Four losses, all at -1R, no exits above the stop. Net -4.00R for the loss-side ledger across Wed-Fri. The same five sessions closed +1.21R on the recap. This is
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.
Four losses. A four-trade losing streak running across Wed afternoon to Fri afternoon, the longest of the rolling sample. Net for the loss-counting window: -4.00R, every loss at -1R. The trough sat at $99,358 Friday afternoon, a 7.45 percent intraweek drawdown from the Tuesday peak. This drawdown report is not the recap. The recap covers all of Apr 13-19 and lands at +1.21R net across nine trades, with three TP3 winners carrying the early arc. See the weekly recap for the full ledger and March's monthly recap for closing-month context. This document opens the books on the loss side: each loss, why each cleared threshold, what failed after entry, and the rolling-window statistics that say a four-trade streak on a 35-40 percent system sits below the median.
The week opened with three winners on Mon and a TP3 XAUUSD winner on Tue afternoon. Equity peaked at $107,358. Forty minutes after the XAUUSD entry, EURUSD Long at 15:27 UTC on a pullback to session support stopped at -1R when the shelf failed inside the hour.
Wed Apr 15 was a zero-trade day. Thu Apr 16 brought a NAS100 Short at 14:31 UTC on a VWAP rejection that stopped at -1R as the index reclaimed VWAP through the cash open. Fri Apr 17 delivered two more losses inside fifty minutes. EURUSD Long at 14:29 UTC stopped. XAUUSD Long at 15:19 UTC stopped as a buyer stack above the shelf swept the position. Four consecutive losses, the longest streak of the rolling sample. Equity at the trough of $99,358.
Six minutes after the XAUUSD stop, US30 Long at 15:25 UTC on a bullish pullback to micro-support cleared confluence at the same 55 percent threshold the morning's losers had used. The Risk Agent did not widen the floor. The trade ran to TP1 for +1.53R and pulled the recap net to +1.21R.
Prior drawdown report: Apr 6-12 drawdown report (2 losses, -2R).
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Apr 14 | 15:27 UTC | EURUSD | Long | Claude Opus 4.6 | Buy EURUSD on Pullback to Session Support | C+ | -1.0R | -$2,000 | Stop hit | - |
| Apr 16 | 14:31 UTC | NAS100 | Short | Claude Opus 4.6 | VWAP Rejection Short | C+ | -1.0R | -$2,000 | Stop hit | - |
| Apr 17 | 14:29 UTC | EURUSD | Long | Claude Opus 4.6 | EURUSD Pullback Long | C+ | -1.0R | -$2,000 | Stop hit | - |
| Apr 17 | 15:19 UTC | XAUUSD | Long | Claude Opus 4.6 | Bullish Pullback to NY Session Support | C+ | -1.0R | -$2,000 | 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.
The four losses did not share a setup. EURUSD ran two pullback longs. NAS100 ran a VWAP rejection short. XAUUSD ran a bullish pullback at NY session support. Three instruments, three distinct families, all stopped at -1R.
Every confluence score landed in the 60-67 percent band, the floor of the actionable range. Macro, structure, and cross-asset were all right. What was not yet present was the second-bar absorption signal that confirms a level is defending, not just being touched.
A 60-67 percent confluence trade has a roughly 40 percent failure rate by construction. That is why 55 percent is the floor, not the target. Four floor-of-the-range trades clustered across three sessions and the failure rate compounded. Wrong tail, not off distribution.
The same setups, at the same scores, will be taken again. Removing the 60-67 percent band would lower expected value and would have skipped the Friday US30 long.
The Risk Agent did not engage a circuit breaker because the system does not have one. Position sizing is fixed per trade. A pause after the third Friday loss would have skipped the US30 long six minutes later that ran +1.53R and pulled the recap window green.
The Trend Agent's confluence threshold stayed at 55 percent through the entire window. The four losses and the streak-breaking Friday US30 long were all evaluated under the same scoring rules. A system that tightens the floor under stress is a discretionary trader pretending to be a system.
The Macro Agent held risk-tolerant through Mon-Tue and consolidating-USD with intraday sweep-reverse tape through Wed-Fri. The regime read was right; what the read does not guarantee is per-trade confirmation.
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.
Same signals, same risk framework, different foundation model.
EURUSD took two losses, both pullback longs at session-support reads. Tuesday and Friday entries stopped on shelf failures inside the first hour. -2.0R net.
All EURUSD this week →XAUUSD took one loss, the Friday Apr 17 long. A buyer stack above the shelf swept the position. -1R.
All XAUUSD this week →US30 took zero losses. Both trades ran to TP1 (Apr 13 short +1.00R, Apr 17 long +1.53R). The Friday long broke the streak.
All US30 this week →NAS100 took one loss, the Thursday VWAP rejection short. The index reclaimed VWAP through the cash open and resolved at -1R.
All NAS100 this week →US500 took zero trades. Consolidation, no patterns above threshold.
All US500 this week →Loss of the week: XAUUSD Long · -1R
What the system saw: bullish pullback into a NY session support shelf. Macro gated bullish-gold on softening yields. Cross-Asset confirmed with US30 bid. Confluence 67 percent, the highest-grade loss of the week.
