Apr 20-26, 2026: Two stops, two index trades, 2R given back
A 33% win rate week where both decisive losses sat on the same playbook shape. The system gave back 2R against +10.67R YTD, and the loss attribution lands entir
Restated: Gold (XAUUSD) was part of SkyAnalyst's coverage from inception (Jan 12, 2026) through May 2026. We've since narrowed coverage to six instruments — EURUSD, GBPUSD, USDJPY, US30, NAS100, US500 — and these numbers are restated for the current lineup. The original publish date is preserved; cumulative figures have been recomputed.
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.
Three master-automation trades fired in this window. Two of them stopped at the planned risk distance, both on index instruments under Claude Opus 4.6, and both inside the same pullback-into-structure family of setups. The math is straightforward: a 33% win rate week, 2R given back, a 3.87% trough on the equity curve before the cadence reset. We are publishing this because every legitimate trading desk publishes its drawdowns. From Jan 12 inception through Apr 27, 2026 the system has banked +10.67R YTD across 77 trades at a 57.14% win rate, and a $100,000 simulated account at 2% risk per trade now sits at $121,345 (static) heading into the next window. The week's 2R give-back is roughly $4,000 of that figure, a normal drawdown line on a sample of 77, and the kind of week that has to happen for the longer record to mean anything. The loss pattern is the interesting part. Both stops were the same shape on different indices, fired on consecutive sessions, by the same model family. That is the lens this report uses.
Act I, Thursday's NAS100 long
The week opened quietly through Monday and Tuesday with no master-automation entries that cleared the threshold. The first decisive trade fired on Thursday Apr 23 at 15:51 UTC, a NAS100 long built on a conditional pullback into the VWAP and structure confluence zone. Claude Opus 4.6 scored the setup at C+ on grade, sized at 2% account risk, and took the entry. Price pressed against the upper structure band and never built the follow-through the playbook required. Stop hit at -1R, and equity dropped from the day's $103,333 peak to $101,333.
Act II, Friday's US500 short
Less than 24 hours later, on Friday Apr 24 at 14:05 UTC, the system fired in the opposite direction. A US500 short on a VWAP rejection plus opening range breakdown read. Same model, same C+ grade, same 2% risk. The setup looked clean on the entry candle and then the index reclaimed its opening range inside thirty minutes. Stop hit at -1R, and the curve printed its trough at $99,333. That moment is the maximum drawdown line: 3.87% from peak, two stops in two sessions, both on the same playbook lineage.
Act III, the window's only decisive winner
The third trade in the window was the only winner, and it kept the give-back from compounding into a worse number. We isolate the losses here because that is the point of a drawdown report, but the curve recovered partially off the trough on the third trade's resolution. Net for the window: -2R, two losses on index instruments, one winner outside this report's scope.
Every trade the system took.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Apr 23 | 15:51 UTC | NAS100 | Long | Claude Opus 4.6 | Conditional Pullback Long at VWAP/Structure Zone | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Apr 24 | 14:05 UTC | US500 | Short | Claude Opus 4.6 | VWAP Rejection / Opening Range Breakdown Short | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
Dollar figures are simulated on a $100,000 account at 2% risk per trade. Actual subscriber P&L varies with account size. Past performance is not a guarantee of future results.
Pattern of the week
Both losses are the same shape with the sign flipped. A conditional pullback into a VWAP-anchored structure zone on an index, taken by Claude Opus 4.6 at a C+ grade, sized at the system's standard 2% risk. The NAS100 trade was the long version, the US500 trade was the short version, and the two fired on consecutive sessions twenty-two hours apart.
The pattern is worth naming because it tells us where the system's edge thinned out this week. Pullback-into-structure entries on US indices rely on the post-pullback follow-through candle building the level. When that candle prints and then fades inside the next ten bars, the structure was a noise level, not a real one. Both of this week's stops show that shape on the post-entry chart. The Trend Agent scored the setup; the Macro Agent did not veto; the Cross-Asset Agent did not flag a correlated rejection; the Risk Agent sized normally. All four agreed, twice, on two trades that the tape disagreed with.
That alignment is the lesson. The system is not broken when four agents agree and the market still says no. It is doing exactly what a probability-based system does on the losing side of its win rate.
Decision highlights
The first decision worth flagging is the entry on Friday's US500 short less than 24 hours after Thursday's NAS100 stop. The system has no rule that gates a fresh entry on the heels of a recent loss in the same instrument class, and on this week that produced two correlated stops back to back. The behavior is correct under the current ruleset. The question is whether the ruleset should change.
