Four Losses, Net Positive: A Drawdown Week That Closed Green
Four losses, -4.00R given back, a 2-trade losing streak inside the window. The same week produced the biggest winner of February. Net for Feb 23-Mar 1 closed at
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.
Four losses. Two on Thursday afternoon back-to-back. Net result for the loss-counting window: -4.00R, every loss at exactly -1R, no exits closer to the stop and no exits past it. The longest losing streak inside the window was 2 trades, set inside a single 22-minute span on Feb 26 when a NAS100 short and a US30 long stopped within the same session. Looked at in isolation, this is a textbook drawdown report week. This drawdown report is not the recap. The recap covers all of Feb 23-Mar 1 and lands at +0.80R net across 9 trades, with a 55.6 percent win rate and the biggest winner of February embedded inside the same five sessions. See the Feb 23-Mar 1 Weekly Recap for the full ledger and the February monthly recap for the closing-month context. This document opens the books on the loss side: each of the four losses, why each cleared threshold, what failed in the tape after entry, and the rolling-window statistics that say a 2-loss streak on a 35-40 percent system is below the median expected. The framing for everything that follows: drawdown is the cost the asymmetry pays. This week the asymmetry actually paid.
Act 1: Tuesday opens the loss column
The week opened with a US30 short at 15:33 UTC on Feb 24, a mean-reversion entry at intraday resistance after the index rallied into the prior day's pivot. The Trend Agent flagged the setup at C+. Macro gated short-friendly on a soft DXY and bid bonds. Cross-Asset confirmed with NAS100 stalling at the same time. The position stopped within the hour at -1R when the index reclaimed the pivot and held the breakout.
By the close of Tuesday the loss-side tally read 1 loss, -1R. Inside the same Tuesday, the recap's column carried two NAS100 winners that ran in parallel. The drawdown column read red. The recap column read green. Two reports of the same day, two slices of the same ledger.
Act 2: Thursday delivers the streak
Thursday Feb 26 was the day this drawdown report exists for. At 15:49 UTC the Trend Agent triggered a NAS100 short on a trend-continuation setup at a weak retest of intraday lows. Confluence cleared 60 percent. The position was paused at -1R when the system's internal stop logic caught a pre-stop reversal. Twenty-two minutes later, at 16:11 UTC, a US30 long flagged on a buy-the-dip setup at session VWAP. Confluence cleared. The position stopped at -1R inside the hour.
That is the 2-loss streak the longest-streak metric counts. It ran on consecutive trades inside a single afternoon. The Risk Agent did not engage a circuit breaker. Sizing stayed fixed. Threshold stayed at 55 percent. The system kept reading the next setup the same way it would have read it on Monday.
Act 3: Friday gives back one more, then takes the week back
Friday Feb 27 opened with a US500 long at 16:39 UTC, a pullback-buy at the rising VWAP after the index held the prior day's lows. Confluence cleared 62 percent. The position stopped at -1R when the broader risk tape softened into the bond auction window. That is the fourth loss and the loss-of-the-window the result reveal flags. Net result for the loss-counting window: -4.00R, equivalent to -$8,000 of simulated drawdown on the $100,000 / 2 percent risk baseline. Trough equity hit 97,771.93 at the second Thursday stop, a -3.93 percent drawdown from peak.
What is not on the loss-side ledger but is the editorial story of the same five sessions: a US30 short later that Friday afternoon ran the entire intraday move and closed at +4.33R, the biggest single-trade winner of February. Two NAS100 winners on Tuesday added meaningful R. A US500 short on Friday added more. The recap's net for the same nine trades closed at +0.80R. One winner, the Friday US30 short at +4.33R, more than covered the four losers in this report. See the Feb 27 US30 short case study for the trade that did the covering. The drawdown report exists to make the loss side visible. This week, the loss side and the win side together closed the books in green.
Related reading: companion weekly recap · February 2026 monthly recap · same-week winner that offset losses.
Every trade the system took.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Feb 24 | 15:33 UTC | US30 | Short | Claude Opus 4.6 | US30 SHORT (mean-revert at resistance) | C+ | -1.0R | -$2,000 | Stop hit | - |
| Feb 26 | 15:49 UTC | NAS100 | Short | Claude Opus 4.6 | Setup #1 · NAS100 SHORT (trend-continuation on weak retest) | C+ | -1.0R | -$2,000 | Stop hit | - |
| Feb 26 | 16:11 UTC | US30 | Long | Claude Opus 4.6 | US30 LONG (Buy-the-dip) | C+ | -1.0R | -$2,000 | Stop hit | - |
| Feb 27 | 16:39 UTC | US500 | Long | Claude Opus 4.6 | US500 LONG (pullback buy) | 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.
