Weekly Losses, Week of Jun 29: Four Stops, One Idea
Four losses for -4.00R inside a week that still closed green. Three of them arrived in fifty-two minutes on the same thesis, and that clustering, not any single
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 trades stopped out this week, and the honest version of the story is that they were not four independent events. One loss opened the week on Monday, a NAS100 washout bounce that failed at the first bell. The other three arrived together on Thursday morning, three pullback longs entered inside fifty-two minutes, all expressing the same belief about the same dip, all stopped for their budgeted -1R. Total damage: -4.00R, the week's longest losing streak at three, inside a week that still finished +1.48R net. Context keeps the number honest in both directions. The system gave back 4.00R across these losses against +27.29R YTD from its Jan 12 inception, and through Jul 5, 2026 a $100,000 simulated account risking 2% per trade sits at $154,585 on the static ledger. A losing cluster this size is not an anomaly to explain away; it is a recurring feature of any system with a real win rate, and the point of this report is to separate what the losses cost from what they taught. Every loss is in the index below, each with its setup and its stop. Two get full teardowns. The statistics section afterward puts the streak and the drawdown in the context of 136 trades of history, which is the only context that matters.
Act one: the Monday toll
The week's first trade was its first loss. A NAS100 opening-spike washout bounce, a mean-reversion long trying to catch the snap-back after an early flush, stopped out for -1R within the first hour of the week. The desk's response was the correct kind of nothing: no immediate re-entry, no size change, and the next NAS100 trade waited until Tuesday's fully qualified session-support long, which won.
Act two: the quiet middle
From Monday afternoon through Wednesday's close the loss ledger stayed empty. Seven consecutive trades finished green, and the simulated account walked from its Monday dip to the week's equity peak of $108,976 late Wednesday. This is the stretch that makes loss reports worth reading: the same pullback playbook that would fail on Thursday was, for three days, the entire engine of the week.
Act three: fifty-two minutes on Thursday
Between 14:24 and 15:16 UTC on July 2, the desk entered a US500 long at a prior-day-high retest, a EURUSD long at 61.8% fib support, and a US30 long into the 50 to 61.8% retrace. Three instruments, one thesis: the morning dip is a pullback, not a turn. The tape disagreed everywhere at once. All three stopped for exactly -1R, the streak counter read three, and the simulated account marked a 5.51% dip from Wednesday's peak. The desk took no fourth attempt and went flat into the holiday weekend.
Every trade the system took.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Jun 29 | 14:35 UTC | NAS100 | Long | Claude Opus 4.7 | VWAP Mean-Reversion LONG — Opening Spike Washout Bounce | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Jul 2 | 14:24 UTC | US500 | Long | Claude Opus 4.7 | Pullback Long — Prior Day High Retest | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Jul 2 | 14:57 UTC | EURUSD | Long | Claude Opus 4.7 | EURUSD Long — Pullback to Session Low / 61.8% Fib Support | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Jul 2 | 15:16 UTC | US30 | Long | GPT-5.5 | Long pullback into 50–61.8% retrace | 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
Every loss this week was a long, and every loss was a pullback entry at a level the system named in advance. That is not an indictment of the pattern; it is the pattern's cost of doing business. The identical family of entries, pullbacks to VWAP, fib, and structural supports, produced the bulk of the week's seven wins, including two trades that ran their complete target ladders.
What separated the winners from the losers was not setup quality on the chart. Monday's washout bounce and Thursday's three retracement longs were built the same way the winners were: defined level, structural stop, laddered targets. The difference was the tape's regime in the hours after entry, and that is precisely the variable no entry checklist can score in advance. The system pays -1R to find out. Four times this week, that is what it paid.
Decision highlights
Monday's stop was followed by the right kind of silence. The washout-bounce loss produced no revenge entry and no size adjustment; the next NAS100 position waited a full day for a setup that qualified on its own terms, and won. A system that treats a fresh loss as information rather than a debt to collect is doing the single hardest thing in discretionary trading by default.
