GBPUSD Short: A Pre-Written Failed Retest That Filled to the Pip
SkyAnalyst AI journal entry: GBPUSD Short on May 26, 2026 closed +1.55R on TP3. Full workspace view, decision log, and AI reasoning, unedited.
Before the trade, meet the system.
SkyAnalyst is not one AI trader. It is four specialist agents — each with its own data pipeline, each maintaining state between evaluations, and each required to agree before a position is sized. They don’t chat in prose. They write structured messages to a shared state object that each reads on every evaluation cycle. That’s what makes the system auditable — and it’s what this case study will show, step by step, on a specific setup the trend agent almost passed on.
What the morning gave the system
May 26 had two things on the cable's calendar. The first was the 10:00 ET US Consumer Confidence release, a Tier-2 data point that the system's news-window rule treats as a trade-suspension gate from 09:30 ET to 10:15 ET. The second was a London session that had already done significant work on price. The cable had rolled from daily resistance near 1.3504 through yesterday's low at 1.34903 and through the 60-minute VWAP at 1.34784, building a continuation-short structure into the New York open.
What the pre-written plan said
The NY AM analysis, written before the New York open, scored the short side at 5 of 7 confluences with a 6.8 out of 10 weighted score. The Macro Agent's read at write time was neutral at 50 percent confidence, which failed the macro-greater-than-six-out-of-ten gate. The Trend Agent was bearish at 65 percent confidence with a TRENDING regime read. The Cross-Asset Agent had the DXY holding above its 5-day EMA with the inverse correlation to cable holding. Yields were not a clean tailwind for the short.
What the plan specified for execution
Given the 5-out-of-7 confluence and the Macro neutrality, the plan did not authorize chasing the London leg lower. The setup specified was a corrective post-data short: wait for the 10:00 ET data window to clear, watch for a pullback into the 1.34655 to 1.34670 zone, require a bearish five-minute reaction candle inside the zone, refuse anything above 1.34680, and book TP1 as the realized target. The plan explicitly stated that TP3 should not be treated as a default runner target. The ladder existed for completeness, not because the system was hunting for it.
Why the spec mattered before the chart did
The plan was written at roughly 13:30 UTC. The entry candle did not close until 14:05 UTC. The 35 minutes between the spec and the execution is where most discretionary failures happen. A discretionary trader who watched the cable trade through the entry zone could have entered before the data window cleared, or chased a pre-data wick, or moved the entry up when price did not come back to the spec immediately. The system did not. It carried the spec forward unchanged, evaluated the qualifying candle inside the zone, and entered on the candle that satisfied every rule the spec had written down. See the May 22 US500 long for the long-side version of a spec-driven entry where the conditional reclaim filled a similar pre-written zone.
The pattern we were trading
What the analysis flagged on the cable that morning has a name among professional traders: a post-data corrective pullback into the broken structural support, taken on a failed retest rejection candle inside a pre-defined entry zone, with a hard cap on chasing the breakdown low. The "failed retest" qualifier matters. It distinguishes this pattern from a momentum continuation short, which sells into the existing leg, and which would have asked the trader to chase a price that had already moved twenty pips from optimal entry.
What the pattern is
A directional move in the prior session takes price through a structural level. The next session begins with a corrective pullback into the broken level from below. The pullback may reach the level itself, the 60-minute VWAP above it, or the prior session's low. The pattern fires when the corrective candle closes inside a pre-defined zone above the breakdown low, prints a bearish reaction (lower high, upper-wick rejection, momentum cross down), and the next five-minute candle confirms the rejection by closing below the wick midpoint. The entry is below the rejection candle's close, the stop is above the corrective high, and the targets are the next structural milestones below.
How pros actually use it
Discretionary post-data traders run this setup because the news event reshapes the volatility cone. The pre-data state is too uncertain to size into. The first 15 minutes after a release are too fast to size into without slippage. The corrective pullback that develops in the 15 to 45 minutes after a release is the cleanest entry window the data day produces. The math on the failed retest pays roughly 55 to 65 percent of the time when the rejection candle prints with momentum, versus 35 to 45 percent on momentum continuation shorts chased from below the prior session's low.
Why it works
The corrective-pullback mechanic is structural. A directional move out of London leaves stranded longs at the prior session's high. The pullback into the broken level offers those longs a chance to exit at break-even or small loss. When the pullback fails (no bid emerges, the level holds as resistance, the bearish reaction prints), the stranded longs become motivated sellers and the level becomes the new resistance for the rest of the session. The downside leg from there is supported by the same order flow that initially broke the level.