Pattern of the week
The four losses this week did not share a single setup. US30 ran one short at intraday resistance and one long at session VWAP. NAS100 ran a trend-continuation short at a weak retest. US500 ran a pullback long at rising VWAP. Four instruments-and-directions, three distinct setup families, all four stopped at -1R.
What the pattern is
The structural commonality across all four entries: every confluence score landed in the 60-67 percent band, the actionable floor of the entry range. The macro was right, the structure was right, the cross-asset was right. What was not yet present in any of these entries was the second-bar absorption signal that confirms a level is defending, not just being touched. A discretionary trader watching the same tapes would call these "early entries." The system calls them "entries at the threshold."
Why these failed
A 60-67 percent confluence trade has a roughly 40 percent failure rate by construction. That is the whole reason 55 percent is the floor, not the target. When four floor-of-the-range trades cluster inside one week, the per-week failure rate compounds. Four independent 40-percent-failure-likelihood trades have a non-trivial probability of all stopping inside one window. Mathematically, this week's loss side is on the wrong tail but not off the distribution.
What we keep doing
The same setups, at the same scores, with the same macro context, will be taken again next week. The system is not a discretionary trader who needs maximum-conviction entries. It is a portfolio of conditional probabilities that earn their expectancy over a rolling window. Removing the 60-67 percent confluence band to "improve" the win rate would lower the expected value of the strategy and would have skipped the Friday US30 short that printed +4.33R and carried the week. The question worth asking is the absorption-scoring question we surface in the teardowns, not the confluence-floor question.
Decision highlights
The Risk Agent did not engage a circuit breaker after the Thursday 2-loss streak because the system does not have one. By design, position sizing is fixed per trade and is not modulated by recent results. A circuit breaker that paused after the second consecutive loss on Feb 26 would have skipped the Friday US30 short later in the recap window that ran +4.33R. That is the exact failure mode of a streak-aware override: the trades you skip while regrouping are the trades that pay for the regrouping. The discipline is that the threshold is the threshold.
The Trend Agent's confluence threshold stayed at 55 percent through the entire window. No setup was rejected for being "below threshold during a streak," and no setup was promoted for being "good enough during a streak." The four losses on this report and the +4.33R Friday US30 short 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. We do not do that.
The Macro Agent held the regime tag at consolidating-USD with a soft-tilt through Tuesday and into Wednesday, then updated to closing-week-strength after the Wednesday session. The four losses on this report split across both regime tags: the Feb 24 US30 short ran under the consolidating tag, the Feb 26 pair ran across the tag transition, and the Feb 27 US500 long ran under the closing-week-strength tag. The regime read was right by what the closing-week tape produced. What the regime read does not guarantee is per-trade confirmation, which is the lesson the absorption-scoring tuning surfaces.
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.
Same signals, same risk framework, different foundation model.
Six instruments, six stories.
EURUSD took zero losses this window. The Trend Agent did not flag a EURUSD setup that cleared confluence on the loss side. On the loss-side ledger this report counts, EURUSD was absent.
All EURUSD this week →XAUUSD took zero losses this window. Gold consolidated through the week without printing the structural setups the system targets. No setup, no entry, no contribution to the -4R draw.
All XAUUSD this week →US30 took two losses, one short and one long. The Tuesday short at intraday resistance stopped on a pivot reclaim. The Thursday long at VWAP stopped on a midday reversal. Two structurally different trades, same -1R outcome each.
All US30 this week →NAS100 took one loss, the Thursday Feb 26 short at 15:49 UTC. The setup was a trend-continuation entry at a weak intraday low retest. The internal stop logic caught a pre-stop reversal and the position resolved at -1R.
All NAS100 this week →USDJPY took zero trades this window. The dollar-yen tape held a tight range without printing the trendline retest the system targets. No setup, no entry, no contribution to the -4R draw.
All USDJPY this week →US500 took one loss, the Friday Feb 27 long at 16:39 UTC. The pullback-buy at rising VWAP stopped on the broader risk tape softening into the bond auction window. The setup was textbook by what the system measures.
All US500 this week →Loss of the week: US500 Long · -1R
Losses worth learning from
Loss 1: NAS100 Short Feb 26 15:49 UTC
What the system saw: the Trend Agent flagged a trend-continuation short on a weak retest of intraday lows. Macro gated short-tilt on the consolidating-USD tag and a softening risk tape. Cross-Asset confirmed with US30 stalling at the same horizon. Confluence read 61 percent, just inside the actionable band. Setup grade C+.