Thursday's cluster is the week's real finding. The confluence gate evaluated each of the three longs against its own chart and approved each honestly, but no layer of the stack asked how many copies of the same thesis the book was accumulating. The judgment failure, to the extent there was one, was architectural rather than analytical: three qualified trades that were collectively one 3R position.
What did not happen after the cluster matters as much as what did. No stop was widened, no position averaged down, and no fourth pullback long was attempted into the afternoon. The desk finished the losing streak at exactly the size its rules budgeted, went flat, and let the week end. Losing 3R in an hour without breaking a single rule is, strange as it sounds, the system working.
Six instruments, six stories.
EURUSD: One loss. Thursday's long into 61.8% fib support stopped for -1R, one day after the euro desk banked a short at the same ratio's exhaustion. The level family is symmetric; Thursday's tape was not.
All EURUSD this week →GBPUSD: No losses this window. The pair's single trade ran its full ladder overnight.
All GBPUSD this week →US30: One loss, the week's loss of the window: Thursday's 50 to 61.8% retrace long, stopped for -1R after three winning pullback continuations earlier in the week.
All US30 this week →NAS100: One loss. Monday's opening-spike washout bounce stopped for -1R on the week's first trade; the desk recovered the instrument to a positive week by Wednesday.
All NAS100 this week →USDJPY: No trades this week, so no losses. Nothing cleared the pair's confluence gate, and a forced trade would have been the worse outcome.
All USDJPY this week →US500: One loss. Thursday's prior-day-high retest long was the first entry of the cluster and stopped for -1R.
All US500 this week →Loss of the week: US30 Long · -1R
Losses worth learning from
US30 Long: the third copy of one idea
What was right: the location. The 50 to 61.8% retracement band is where trend continuations legitimately reload, the stop sat beneath the zone at a level that falsified the read, and the loss was the budgeted -1R without slippage drama. After three winning US30 continuations earlier in the week, taking a fourth qualified pullback was consistent, not greedy.
What was wrong: by 15:16 UTC this was the third same-direction pullback entry of the morning, and the thesis it expressed had already been challenged twice. Nothing in the entry logic was misread on the chart; what went unpriced was the accumulating portfolio bet on a single interpretation of one morning's dip.
What we would do the same: take this exact trade in isolation, at this level, with this stop, every time. The correction this teardown argues for lives in aggregate exposure math, not in the read.
EURUSD Long: the ratio that works both ways
What was right: the 61.8% retracement is a legitimate decision level, and the desk's own week proves it, having banked a winning short at the same ratio's exhaustion the day before. Support location, defined invalidation, clean -1R exit.
What was wrong: the symmetry argument hides an asymmetry. Wednesday's short faded a corrective bounce inside an established downtrend; Thursday's long tried to catch a falling tape at a measured level while the morning's flow was pressing through supports across the board. Same ratio, different regime, and the regime is the part that failed.
What we would do the same: trust the level family. Fibonacci retracements earned their week even counting this loss. The lesson filed is about reading the day's regime before trusting the day's geometry, not about abandoning the geometry.
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
From the desk
We publish the losses with the same typography as the wins because the alternative is a marketing document. This week's report is unusually clean in one respect: every loss cost exactly what it was budgeted to cost, -1R, four times. The damage was never about any single trade. It was about three of them sharing a birthday.
In dollars, the window's give-back is approximately $8,000 of the static ledger at $2,000 per R. That ledger, through Jul 5, stands at $154,585 on a $100,000 simulated account risking a fixed 2% per trade since the Jan 12 inception, against $167,370 if the same trade sequence is compounded at 2% of the growing balance. Both figures absorbed this week's losses and both remain where they are because sizing never flexed, not on Wednesday when everything was working, and not on Thursday when nothing was. That consistency, more than any win, is the property we would show a skeptic first.
What we're tuning
The examination this window forces is a session-level correlation check. Today the confluence gate scores each instrument in isolation, which is correct for read quality and blind to accumulation: on Thursday it approved three same-direction pullback entries inside an hour without anything pricing the shared thesis. We are testing a rule that treats stacked same-direction entries within a session as one growing exposure for sizing purposes, so the second and third expressions of an idea pay a progressively higher admission price. Nothing ships until it survives backtesting against the full trade history; this note is the commitment to run that test, not the announcement of a change.