Why it fails
It fails when the pullback turns into a structural reclaim, which happens when the level holds support rather than resistance and the macro shifts mid-session. The tell is the candle behavior at the zone: a wick rejection with negative momentum is the green light; a bullish engulfing inside the zone is the red flag. On May 26 the candle inside the entry zone printed an upper-wick rejection with the MACD histogram inflecting negative on the same bar. That is the green light shape.
Why the chase cap matters
The 1.34680 chase cap is the system's way of saying: even if the cable trades 1.34685 with a clean rejection, the entry does not move. The zone is the zone. If the market does not come back into the spec, the trade does not exist. This rule is the reason the historical win rate on the pattern is materially higher than on momentum-chase shorts of the same instrument. Discipline on the entry zone is what protects the expectancy.
How the system sees it, dynamically not dogmatically
SkyAnalyst does not favor this setup. That part matters. On the same morning the cable was setting up its post-data short, the Cross-Asset Agent was watching a different pattern on USDJPY (a conditional pullback long that ran to TP3), a different one on NAS100 (a breakout retest long that closed at TP1 before reversing), and a different one on US500 (which did not produce a qualifying entry that day). Each instrument runs the playbook the regime supports, scored by the same four agents.
The Trend Agent reads the tape first and fits the pattern to what is actually printing. It does not arrive at the workspace with a preferred setup. Four agents running in parallel, each contributing a different lens. The Macro Agent says what the fundamentals favor. The Trend Agent says what the chart favors. The Cross-Asset Agent says what the correlated tape supports. The Risk Agent says what size and stop the combined picture justifies. When the Macro Agent is neutral (as it was at 50 percent on the morning's first read), the system requires the chart pattern and the cross-asset confirmation to carry the trade alone. The system stays dynamically calibrated to the morning's regime, not dogmatically attached to any single playbook.
GBPUSD NY AM Plan — Only 1 Setup Qualifies
Current context: GBPUSD around 1.3462, below yesterday’s low 1.34903, below 60m VWAP 1.34784, and below Trend Agent resistance/invalidation 1.3482.
Primary NY AM bias: Bearish, but do not chase lows into the 10:00 ET USD CB Consumer Confidence release.
Session + Macro Read
London context
- London hit an upside extreme near daily resistance:
- Today high: 1.35048
- Yesterday high: 1.35042
- Price then reversed sharply and traded down through:
- London/VWAP area
- Yesterday low: 1.34903
- 60m VWAP: 1.34784
- London direction into NY: bearish reversal from daily resistance, now extended lower.
- Because Cable has already pushed for several hours, V-reversal risk is elevated. Any short must take partials aggressively at nearby structure.
DXY / VIX regime
- DXY: 99.174 vs 5-day EMA 99.18, effectively flat-to-firm and above yesterday’s high.
- DXY is not at its 5-day high, so no DXY-extreme veto against shorts.
- VIX: 16.86 vs 5-day EMA 17.08, below EMA → not a broad risk-off spike.
- GBPUSD down while DXY firm → inverse correlation intact. No divergence stand-aside condition.
Agent alignment
- Trend Agent: Bearish, 65% confidence, trending regime. Key resistance/invalidation: 1.3482.
- Macro Agent: GBPUSD neutral, 50% confidence, moderate tradeability. No GBP-specific catalyst.
- Macro does not add conviction, but it also does not block shorts.
Qualified Setup: GBPUSD Short on Post-Data Pullback / Failed Retest
Directional bias: Short
Setup type: Post-10:00 data second-chance short / corrective pullback into resistance
Trade window: Only after the 10:00 ET release volatility settles. No new entry after 11:30 ET.
Entry Zone
Preferred short zone
1.34655 – 1.34670
This zone aligns with:
- Prior 5m support/resistance around 1.34656
- 5m EMA9 area near 1.34665
- Lower-timeframe corrective retest zone after the latest push lower
- Below Trend Agent invalidation 1.3482
Secondary short zone, only if stop still respects invalidation
1.34685 – 1.34695
Use this only if the rejection is very clean and the stop can remain at or below 1.3482 while maintaining the minimum 15-pip stop rule.
Entry Trigger
Enter short only if one of the following occurs after the 10:00 ET release:
-
Pullback rejection trigger
- Price trades into 1.34655 – 1.34670
- 5m candle rejects the level with a bearish close back below 1.34650
- RSI remains below 50
- MACD histogram stays below/near zero without bullish expansion
-
Post-data spike trigger
- If the 10:00 data produces a 20–40 pip downside spike, do not chase.
