Mar 16-22, 2026: Seven Losses, Seven R Given Back to the Tape
Seven losses, -7.0R given back across four instruments, longest losing streak of 4. The deepest trough in the published record. The recap counts the full 12-tra
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.
Seven losses. Four in a row from Tuesday's open into Wednesday's first hour, all four printing -1R stops on confluence-cleared setups across the US index complex and USDJPY. Net for the loss-counting window: -7.0R, equivalent to -14,000 dollars of simulated drawdown on the 100,000 / 2 percent risk baseline. Trough equity 95,167.32 by Friday close, -7.60 percent from peak. The longest losing streak: 4 trades, the deepest in the published record. Every loss a clean -1R stop, no slippage, no excursions worth flagging. Through Mar 23, 2026, the cumulative ledger reads +4.63R YTD across 48 trades from Jan 12 inception. The simulated $100,000 account at 2 percent risk per trade sits at $109,259.29 on the static line and $108,455.73 on the compounded line. The running total has absorbed this window and stayed well above the starting balance even at the trough. This is not the recap. The recap covers all 12 canonical trades inside the same five sessions and runs a different scoreboard on the win side. The drawdown report exists to make the loss side legible on its own terms: each of the seven losses, why each cleared threshold at entry, what failed in the tape after entry, and the rolling-window statistics that frame a -7R week on a 0.92R-average-winner system. The framing for everything that follows: drawdown is the cost the asymmetry pays. This week, the cost printed deeper than any window since inception.
Act 1: Tuesday's index cluster opens the books
Tuesday Mar 17 opened the loss column with three correlated index entries inside 26 minutes. US500 long at 14:10 UTC on a pullback-buy into prior breakout support, grade C+. NAS100 long at 14:31 UTC on the correlated long-side confluence read, grade C+. US30 long at 14:36 UTC, the third leg of the index complex, grade C+ again. All three setups cleared confluence in isolation; all three stopped at -1R inside the same session window. Three trades, three stops, minus 3R, on a single Tuesday afternoon.
The Cross-Asset Agent did not gate the cluster because the current logic checks correlation by instrument at evaluation time, not by intra-window timing across the index complex. The per-instrument confluence math read the three setups as three independent positions. The tape read them as one trade with triple sizing.
Act 2: Wednesday extends the streak to four
Wednesday Mar 18 opened with a USDJPY pullback long retest-and-hold at 14:15 UTC, grade C+. The retest held for nine minutes before the dollar bid faded into the New York session and the stop printed at -1R. That was the fourth consecutive loss, the longest losing streak in the published record. Trough equity by Wednesday afternoon: 95,000.00, a -7.77 percent drawdown from the Monday close.
The Risk Agent did not engage a streak-aware override because there is no streak-aware override in the architecture. Sizing stayed fixed at 2 percent. The confluence threshold stayed at the actionable floor. The system continued evaluating the next setup the same way it would have on Monday.
Act 3: Thursday and Friday close the give-back
Thursday Mar 19 ran the only B-grade loss of the week: NAS100 short at 14:50 UTC on a clean breakdown read. The Trend Agent scored the structure, the Macro Agent did not flag a regime gate, the Cross-Asset Agent confirmed the divergence against US500. The index reclaimed the breakdown on a higher-volume bar than the rejection bar that triggered entry, and the stop printed at -1R. The cleanest loss of the window: every step of the framework executed as designed, the market simply disagreed.
Friday Mar 20 closed the give-back with two more stops. US30 short at 15:13 UTC on a pullback-failure read into resistance, grade C+. EURUSD short at 15:28 UTC on a retracement into resistance, grade C+. The Risk Agent allowed both. Final equity by Friday close: 95,167.32, a -7.60 percent drawdown from the peak set the prior Monday.