What the numbers actually mean
The numbers above need a framework more than they need a defense. The system's full record stands at a 60.29% win rate across 136 trades, with the average winner's first target paying just under 0.8R. Van Tharp's framing in Trade Your Way to Financial Freedom is the right lens: what matters is not any single outcome but the distribution of R-multiples the system generates, and a distribution with a 60% hit rate and capped -1R losses produces losing runs as a matter of arithmetic, not malfunction.
A three-loss streak deserves the same treatment. For any sequence of trades where roughly four in ten lose, three consecutive losses are not a tail event; across a sample of 136 trades, runs of that length should appear repeatedly, and Jack Schwager's interviews in the Market Wizards series return to this point constantly: professionals distinguish between drawdowns that fall inside their system's expected streak behavior and those that break it. This one sits comfortably inside. The 5.51% peak-to-trough dip on the simulated account is likewise a direct mechanical consequence of 2% fixed risk meeting three stops plus an earlier loss, not a structural warning.
The honest caveats run in both directions. A week is a dozen trades; nothing about one window, good or bad, moves the estimate of the system's edge by much, and we resist reading Thursday as either disaster or bargain. What a window can legitimately teach is structural, and this one did: the streak arrived compressed into fifty-two minutes because the entries were correlated, and streak math assumes independence. Correlated entries make expected streaks arrive faster and cut deeper than the textbook arithmetic suggests, which is why the tuning note below is about exposure, not about win rate.
At a Glance
Drawdown questions
Why publish a losses report at all?
Because every legitimate trading operation has losses, and the ones that do not show them are choosing what you see. Drawdown reporting is standard practice for funds precisely because losses carry most of the useful information about how a system behaves under stress: whether stops are honored, whether sizing stays fixed, whether losing streaks fall inside expected ranges. A record with the red removed is not a record.
How normal is a three-loss streak for a system that wins about 60% of the time?
Entirely normal. With roughly four losses in every ten trades, three consecutive losses should appear repeatedly across a 136-trade sample; their absence would be the statistical surprise. The caveat this week adds is independence: the three losses were correlated entries on one thesis, and correlated entries compress expected streaks into shorter windows. The streak length was ordinary. The fifty-two minute delivery was the finding.
What does the 5.51% drawdown figure actually describe?
It is the peak-to-trough dip on the simulated $100,000 account, measured from Wednesday's equity high near $108,976 to the level after Thursday's third stop, with every trade risked at a fixed 2%. It is a mechanical consequence of four -1R losses landing near a peak. At fixed fractional risk, drawdown depth maps directly to losing-run length, which is why streak behavior, not any single loss, is the number worth watching.
Why were all four losses long trades?
Not because the system carries a bullish bias. The week's tape offered mostly pullback-long setups, seven of which won, and the losses came from the same distribution the wins did. Three of the four losses failed together because they expressed one belief about one Thursday morning. Direction was incidental; concentration was the cause, and it is the subject of this report's tuning note.
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: +27.29R YTD across 136 trades, see stats strip.
From the SkyAnalyst Journal
Weekly Recap: The Week the Pullback Paid, Then Charged Tuition
Eleven trades across five instruments, seven wins, four losses, +1.48R net. One entry shape ran through the whole week, paying through Wednesday's four-winner burst and presenting the bill on Thursday morning.
NAS100 Long: The Level Changed Sides, and So Did the Desk
Five days after shorting the Nasdaq off a failed bounce, the desk went long it at a resistance level that had flipped to support. Case study #108 is about a system with no memory of its last opinion.
The GBPUSD Long That Never Spent a Single Pip Underwater
Cable V-reversed 72 pips off the ISM print and the system refused to chase it. Case study #107 is about entry location: a pullback buy that spent nearly 16 hours in the market without a pip of drawdown.