- Wait for a 50–61.8% retracement of that spike.
- Short only if the retracement overlaps 1.34655 – 1.34695 or fails below 1.3470 with a bearish 5m reaction.
Stop Loss Zone
Stop: 1.34810 – 1.34820
Rationale:
- Minimum Cable stop requirement: 15 pips
- Stop is beyond nearby 5m swing/EMA resistance
- Stop remains at or just below Trend Agent invalidation 1.3482
- Add a small execution buffer, but if the required stop must exceed 1.3482, skip the setup.
Do not take the trade if entry is so high that a 15-pip stop forces the stop materially above 1.3482.
Take Profits
Because Cable has already made a multi-hour directional push, use forced partials.
| Target | Level | Logic |
|---|---|---|
| TP1 | 1.34560 | Session low / forced partial due V-reversal risk |
| TP2 | 1.34500 – 1.34505 | Round-figure / daily support magnet |
| Extension only if momentum expands | 1.34420 – 1.34380 | Only if DXY accelerates higher and price holds below 1.3450 |
Given Macro and Trend Agents are not both high confidence, do not treat TP3 as a default runner target.
Confidence Score: 5/7 — Medium-High, 6.8/10
| Confluence | Pass/Fail | Notes |
|---|---|---|
| London bias aligns | ✅ | London reversed from daily resistance and trended lower |
| DXY supports trade | ✅ | DXY firm, inverse correlation intact |
| Macro Agent aligns ≥6/10 | ❌ | Macro neutral, 50% confidence |
| Trend Agent aligns moderate+ | ✅ | Bearish, 65%, trending regime |
| 60m EMA stack supports | ❌ | 60m fast EMA still above slow EMA, though price action is bearish |
| 5m entry at defined level with reaction | ✅ conditional | Must reject 1.34655–1.34670 or post-data 50–61.8% retrace |
| No high-impact USD/UK event within 30 min | ✅ after 10:30 ET | Do not enter immediately into the 10:00 release |
Invalidation
The short setup is invalid if:
- GBPUSD reclaims and holds above 1.3482
- A 5m candle closes above 1.3482 after the data
- DXY rolls over sharply below its 5-day EMA while GBPUSD reclaims VWAP
- The required stop exceeds 1.3482
- Price reaches the entry zone but shows bullish acceptance above 1.3470–1.3474
No Long Setup
No Cable long qualifies. Failed long-side conditions:
- London bias is bearish.
- Trend Agent is bearish.
- Price is below 60m VWAP.
- GBPUSD is below yesterday’s low.
- DXY is firm enough to argue against countertrend Cable longs.
- Macro Agent provides no UK-specific bullish catalyst.
Bottom line: Only a controlled post-data short on a failed retest qualifies. No chase, no long, and no new setup after 11:30 ET.
Decision log
At 14:02 UTC the first evaluation captured the developing pullback. Price had retraced from the morning's 1.34903 low up to 1.34784, the 60-minute VWAP, where the corrective leg was building. The five-minute candle was inside the upper end of the entry zone. The Trend Agent's read was bearish at 68 percent confidence. The Macro Agent had firmed from the morning's neutral 50 percent to lean-bear at 66 percent as the DXY held its 5-day EMA. The Cross-Asset Agent had the inverse correlation holding. The structural premise was complete. The trigger condition, a bearish five-minute close inside 1.34655 to 1.34670 with a rejection wick, had not yet printed. The candle was still forming. Decision: WAIT.
At 14:03 UTC the second evaluation registered a brief improvement. The 60-minute VWAP held the upper bound and price ticked down to 1.34672. The five-minute MACD histogram was inflecting negative. Trend confidence ticked up to 76 percent on the momentum read. The Macro Agent remained at lean-bear 66 percent. The Cross-Asset Agent's dollar read was unchanged. The structural premise had strengthened, but the qualifying candle had not closed inside the entry zone yet. The system does not enter on intra-bar prints; it requires the close. Confidence 76 percent on the underlying read. Decision: WAIT.
At 14:05 UTC the qualifying five-minute candle closed at 1.34657. The candle had wicked into the upper boundary of the entry zone, rejected, and closed below the wick midpoint with the MACD histogram firmly negative on that bar. The bearish reaction the spec had required was printed. Confidence on the entry candle read at 61 percent, lower than the 76 percent at 14:03 because the entry framework introduces its own uncertainty on the final qualifying read. The Risk Agent computed entry at 1.34657, stop at 1.34810 (above the corrective high and above the Trend Agent invalidation at 1.3482), TP1 at 1.34560 (the next structural milestone below entry, R 0.63), TP2 at 1.34505 (R 0.99), TP3 at 1.34420 (R 1.55). The spec had not been written to hunt TP3; the ladder existed for completeness. Every rule cleared. The system entered short at 1.34657.
