XAUUSD Session Review: September 25, 2026, The Bullish Breakout Failed
Gold rejected the preparation's bearish Trend lead, confirmed a bullish repair after durable-goods data, then erased that break in a two-sided rotation. Neutral / Wait was the right posture, but the next preparation must give event-driven trap risk more weight and demand a defended retest before accepting a repair.
Every prep call is graded like this. See the full track record →
Graded against live session data from Cortiq
Every call in this review is checked against the session data captured in Cortiq — the AI trading workspace that powers our daily preparation, review, and calibration loop.
- Session
- GOLD Durable Goods Breakout Trap
- Symbol
- XAUUSD
- Window
- 00:00 to 23:00 UTC
- Day type called
- Trend
- Day type actual
- Trap
- Lean outcome
- Correct, Neutral / Wait
- Regime
- News-driven breakout failure with two-sided rotation
- Preparation
- Partially accurate
- Surprises
- Moderate
Grade card
9 of 15 correct- Day-type callIncorrect
- Called
- Trend, backed by bearish H4 structure and aligned yield and dollar pressure
- Actual
- Price broke higher after the data, reversed through the range, then recovered. No direction held.
- LeanCorrect
- Called
- Neutral / Wait because no branch reached 55%
- Actual
- The session produced two failed directional moves and closed inside the main 4,244.28 to 4,303.11 decision range.
- Conditional leanPartial
- Called
- Short after confirmed acceptance below 4,244.28, or long after a confirmed reclaim of 4,291.40
- Actual
- The short trigger never fired. The long trigger confirmed, reached 4,315.55, then failed back below 4,291.40.
- Lead scenarioIncorrect
- Called
- Bearish continuation through the twenty-day floor, 37%
- Actual
- Price never reached 4,244.28 and closed 15.77 dollars above the open.
- Scenario that firedCorrect · ly mapped, underweighted
- Called
- Two-sided balance around the data, 30%
- Actual
- Price broke 4,303.11, returned inside within one H1 close, printed a second close inside, then rotated through 4,273.80.
- Scenario mapPartial
- Called
- Bearish continuation 37%, bullish repair 33%, two-sided balance 30%
- Actual
- The lowest-weighted branch described the realized path. The map covered the outcome, but the operative call did not.
- Driver stackIncorrect
- Called
- Bearish structure, rising real yields, and a firm dollar favored continuation
- Actual
- Durable Goods Orders printed 0.0% against a -3.4% forecast, yet gold first broke higher and later recovered from the selloff. The standing bearish read did not control the close.
- 4,342.52 resistanceCorrect · not tested
- Called
- Upper repair target and structural repair threshold
- Actual
- Session high was 4,315.55. The level was not tested.
- 4,334.35 resistanceCorrect · not tested
- Called
- Secondary repair target after a held reclaim
- Actual
- Session high was 4,315.55. The level was not tested.
- 4,322.66 resistanceCorrect · not tested
- Called
- First wider repair target
- Actual
- The rally stopped 7.11 dollars below the level.
- 4,303.11 resistanceCorrect
- Called
- Immediate bearish invalidation edge, with a held break shifting the day toward repair
- Actual
- Price closed above it once, reached 4,315.55, then returned below it on the next H1 close. It marked the failed breakout cleanly.
- 4,291.40 decision edgePartial
- Called
- A confirmed reclaim would activate bullish repair
- Actual
- H1 closes at 4,294.87 and 4,310.79 confirmed the reclaim, but price later lost the level and closed at 4,288.49.
- 4,273.80 pivotCorrect
- Called
- Repeated crossings would favor two-sided balance
- Actual
- Price crossed the area several times and used it as the center of the late rotation.
- 4,244.28 supportCorrect · not tested
- Called
- Bearish continuation trigger after displaced, sustained acceptance below
- Actual
- The session low was 4,254.51. The trigger never fired.
- 4,235.17 supportCorrect · not tested
- Called
- Final confirmed range floor
- Actual
- Price stayed 19.34 dollars above it.
The tape
- Open, first 60 minutes
Gold opened at 4,272.72 and stayed close to the 4,273.80 pivot. Asia then dipped to 4,264.17 before rallying through 4,291.40, but the move did not hold. By 06:00 UTC price had returned below the pivot, leaving the session balanced rather than directional.
- Mid-session
London extended the decline to 4,256.20, still above the 4,244.28 bearish trigger. Gold then recovered steadily into the 12:30 UTC Durable Goods Orders release. The report printed 0.0% against a -3.4% forecast. Price closed the 12:00 hour at 4,294.87, closed the next hour at 4,310.79, and reached 4,315.55 during the 14:00 hour. That sequence confirmed the prep's bullish repair trigger, cleared 4,303.11, and then failed before reaching 4,322.66.
- Late and close
The failed breakout accelerated after 15:00 UTC. Gold fell through 4,291.40 and 4,273.80, then set the day's low at 4,254.51 during the 17:00 hour. Sellers still could not reach 4,244.28. A late recovery carried price back to 4,294.80 before the final close at 4,288.49, above the session open but below the main repair edge.
What we learned
The moderate surprise was not that a two-sided outcome existed in the map. It was the sequence: a large positive data surprise coincided with a confirmed bullish repair, that repair failed across the full intraday range, and the later selloff also failed to secure acceptance below support. The prep could have anticipated this more directly by making the event-day trap branch co-equal with either directional branch.
- Precondition checks
A Trend call needs more than a bearish H4 swing sequence when a high-impact release sits inside the primary breakout window. For the next event session, require post-release acceptance outside 4,244.28 to 4,303.11 before calling the day directional; otherwise give trap or whipsaw at least equal weight.
- Driver-stack ordering
The observed gold response to the release must outrank the assumed yield and dollar reaction. When stronger US data fails to produce sustained gold selling, mark that refusal as live evidence and stop carrying the pre-release bearish stack forward unchanged.
- Tradability standard
Two H1 closes above 4,291.40 confirmed repair mechanically, but the move never defended a retest after the data impulse. On high-impact days, require both meaningful displacement and a held retest before treating a repair trigger as durable.
- Precondition checks
The 30% two-sided branch described the actual path while the 37% lead missed. When the prior session closes near mid-range and the next session contains a major release, narrow lead margins should be treated as a mixed regime, even when the nominal margin reaches four points. --- Reviewed prep: 2026-09-25-xauusd-session-preparation