What went wrong: within the first hour, a buyer stack we had not weighted swept the support level. The volume model under-weighted resting orders built above the shelf over the prior NY session without recent price interaction. Stop at -1R; equity touched the trough at $99,358.
Lesson: every macro input was right. Resting-order volume-scoring is what we are tuning. We would take the trade again.
What the system saw: VWAP rejection short after the index rejected the prior session's high. Macro gated bearish-equities on bid bonds. Cross-Asset confirmed with US30 stalling. Confluence 60 percent.
What went wrong: within the first hour the index reclaimed VWAP through the cash open and the position stopped at -1R. A 60 percent confluence trade has roughly a 40 percent failure rate by design.
Lesson: the bearish read was not wrong. The timing was right by what the system measures, wrong by what the cash open printed.
What the system saw: pullback to session support after the pair held the prior day's lows. Macro gated risk-tolerant with a soft DXY. Cross-Asset confirmed with bid yields. Confluence 62 percent.
What went wrong: the support shelf failed inside the first hour as a broader risk-off ripple ran through the dollar pairs. The position stopped at -1R on a cross-asset risk shift inside the trade lifecycle.
Lesson: the bullish-EUR read was inside the regime envelope. The variance is the cost of a 35-40 percent win-rate system on the way to long-run expectancy.
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 | -4R | −$8,000 |
The honest reading is that the system traded its full sizing, took every setup that cleared threshold, and gave back -4R on four independent trades, while the same week produced three TP3 winners on Mon and Tue and a streak-breaking US30 long on Friday. Recap +1.21R net; drawdown -4.00R on the loss-side ledger.
This is the editorial reason a drawdown report on a positive week is worth publishing. The asymmetry is usually only visible at the rolling-100-trade resolution. This week it was visible at week resolution because the winners and losers landed inside the same five sessions. From the SkyAnalyst Team.
The volume-scoring treatment of resting orders is the operational item out of this week. When buyers (or sellers) have built positions above (or below) a level over hours without recent price interaction, the model under-weights them on the sub-hour aggregation window. The Friday XAUUSD long matches this profile. A fix is in testing. Backtests show a 6-8 point confluence-score reduction on similar scenarios, which would have moved the XAUUSD setup below threshold.
A 35-40 percent win rate paired with a 1.67R average winner target and the asymmetric tail the three TP3 winners on Mon and Tue illustrate is the rate-and-reward profile this system was designed around. One 1.67R winner covers 1.67 losers, and the three TP3 winners covered the four losers and then some at the broker fill. At a 37 percent win rate the rolling expectancy on 100 trades sits modestly positive even when individual weeks land deep in the red. Van Tharp's R-multiple framework, Schwager's analysis of trend-following systems, and standard binomial treatment of independent trial outcomes all conclude that a 35-40 percent system has expected longest losing streaks of 5-8 trades inside any rolling 100-trade window. This week's 4-loss streak is below that median.
A -4R intraweek drawdown on the $100,000 / 2 percent risk baseline represents 7.45 percent of equity at the trough. For a system with this volatility profile, drawdowns of 5-10 percent are inside the first standard deviation of expected variance. A 7.45 percent intraweek draw that closed the week at +1.21R net is well inside that envelope. Drawdowns become signal rather than noise when they exceed the historical 95th percentile of the equity curve. We are not close.
The single concept worth holding onto is this: judge a system on its 100-trade rolling window, not its weekly one. The shorter the window, the more variance dominates the signal. A drawdown report exists to make the variance visible. The math, extended to the right horizon, is what makes the variance pay.
The system has no streak-aware circuit breaker. A pause after the third Friday loss would have skipped the US30 long six minutes later that ran +1.53R. The discipline is that the threshold is the threshold.
A -4R window with a 4-trade streak on a 35-40 percent system is well inside the first standard deviation of expected variance. Binomial treatment predicts longest streaks of 5-8 inside any rolling 100-trade window. This week's streak is below that median.
The recap projects winners at TP1 baseline and counts every loss the same way, landing at +1.21R net for Apr 13-19. The drawdown report counts only losses and reports -4.00R. Both apply the same methodology to different slices of the same five sessions.
Claude Opus 4.6 produced every entry on both sides, including the three TP3 winners and the streak-breaking US30 long. The asymmetry comes from the rate-and-reward profile, not from a model running differently in different conditions.
No. Drawdowns of 5-10 percent are routine for this risk profile. The honest signal in this week's data is the volume-scoring fix, not the drawdown itself.
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.
Three losses, all at minus 1R. Net minus 3R for the loss-side ledger. Longest streak of 1. Original printed 4 losses and a streak of 2; the cancelled-trade fix dropped one phantom NAS100 row from Feb 26.
Seven trades, four winners, three losses, +1.21R net on a TP1 baseline. Original printed nine trades and +0.80R; the cancelled-trade fix dropped one paused NAS100 row from Feb 26. Corrected ledger.
Twenty-one trades, thirteen winners, eight losers, +4.41R net on a TP1 baseline. Original published as 24 trades and +6.64R; the cancelled-trade fix dropped 3 paused rows the dashboard never had.