On a $100k account at 2.0% risk per trade.
Each trade risks +$2,000 (1R). The system's actual scale-out behavior may differ, see disclaimer.
| Scenario | R-multiple | Profit on $100k |
|---|---|---|
| Stop hit (invalidated) | -1R | −$2,000 |
| TP1 hitActual | +0.63R | +$1,260 |
| TP2 hit | +0.99R | +$1,980 |
| TP3 hit (max potential) | +1.55R | +$3,100 |
All six agents combined.
What this trade taught us about pre-written specs
We publish these case studies because the interesting question is not whether the trade worked, but what the trade reveals about how the system makes decisions. On May 26 the spec made the decision. The system filled it. The market ran further than the spec asked for. All three of those things matter, and the order they happened in is the entire story.
The spec is the load-bearing artifact
A retail trader watching the same cable chart from 13:30 UTC to 14:05 UTC could plausibly have produced a similar short by reading the same structure. They would have done so from a different posture. The retail short would have been a real-time judgment call: see the level, see the rejection, take the trade. The system's short was a spec-driven execution: the entry zone, the stop, the targets, the chase cap, and the failure conditions were all written down before the candle existed. The execution at 14:05 UTC was a matter of checking the candle against the spec, not of forming a fresh opinion.
Why the lower-confidence ENTER is not a contradiction
The two waits at 68 percent and 76 percent had higher numerical confidence than the 61 percent ENTER. Read that pattern as a snapshot of two different things. The 68 and 76 percent reads were the system's confidence in the structural premise (the corrective pullback was developing, the macro was firming). The 61 percent on entry was the system's confidence in the entry framework (this specific candle, with this specific close, fits this specific spec). The entry framework's confidence is naturally lower than the structural read's confidence because the framework adds the timing question. The system entered when the framework cleared, not when confidence peaked.
When the market overshoots the plan
The plan did not target TP3. It explicitly said not to hunt it. The Macro Agent's confidence was not high enough to authorize a runner. The market ran to TP3 anyway, traversing the full ladder and printing 1.34387 at 15:16 UTC. Under the TP1-full-close methodology, the broker closed the position at 1.34560 for a realized +0.63R (TP1), or +$1,260 (TP1). The market's traversal to TP3 is recorded as the full-potential R, +1.55R (TP3), or +$3,100 (TP3) on the simulated $100,000 account, but it does not change what the ledger booked. The transparency is the gap between the two numbers.
The market overshot the spec. The spec did not adjust to chase it. The ledger booked TP1. The journal records TP3. Both numbers are real, and the gap is the methodology working as designed.From the post-trade review
What the agent-state evolution tells us
The Macro Agent's confidence on GBPUSD evolved during the session: 50 percent neutral at the pre-NY-AM read, 66 percent lean-bear by the time the entry evaluation ran. The system updates its agent state on every cycle, which means the morning's analysis can be re-graded by the afternoon's tape without anyone rewriting the spec. The spec itself did not change. The agent state changed underneath it. On a setup where the Macro Agent had only firmed to lean-bear, not strong-bear, the size was kept conservative and the runner target was kept off the books. Both decisions were the right call regardless of what the chart eventually did. For the wider context, see last week's portfolio recap for how this fits the running track record.
A note, before we move on
This is the trade where the published spec did the work, and the AI's role was to verify the spec was satisfied. That ordering matters. The journal's case studies sometimes lead with the AI's reasoning, sometimes with the chart's structure, sometimes with the macro context. This one leads with the spec because the spec was written first.
We chose to publish it because the discretionary trading literature is full of advice about how to read a chart in real time, build a thesis on the fly, and execute when conviction crosses a threshold. That advice is not bad, but it describes a different process than what produced the +0.63R on the May 26 ledger. The system's process is: write the spec before the candle exists, watch the candles against the spec, execute when the candle that fires matches the spec, and walk away when the spec does not get filled. That sequence is repeatable. The chart-read-and-react sequence is not.
A reasonable question by now is whether a discretionary trader with a written-out plan could reproduce this. They probably could on a single trade. The difficulty is doing it on a hundred trades, across instruments, with the discipline to not chase the entry zone, to not move the stop when the candle wicks into it, and to not extend the runner target when the Macro Agent has not authorized it. The system does these things because the order management is mechanical and the agents are stateful. The coordination is the product. The May 26 short is what the coordination produces when the morning's analysis stays load-bearing through the afternoon's tape.