The second is the Risk Agent's sizing on both trades. Both fired at the standard 2% of account, with no haircut for the C+ setup grade or for the fact that both setups sit in the same playbook family. A grade-aware or family-correlated sizing tier is something we have discussed and not yet wired in; this week's losses are the cleanest case yet for revisiting that conversation.
The third is the absence of a Macro Agent veto on either trade. Both setups were taken in macro regimes the agent reads as neutral, which is the agent operating as designed. The honest review here is whether neutral should imply full-size, or whether neutral should imply a sizing haircut on the assumption that edge is thinner outside high-conviction regimes.
Claude vs GPT: who led the week.
SkyAnalyst runs multiple foundation models in parallel across its four-agent system. When two models trade the same instrument in the same week, the results are directly comparable. This is that comparison.
- NAS100-1.0R · 1 trade
- US500-1.0R · 1 trade
Same signals, same risk framework, different foundation model.
Six instruments, six stories.
EURUSD: no losses this window. The pair did not produce a setup that cleared the Trend Agent's threshold during the window, so it sits at zero trades on the loss-side scoreboard.
All EURUSD this week →GBPUSD: no losses this window. Macro conditions through Apr 20-26 kept Sterling outside the conditions where our setups typically trigger, and the pair did not fire a decisive entry.
All GBPUSD this week →US30: no losses this window. The Dow stayed inside the Trend Agent's neutral band on the relevant timeframes, and no entry signals fired during the window.
All US30 this week →NAS100: one loss, -1.0R. The Apr 23 Claude Opus 4.6 long on a conditional pullback at VWAP/structure stopped when the post-entry candle failed to build the level. This is the first half of the week's repeated-shape pattern.
All NAS100 this week →USDJPY: no losses this window. The pair stayed inside the macro regime that the Macro Agent does not gate for our setups, so no decisive entries fired during the window.
All USDJPY this week →US500: one loss, -1.0R. The Apr 24 Claude Opus 4.6 short on a VWAP rejection and opening range breakdown stopped when the index reclaimed its opening range inside thirty minutes. This is the second half of the repeated-shape pattern.
All US500 this week →Loss of the week: US500 Short · -1R
Losses worth learning from
NAS100 long, Apr 23, -1R
What was right: The Trend Agent's score on the 15m structure read was clean. VWAP and the structure zone aligned within a tight band, and the pullback candle into that zone printed the shape the playbook calls for. Cross-Asset confirmed no correlated rejection on US30 or US500 at the entry moment. Risk sized at the standard 2%, stop placed below structure at -1R distance.
What was wrong: The post-entry follow-through never built. The next ten 15m bars failed to print a close above the entry candle's high, which is the playbook's implicit confirmation that the structure was real. By the time price tagged the stop, the structure level had already failed twice on lower timeframes.
What we'd do same: Take the setup. The read was internally consistent across all four agents and matched the historical pattern. The miss is on the market side of the probability, not the system side.
US500 short, Apr 24, -1R
What was right: The VWAP rejection candle on the 5m printed cleanly at the opening range high, and the opening range breakdown read followed within two bars. Trend scored the structure, Cross-Asset confirmed correlated weakness on NAS100, Risk sized normally. The entry trigger fired exactly where the playbook places it.
What was wrong: The opening range was thinner than the system's threshold typically tolerates, which made the breakdown a more fragile read than the grade suggested. The index reclaimed the range inside thirty minutes, which is the failure signature for this exact setup family. Stop hit cleanly.
What we'd do same: Take the setup. The grade and the agent confluence supported entry under the current rules. The question for the QC cycle is whether the opening range thickness check should move from advisory to gating; this trade is a clean case for that tightening.
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 |
|---|---|---|
| Window drawdownActual | -2R | −$4,000 |
All six agents combined.
From the desk
The honest read on this week is that the system did its job. Four agents agreed twice, on two setups that fit the playbook, and the tape disagreed with both of them. That is the cost of running a probability-based system whose edge plays out over hundreds of trades, not three.
A $100,000 simulated account at 2% risk per trade sits at $121,345 (static) or $121,678 (compounded) through Apr 27, 2026, which is the YTD context for the window. The 2R give-back this week is approximately $4,000 of that figure. The gap between static and compounded balance, modest at this sample size, is the visible evidence that disciplined sizing through 77 trades produces a different dollar path than naive flat-stake math would. That gap widens over time, and on a losing week it is the reminder that the rules under the surface are still working.
We will keep publishing these reports on weeks the gate opens. The cadence has organic holes by design, because a drawdown report on a week that did not draw down is just noise. This week drew down, and so this week gets the report.