What went wrong: within the evaluation window after entry, the index printed a higher-low absorption signal the volume model under-weighted. The retest had been weak, but the absorption that followed was not. The internal pre-stop reversal logic caught the move and resolved the position at -1R rather than waiting for the SL print.
Lesson: every macro-level input was right. The absorption-scoring treatment of post-retest reversal candles is the operational item we are tuning. We would take the trade again at the same score. The tuning is housekeeping inside an actionable confluence band, not a redesign of the threshold.
Loss 2: US30 Long Feb 26 16:11 UTC
What the system saw: the Trend Agent triggered long at session VWAP after a clean intraday pullback into a previously defended level. Macro gated bullish-risk on the regime transition into closing-week-strength. Cross-Asset confirmed with bid yields. Confluence read 64 percent, clean entry inside the actionable band. Setup grade C+.
What went wrong: a midday auction reversal swept the VWAP support and stopped the position inside the hour. The intraday mean-reversion was the kind of pattern the strategy is built to absorb at this confluence level. A 64 percent confluence trade has roughly a 40 percent failure rate by design, and this trade landed inside that 40 percent.
Lesson: the bullish read on the closing-week tape was not wrong, and the same closing-week-strength regime produced the +4.33R US30 short on Friday afternoon. The timing of this specific entry was right by what the system measures and wrong by what the auction printed. We would take the trade again. There is no edit to the model that fits this loss without overfitting to the auction-pivot artifact.
Loss 3: US500 Long Feb 27 16:39 UTC
What the system saw: the Trend Agent triggered long on a pullback-buy at rising VWAP after the index held the prior day's lows. Macro gated bullish on the closing-week-strength regime. Cross-Asset confirmed with US30 trending and bonds offered. Confluence read 62 percent, inside the actionable band. Setup grade C+.
What went wrong: the broader risk tape softened into the bond auction window. The position stopped at -1R when the index gave back the pullback hold. The setup was textbook by what the system measures at entry. The catalyst that resolved it was the bond auction print, which is the kind of timed-event reversal the model accepts as part of the variance envelope at this confluence level.
Lesson: the bullish read for the closing-week tape was right at portfolio level, as the +4.33R Friday US30 short demonstrates. This specific US500 entry caught the auction reversal. We would take the trade again at the same score. The variance is the cost of running a system at a 35-40 percent win rate on the way to its long-run expectancy.
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 | -4R | −$8,000 |
All six agents combined.
From the desk
The honest reading of this week 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 the largest single-trade winner of February. The recap closes at +0.80R net for the full window, which is the literal definition of a green week, and the drawdown report closes at -4.00R on the loss-side ledger, which is the literal accounting of what the four losers cost. Both numbers are correct under the same methodology applied to different slices of the same five sessions.
This is the editorial reason a drawdown report on a positive week is worth publishing. The asymmetry the strategy is engineered around is usually only visible at the rolling-100-trade resolution. This week it was visible at week resolution because the +4.33R Friday US30 short happened to land inside the same five sessions as the four losses. The math that makes the strategy work over the long run does not change between weeks. The visibility of the math changes. We can show it more cleanly this week than most weeks. The arithmetic underneath is the same arithmetic that runs every week.
What carries into next week is the absorption-scoring release in test, the macro question of whether closing-week-strength carries into the early-March open, and the operational reality that the next economic calendar window is heavier through Wednesday. We expect a fuller week of setups. We expect some of them to win, some of them to lose, and the per-trade variance to do what it does. We will report whatever happens.
What we're tuning
The absorption-scoring treatment of post-retest reversal candles is the operational item out of this week. When a weak retest is followed by a higher-low absorption print on the next bar, the current volume model under-weights the absorption on the sub-hour aggregation window. The Feb 26 NAS100 short matches this profile most cleanly. A fix is in testing for the next signal cycle. Initial backtests show a 5-7 point reduction in confluence-score on similar post-retest reversal scenarios, which would have moved the NAS100 short below the 55 percent entry threshold and the system would have skipped it. The other three losses do not share this artifact and the tuning would not have caught them. Whether the fix generalizes only emerges after a few weeks of live signal. We will report whatever the data shows in the next drawdown window that produces a similar setup.