Every trade the system took.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Mar 17 | 14:10 UTC | US500 | Long | GPT-5.4 | US500 LONG — Pullback buy into prior breakout support | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Mar 17 | 14:31 UTC | NAS100 | Long | GPT-5.4 | NAS100 LONG | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Mar 17 | 14:36 UTC | US30 | Long | GPT-5.4 | US30 LONG | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Mar 18 | 14:15 UTC | USDJPY | Long | GPT-5.4 | USDJPY pullback long retest-and-hold | C+ | -0.25R(SL) | -$500(SL) | Stop hit | - |
| Mar 19 | 14:50 UTC | NAS100 | Short | GPT-5.4 | NAS100 SHORT | B | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Mar 20 | 15:13 UTC | US30 | Short | GPT-5.4 | US30 SHORT (pullback failure into resistance) | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Mar 20 | 15:28 UTC | EURUSD | Short | GPT-5.4 | EURUSD SHORT retracement into resistance | 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
The seven losses this week share one structural commonality more cleanly than most loss windows: five of the seven were directional bets on the US risk complex, and four of those landed inside windows where the index family moved as a single instrument. The Tuesday cluster is the textbook case. The Friday US30 short joins the same family at a different session.
What the pattern is
Six C+ setups, one B-grade (Thursday's NAS100 short). Every loss cleared confluence in isolation. The Macro Agent did not flag a regime gate on any of the seven. The Cross-Asset Agent confirmed correlated tape on the index entries. The pattern is not bad reads at the per-trade layer; the pattern is correlation arithmetic at the portfolio layer. Three correlated index longs sized independently behaved like one trade at triple sizing once the tape rolled in unison.
Why these failed
A C+ setup has a stop probability around 60 to 65 percent by construction; that is the band the system sizes for. Seven losses on twelve canonical trades sits inside the variance envelope at the rolling-100-trade horizon. The deeper problem is the cluster math: Tuesday's three correlated stops did the structural damage to the equity curve. The per-instrument reads were defensible. The combined exposure was not.
What we keep doing
Same setup catalog, same confluence scores, same macro gate next week. The architecture is built to absorb 35 to 40 percent win-rate variance over the rolling-100-trade window. Removing the C+ band on the strength of a single sample would skip the multi-R outliers that carry expectancy on the win side. The structural item, the correlation cap across the index complex, gets scoped on the engineering side, not papered over by changing the entry rules.
Decision highlights
The Risk Agent allowed three correlated index entries inside 26 minutes on Tuesday afternoon. Each setup cleared confluence in isolation, which is the rule the Risk Agent enforces today. The portfolio-correlation cap on the index complex, which would have sized the second and third entries down or rejected them, is the clearest structural item this report logs.
Thursday's NAS100 short took a full stop at the B grade. Every step of the framework executed as designed. The market disagreed. The decision to publish that loss in the cleanest-loss column rather than the avoidable-loss column is the editorial line the drawdown report holds. A B-grade stopping at -1R is the cost of running a system whose edge plays out over hundreds of trades.
Friday's two pullback-failure entries went into resistance on US30 and EURUSD inside fifteen minutes. Both within rules. Both stopped. The decision the desk made on Friday close was not to change the entry catalog on a sample of two. Recency bias at the trough is the most reliable way to convert a normal drawdown into a permanent loss; the literature is unambiguous on that point.
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.
- USDJPY-0.3R · 1 trade
- US500-1.0R · 1 trade
- EURUSD-1.0R · 1 trade
- NAS100-2.0R · 2 trades
- US30-2.0R · 2 trades
Same signals, same risk framework, different foundation model.
Six instruments, six stories.
EURUSD: one loss (Fri Mar 20 short at 15:28 UTC). Retracement into resistance, C+. The European close bid lifted the pair through the stop in twelve minutes.
All EURUSD this week →GBPUSD: no trades this week. The pair sat outside our setup criteria for the full window; neither the Trend nor the Cross-Asset Agent flagged a confluence read worth sizing.
All GBPUSD this week →US30: two losses (Tue Mar 17 long at 14:36 UTC in the index cluster; Fri Mar 20 short at 15:13 UTC on a pullback-failure read). Both C+. The index traded in lockstep with NAS100 and US500 on both sessions, compounding the correlation cost.