What we're tuning
Two items go on the next QC cycle. First, a family-correlated sizing haircut for consecutive same-shape entries on related instruments. Two index pullbacks in 22 hours under the same model family is the cleanest case we have seen for that tier, and the implementation cost is low. Second, a gating move on the opening range thickness check for US500 breakdown setups, lifting it from advisory to a hard threshold the Trend Agent must clear before scoring the setup at C+ or better.
Neither change is urgent. The system's longer-record edge is intact, the drawdown is inside the expected band, and a week of two stops is not a mandate to redesign the playbook. The changes go in the queue, get tested on the historical sample, and ship only if the backtest holds.
What the numbers actually mean
A 33% win rate week, on a sample of three trades, is not a signal. It is a slice. The system's longer record sits at 57.14% across 77 trades from Jan 12 inception, which is the number that carries statistical weight. Van Tharp's framework in Trade Your Way to Financial Freedom puts it plainly: a system's edge is the product of win rate and average R, evaluated across a sample large enough for the law of large numbers to do its work. Three trades is not that sample. Seventy-seven is closer.
Schwager's Market Wizards interviews repeat a version of the same point. The traders who survive long enough to be interviewed share one habit. They size for the drawdown they will get, not the drawdown they hope for. A current window drawdown of 3.87% from peak is well inside the range a 2%-risk system with a 57% historical win rate should expect, and the longest losing streak observed this week (two) is below the streak distribution a system of this profile produces routinely. Schwager would call this a normal Tuesday.
The statistics we publish in the scoreboard, current win percentage, current R target, current drawdown percentage, current longest streak, are the numbers a fund's risk committee reviews on its own books every Monday morning. The point of putting them in front of readers is the same point that animates this entire article: drawdown reporting is what every legitimate trading desk does, every week, regardless of whether the week was loud or quiet. We are showing the math because the math is the product.
At a Glance
Drawdown questions
A 33% win rate week sounds bad. Is the system breaking?
No. Three trades is not a statistically meaningful sample, and the longer record sits at 57.14% across 77 trades. A 33% slice on N=3 is well inside the variance any system of this profile will produce, and the math behind that is covered in the statistics section above.
Why did both losses come from the same model?
Claude Opus 4.6 produced both decisive losing trades because it scored both setups above the entry threshold while GPT did not fire a decisive entry this window. That is sample variance on a two-trade slice, not a model verdict. Our longer head-to-head record shows both families winning and losing across instruments.
How does this 2R loss compare to YTD performance?
The system is at +10.67R YTD across 77 trades through Apr 27, 2026. The 2R give-back this week represents roughly 18% of cumulative YTD R, which on a $100,000 simulated account at 2% risk per trade is approximately $4,000 against a static balance of $121,345.
What changes are you making after this week?
Two items go on the next QC cycle: a family-correlated sizing haircut for consecutive same-shape entries on related instruments, and a gating move on the opening range thickness check for US500 breakdown setups. Neither is urgent, both will be backtested on the historical sample before shipping.
Why publish your losses at all?
Because every legitimate trading desk publishes its drawdowns. Hiding losing weeks would make every winning week meaningless. The math behind a 2% risk system with a 57% win rate is the product, and you cannot see the math without seeing the losses.
Trade with the system that publishes its drawdowns.
Subscribers receive every signal — winners and losers — three minutes before entry, with full reasoning.
Dollar figures are simulated on a $100,000 account at 2% risk per trade. Drawdown trajectories shown reflect a small window sample size and are not projections of forward performance. Past performance — including losses — is not a guarantee of future results. Actual subscriber P&L varies with account size and execution. YTD context: +10.77R YTD across 77 trades, see stats strip.
From the SkyAnalyst Journal
EURUSD Short on May 15: Six Waits, One Enter, a Clean TP1 Fill
A risk-off Euro short where the system scored six consecutive waits in the low 40s, then flipped to enter at 62 percent, and the position closed TP1 for +2.00R (TP1) with zero recorded drawdown.
Three stops, one regime: the week the tape refused to confirm
Three losses, 2.25R given back against a year that still reads +20.43R. The honest portfolio view: what every stop taught us, and what the drawdown curve says about a week that drew down 2.4 percent and recovered.
May 11-17, 2026: Seven-Three Week, +5.96R Net Across a Model Upgrade
Ten canonical trades, seven winners, three losers, +5.96R net at the TP1 baseline. Tuesday and Wednesday ran on Claude Opus 4.6, Friday switched to Opus 4.7, and GBPUSD came online as a new instrument and won both its trades.