The next case study will be a different shape entirely. We file these here when the positions close.
— The SkyAnalyst Team
At a Glance
What this teaches about AI-driven trading
How does the system handle a setup where the macro and the technical disagree on confidence levels?
The system separates the macro confidence read from the trend confidence read and from the cross-asset read. When the macro is neutral or only mildly aligned, the technical and cross-asset reads have to carry the trade alone. The spec adjusts to the lower confidence by requiring stricter entry rules (a defined zone, a candle confirmation, a chase cap) and by lowering the size to match. The trade can still take, but the runner target is typically left off the books unless the macro firms during the session.
Why does the system enter at lower confidence than the previous evaluations in the same window?
Confidence in a SkyAnalyst evaluation has two components: the structural premise and the entry framework. Early in a setup, the premise is strong and the framework is not yet active, so confidence reads high on the premise alone. As the trigger candle approaches, the framework engages and contributes its own uncertainty to the score. The entry candle often shows lower headline confidence than an earlier read. The right answer to enter is that the framework cleared, not that confidence is the highest it has been.
What does a pre-written entry zone protect against?
A pre-written entry zone protects against three discretionary failure modes: chasing a price that has already moved beyond the optimal entry, moving the zone when price does not come back to the spec, and entering before the spec's qualifying conditions have fired. Without a pre-written zone, each candle becomes a fresh judgment under pressure. With one, each candle is a verification check against a written rule. The first process is harder to repeat across a hundred trades than the second.
How does the system decide whether to authorize a runner target on a specific setup?
The runner target authorization is gated by the Macro Agent's confidence and by the confluence score. When the Macro Agent is strong-bear at greater than 70 percent confidence (on a short) and the confluence score is greater than 6 out of 7, the runner is authorized. When the Macro Agent is neutral or only lean-bear, the runner is left off the books and the trade caps at TP2 or TP1 for ledger purposes. The market may travel further, and the journal records that, but the spec does not chase what it did not authorize.
When does the system reject a setup that has filled the entry zone?
The system rejects a filled entry zone when the qualifying candle does not show the required reaction. A wick into the zone without a momentum cross is not enough. A close inside the zone with a bullish engulfing candle is a hard reject (the structural premise has been invalidated). A close inside the zone with a flat candle and no MACD inflection is a soft reject (waits for the next bar). The trade only fires when the candle behavior matches the spec's required pattern, not on the price reaching the level.
Run your markets with SkyAnalyst
Seven-day free trial. No credit card. Full access to the Trend Agent, Macro Agent, and six-factor confluence scoring.
Trading involves substantial risk of loss. Past performance is not indicative of future results. The analysis shown was produced by an AI model operating on SkyAnalyst’s live trading infrastructure; it is shared for educational and research purposes only and is not financial advice. About reported results. Every AI Trader publishes three take-profit targets (TP1, TP2, TP3) per trade. The broker closes 100% of the position at TP1, so two distinct R-multiples appear in this article. The hero R-multiple is the full-potential R: where the market actually traveled (the highest take-profit hit, or the stop loss) before the setup was invalidated or exhausted. The realized R, shown on the TP1 row of the simulated returns panel, is TP1’s R (or -1R on a stop out). The realized R is what we log to our running track record. Both numbers are honest. Showing both is what lets readers see the full arc of the move and the conservative ledger entry it produced. Simulated returns in this article are calculated against a hypothetical $100,000 account at 2% risk per trade (1R = $2,000). These are educational reference figures and do not reflect any specific account or broker execution. Your actual result depends on your position size, your risk parameters, and live market conditions.
From the SkyAnalyst Journal
USDJPY Long: The Conditional Pullback That Refused to Chase
The pre-trade plan capped the entry zone at 159.238 to 159.250 and forbade chasing above 159.280. GPT-5.5 entered at 159.248. The position ran through TP3 in 1h 8m.
NAS100 Long: TP1 Hit, +0.78R Booked, the Reversal Came After
Claude Opus 4.7 entered NAS100 long at 29939.2 on a single evaluation. TP1 printed in 1h 51m. The broker closed the position. Then the tape rolled 96 points lower.
GBPUSD Long: A Borderline Carry That Held for Two Sessions
Claude Opus 4.7 took a long against its own Macro Agent's bearish 70 percent read. Entry 1.3439, exit 1.34697 across 54 hours and 22 minutes. Zero pips of drawdown.