What the numbers actually mean
A 35-40 percent win rate, paired with a 1.67R average winner target and the asymmetric tail this week's +4.33R outlier illustrates, is the rate-and-reward profile this system was designed around. The arithmetic: one 1.67R winner covers 1.67 losers, and one 4.33R outlier covers more than four losers all on its own. At a 37 percent win rate the rolling expectancy on 100 trades sits modestly positive even when individual weeks land deep in the red. This is the inverse-relationship between win rate and reward target that any honest book on system trading walks the reader through. Van Tharp's R-multiple framework, Schwager's analysis of trend-following systems and their drawdown distributions, and any standard binomial-distribution treatment of independent trial outcomes converge on the same conclusion: a 35-40 percent system has expected longest losing streaks of 5-8 trades inside any rolling 100-trade window, with the worst-case observation across many simulations falling between 6 and 10. This week's 2-loss longest streak is well below that median. The 4-loss window total is uncomfortable to read at week resolution and inside the distribution at 100-trade resolution.
What is also worth knowing: a -4R intraweek drawdown on the $100,000 / 2 percent risk baseline represents 3.93 percent of equity at the trough on Thursday afternoon. For a system with the rolling 12-month volatility characteristics this one is calibrated to, drawdowns of 5-10 percent are inside the first standard deviation of expected variance. A 3.93 percent intraweek draw that closed the week at +0.80R 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 to that threshold. The week closed green at the equity level the strategy is measured on.
The single concept worth holding onto: judge a system on its 100-trade rolling window, not on its weekly window. The shorter the window, the more variance dominates the signal. The longer the window, the more the underlying expectancy emerges. A drawdown report exists to make the variance visible. The math, when extended to the right horizon, is what makes the variance pay. This week the math paid inside the same five sessions that produced the variance, which is rare at week resolution and routine at the rolling-window resolution the strategy is built for.
At a Glance
Drawdown questions
Why did the system not pause after the Thursday 2-loss streak?
The system has no streak-aware circuit breaker, by design. Position sizing is fixed per trade and is not modulated by recent results. A pause after the second consecutive Feb 26 loss would have skipped the Friday US30 short that ran +4.33R, the biggest winner of February. A streak-aware override is the exact failure mode that converts a positive-expectancy system into a discretionary one. The discipline is that the threshold is the threshold.
How does this week compare to the long-run drawdown distribution?
A -4R window with a 2-loss longest streak on a system with a 35-40 percent win rate and a 1.67R average target is well inside the first standard deviation of expected variance. Standard binomial treatment of independent 60 percent failure rates predicts longest losing streaks of 5-8 trades inside any rolling 100-trade window. This week's 2-loss streak is well below that median. The 3.93 percent intraweek equity drawdown sits inside the routine band for this rate-and-reward profile.
How does the recap's net R relate to the drawdown report's net R?
The recap projects winner outcomes at the TP1 baseline and counts every loss the same way this report does. The recap therefore includes the +4.33R Friday US30 short and the other winners in the window and lands at +0.80R net for the full Feb 23-Mar 1 window. The drawdown report counts only the loss-side ledger and reports -4.00R. Both numbers are correct under the same methodology applied to different slices of the same five sessions. See the <a href="/blog/weekly-recap-2026-02-23">Feb 23-Mar 1 Weekly Recap</a> for the full window.
What does it mean that the winners and losers came from the same model?
Claude Opus 4.6 produced every entry on the loss side and every entry on the recap's win side, including the +4.33R Friday US30 short. The same scoring rules ran on every setup. The asymmetry that makes the strategy profitable comes from the rate-and-reward profile, not from a model running differently in different conditions. A 60-67 percent confluence trade has a 40 percent failure rate. Some of those failures cluster, and some of the winners run to multi-R outcomes. Both are products of the same evaluation logic.
Should a 3.93 percent intraweek drawdown change how I think about this system?
No. A 3.93 percent intraweek drawdown that closed the week at +0.80R net is well inside the first standard deviation of expected variance for a strategy targeting 35-40 percent win rates with 1.67R average targets and an asymmetric tail. Drawdowns of 5-10 percent are routine for this risk profile. Drawdowns become signal rather than noise when they exceed the historical 95th percentile of the equity curve. The honest signal in this week's data is the absorption-scoring fix going into test, not the drawdown itself.
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.
From the SkyAnalyst Journal
Mar 2-8, 2026: Four Losses, Three Offsetting Winners
Four US500 losses, -4.00R given back, a 2-trade losing streak Thu into Fri. Three winners in the same five sessions covered most of the draw. The companion recap nets -0.62R.
US30 Short on March 2: Eleven Waits, One Enter, +1.2R at TP1
March opens with a sell-the-rally on the Dow. Twelve evaluations across fourteen minutes, eleven of them wait. The twelfth fired short at 48842 and banked TP1 at 48700.
NAS100 Long on March 4: TP1 Banked, Runner Stopped, +0.93R
A breakout continuation on the Nasdaq 100 cleared TP1 inside the New York session, then the runner reversed and tagged the original stop. Reported result reflects the TP1-baseline R.