All US30 this week →NAS100: two losses (Tue Mar 17 long at 14:31 UTC in the cluster; Thu Mar 19 short at 14:50 UTC). The Thursday short was the only B-grade loss of the week, the cleanest read of the seven.
All NAS100 this week →USDJPY: one loss (Wed Mar 18 long at 14:15 UTC). Pullback retest-and-hold, C+. The retest held for nine minutes before the dollar bid faded into the New York session and rolled the entry through the stop.
All USDJPY this week →US500: one loss (Tue Mar 17 long at 14:10 UTC). Pullback-buy into prior breakout support, C+. First leg of the Tuesday cluster, the entry that set the tone for the session.
All US500 this week →Loss of the week: US30 Short · -1R
Losses worth learning from
Loss 1: Tuesday Index Cluster (US500 / NAS100 / US30 Longs Mar 17 14:10-14:36 UTC)
What the system saw: pullback-buy on US500 into prior breakout support at 14:10 UTC; NAS100 long on the correlated long-side confluence 21 minutes later at 14:31; US30 long five minutes after at 14:36. Macro long-tilt across the complex, Cross-Asset confirmed correlated strength rather than vetoing it, all three setups landed in the actionable band at C+ grade.
What went wrong: prior breakout support did not defend. Structure rolled within the hour and stopped all three at -1R. The Cross-Asset Agent did not gate the cluster because the current logic checks correlation by instrument at evaluation time, not by intra-window timing across the index complex.
Lesson: the per-instrument reads were clean. The correlation gap is the structural item this report logs. The fix is a portfolio-correlation cap at the Risk Agent layer, not a change to the per-trade confluence math. The next release scopes the cap.
Loss 2: NAS100 Short Mar 19 14:50 UTC (B-grade)
What the system saw: a clean breakdown read after the index lost intraday support on a rejection bar at prior session resistance. Macro short-tilt on a firming session tape, Cross-Asset confirmed the divergence against US500. The only B-grade setup on the loss side this week.
What went wrong: the index reclaimed the breakdown on a higher-volume bar than the rejection bar that triggered entry. The stop printed at -1R on a clean SL fill.
Lesson: a B-grade has a lower stop probability than a C+, and this one stopped anyway. The higher-conviction band still sits inside the variance envelope at sample size seven. This is the cleanest illustration of the R-multiple distribution this week, the trade that proves the edge is honest.
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 | -6.25R | −$12,500 |
All six agents combined.
From the desk
Through Mar 23, 2026, the cumulative ledger reads +4.63R YTD across 48 trades from Jan 12 inception. The simulated $100,000 account at 2 percent risk per trade sits at $109,259.29 on the static line and $108,455.73 on the compounded line. The spread between the two figures, roughly $800, is the geometric cost of compounding 2 percent risk through a series of mixed-sign R-multiples; on a winning week the compounded path runs hotter than the static, on a losing week it runs cooler, and over a long enough record the two converge inside disciplined sizing. This week's give-back was approximately $7,800 against the prior peak and approximately $14,000 against an unleveraged 7R count, both figures resolved on the same record that carries the +4.63R YTD.
The honest reading: the system took every setup that cleared threshold, gave back -7R on seven trades, and printed the deepest weekly trough in the published record at -7.77 percent. The recap counts the full 12-trade canonical window across the same five sessions; the drawdown report counts only the loss-side ledger.
Does the system need a streak-aware override after the four-trade streak from Tuesday's open into Wednesday's first hour? No. The architecture is built to absorb 35 to 40 percent win-rate variance over the rolling-100-trade horizon, and a streak-aware pause at the trough would skip more winners than losers across history. The structural fix is the portfolio-correlation cap, scoped at the Risk Agent layer and shipping in the next release. The architecture absorbed the deepest drawdown in the published record without intervention, by design.
