50 low preceded the morning research cutoff; the later pullback stopped at 7,822.50.

• The $69 billion two-year auction cleared at 4.787% with a reported 2.63 bid/cover and only a 0.2 bp tail to the when-issued yield—orderly absorption rather than the failed-auction signal required by the short thesis. This result is from secondary reporting because a readable Treasury result was unavailable at cutoff. WSJ

• The New York Fed said reserve-management bill purchases have been zero since mid-August because reserves remain ample. Roughly $400 billion of recent net bill issuance produced only modest repo pressure. This is not a QE signal. New York Fed

Next 72 hours

• September 23, 9:45 a.m. Toronto: U.S. flash PMIs; 10:05 a.m.: Fed Governor Barr on the outlook and housing; 10:30 a.m.: EIA petroleum report.

• September 24, 3:30 a.m.: SNB monetary-policy assessment; 8:30 a.m.: U.S. claims, Canadian releases and USDA export sales; 10:00 a.m.: U.S. new-home sales; 10:30 a.m.: EIA natural-gas storage.

• September 25, 8:30 a.m.: U.S. durable-goods orders; 10:00 a.m.: final Michigan sentiment and inflation expectations; 3:30 p.m.: CFTC positioning for September 22.

Primary schedules: Federal Reserve, EIA petroleum, EIA gas, SNB, Census, Michigan, CFTC.

Calendar correction: Q2 GDP’s third estimate and August PCE are September 30 at 8:30 a.m., outside this 72-hour window. BEA

The research horizon remains January 13, 2027. The paper book is flat; the next entry must earn its risk from a fresh exact-contract observation and a new prospective plan.

ATG / GLOBAL MARKETS

Follow the evidence.

New research, useful explanations, and the decisions that follow.

Follow on Instagram Subscribe by RSS