What we're tuning
The structural item this report logs is the portfolio-correlation cap across the US index complex. Tuesday's cluster of three correlated stops did the deepest structural damage to the equity curve, and the Friday US30 short joins the same family on a different session. The fix scopes at the Risk Agent layer: a check that flags concurrent confluence reads on US30, NAS100, and US500 inside a defined intra-window timing band and either rejects the second and third entries or sizes them down to preserve the per-trade risk budget. The next release ships the gate.
We do not change the entry catalog on the strength of Friday's two pullback-failure stops. Two correlated stops inside fifteen minutes is inside the expected variance of a system targeting 0.92R average winners on a 54 percent rolling win rate. Adjusting the entry rules on a sample of two would convert a normal loss week into a methodology change driven by recency bias. The framework holds. The portfolio layer gets a new check. That is the boundary between disciplined iteration and reactive overfitting, and we hold it.
What the numbers actually mean
The single most useful question on a -7R week: how often does a system with a 54 percent rolling win rate and a 0.92R average winner produce a 4-trade losing streak and a -7R window? Treat each trade as an independent Bernoulli trial with the empirical loss probability the trade record carries through Mar 23. The probability of a 4-trade losing streak over any 12-trade window is roughly 12 to 15 percent, about one window in eight at this cadence. The probability of seven losses on twelve trades is lower, in the body of the distribution but on the unfavorable side of the median. Van Tharp's work in Trade Your Way to Financial Freedom is the standard reference on the R-multiple distribution math; the framework predicts windows of this depth several times in any hundred-trade sample purely as a function of variance, not as evidence that the edge has changed.
Schwager's interviews in the Market Wizards series make the same point from the trader's seat. Every system with positive expectancy posts losing weeks, losing months, and occasionally losing quarters. The discipline that separates a fund that survives the drawdown from one that does not is whether the operator changes the rules at the trough or holds them through the recovery. The 7.77 percent intraweek drawdown printed on this report sits inside the first standard deviation of expected variance for a 0.92R-average-winner system; drawdowns of 5 to 10 percent are routine on the rolling horizon. A drawdown report is one slice of a much longer record, and the sample size in any single window is too small to update the prior on the edge.
The longest-losing-streak math extends the same logic. A standard binomial treatment over the rolling-100-trade window predicts expected longest losing streaks of 5 to 8 trades for a system in this win-rate band, with the worst case stretching to 9 or 10 in any hundred-trade sample. This week's 4-trade streak is below the rolling-window median, uncomfortable at week resolution and unremarkable at hundred-trade resolution. The publisher's commitment is to put the drawdown on the record alongside the year-to-date, not to bury it inside a quarterly summary. Through Mar 23, the YTD line reads +4.63R across 48 trades at 54.17 percent. The week we just walked through is inside the expected envelope for that record.
At a Glance
Drawdown questions
How likely is a 4-trade losing streak on a system in this win-rate band?
On the rolling 54 percent win rate the YTD ledger carries, the probability of a 4-trade streak over any 12-trade window is roughly 12 to 15 percent, about one window in eight at this cadence. Inside the body of the distribution, not a tail event.
Why publish a drawdown report on a -7R window instead of waiting for a recovery?
Because every legitimate trading fund publishes its drawdowns alongside its winners, and the publisher's commitment is to put both numbers on the record at the same time. A drawdown report on a flat week would be noise; a drawdown report on the deepest weekly trough in the record is the cadence working as designed.
What does a 7.77 percent intraweek drawdown mean for forward performance?
It sits inside the first standard deviation of expected variance for a 0.92R-average-winner system on a 54 percent rolling win rate. Drawdowns of 5 to 10 percent are routine on the rolling horizon. They become signal when they exceed the 95th percentile of the variance envelope, which this window does not.
Why did the system not pause after the Tuesday cluster or the Wednesday fourth loss?
Sizing is fixed per trade by design, and the architecture does not carry a streak-aware override. A pause after the third Tuesday stop would have skipped the Thursday and Friday evaluations regardless of their setup quality. Across the rolling-100-trade history, a streak-aware override skips more winners than losers. The structural fix that addresses this week's pattern is the portfolio-correlation cap on the index complex, not a streak pause.
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: +4.79R YTD across 48 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.