Three Losses, Three Identical Sizes, and a 3.61 Percent Drawdown
Every loss this week cost exactly 1R. Not approximately, exactly. Two came from the same NAS100 book that also won twice, and the deepest the account got was 3.
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 trades lost money this week, and all three lost exactly the same amount. Minus 1R on Monday's NAS100 short, minus 1R on Thursday's USDCAD short, minus 1R on Friday's NAS100 long. Total: minus 3.00R against a week that still finished at +3.43R net. That identical figure is the point of this report. The system does not attempt to make its losses small, because trying to make a loss small is how a stop gets moved. It makes them the same size, every time, so that the only variable left in the equation is what the winners do. Through Sep 7, 2026, the desk stands at +28.75R across 220 trades since the January 12 inception, at a 56.82% win rate. A $100,000 account risking 2% per trade sits at $157,519 on static sizing, having been 3.61 percent below its peak at the worst point of this week.
Act I: The loss that was taken correctly
Monday's NAS100 short is graded B, the highest grade of the three, and it is the one we would take again.
The setup was the VWAP rejection pattern that had produced repeated winners the previous week, in a macro configuration that still leaned bearish. The entry cleared the gate. The stop was placed against structure. Price went the other way and the position closed at its predetermined level for -1R.
There is no lesson here about the entry, and that is precisely why it leads this report. A system that only loses on bad entries is a system that has not been running long enough.
Act II: The streak of two
Thursday afternoon a USDCAD short stopped out. Friday a NAS100 long stopped out. Two in a row, which was the longest losing streak of the window.
Two consecutive losses is statistically unremarkable at a 56.82% win rate. It is also the point at which a discretionary trader typically starts adjusting something, which is the behaviour the flat-risk rule exists to make impossible. Position size on Friday's trade was calculated exactly as it had been on Monday's, from the same 2% of account, with no memory of the two stop-outs in between.
Act III: 3.61 percent
The simulated account peaked at $110,875.80 during Thursday morning's run of winners and fell to $98,000 at the trough. That is a maximum drawdown of 3.61 percent.
Reported honestly, a 3.61 percent drawdown in a week that finished up is barely an event. Reported dishonestly, by quoting only the closing number, the week looks smooth and the reader learns nothing about what holding the position actually felt like. The curve in this article shows the whole path.
Every trade the system took.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Aug 31 | 14:05 UTC | NAS100 | Short | NAS100 SHORT, VWAP Rejection / Failed Rally Continuation | B | -1.0R(SL) | -$2,000(SL) | Stop hit | - | |
| Sep 3 | 14:41 UTC | USDCAD | Short | USDCAD Short Bearish Continuation Pullback | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - | |
| Sep 4 | 14:17 UTC | NAS100 | Long | Claude Opus 4.7 | NAS100 LONG, Pullback to VWAP / Yesterday's High Support | 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 pattern in this week's losses is that there is no pattern, and that is a finding rather than an absence of one.
Two of the three losses came from NAS100, in opposite directions, in a week where NAS100 also won in both directions. One came from USDCAD, a book with a single September trade to its name. The three losses share no setup type, no session, no direction and no macro condition.
Compare that to August, where the losses did have a pattern: the GBPUSD long book took ten trades in one direction, won two, and gave back 6.18R. That is a book on the wrong side of a regime, and it needed switching off, which is what the direction split on August 27 made possible. This week's losses look nothing like it. They look like the ordinary cost of participating.
The distinction matters because the two situations demand opposite responses. One requires intervention. The other requires leaving the system alone, which is harder.
Decision highlights
We held risk at a flat 2% through two consecutive losses. The point at which most discretionary traders reduce or increase size is exactly the point where a mechanical desk is required to do neither. Friday's position was sized from the same rule as Monday's, with no adjustment for what had happened in between.
We published Monday's NAS100 short as the most instructive loss of the week despite it being a textbook entry. The temptation with a loss report is to select the trades with an obvious flaw, because those come with a tidy lesson attached. The more useful publication is the well-formed trade that lost anyway, since that is the one that describes what the other 43% of trades look like.
We changed nothing in response to any of the three losses. No threshold was tightened, no book was pulled, no allocation was moved. That is a decision rather than an omission, and the standard for acting is a pattern across a sample, not a stop-out on a book that won twice in the same week.
Six instruments, six stories.
EURUSD lost nothing this week. It entered a short on Thursday at 1.1616 with a stop at 1.1629 that has not yet resolved, so it appears in no figure in this report and will be counted in the week it closes.
All EURUSD this week →GBPUSD lost nothing this week. Its single trade was a winning short, and the long book that cost 6.18R across August remains offline following the direction split.
All GBPUSD this week →US30 lost nothing this week. Its single trade was a winning short taken on Tuesday that cleared the gate on one evaluation.
All US30 this week →NAS100 accounts for two of the week's three losses, one short on Monday and one long on Friday, for -2R combined. The same book also won twice in the same week, which is why neither loss triggers any change to it.
All NAS100 this week →USDJPY did not trade this week, so it lost nothing. The book has taken six trades all year and remains deliberately on a short leash.
All USDJPY this week →US500 lost nothing this week. Its single trade was the best result of the week, a short for +2.14R on the recap baseline.
All US500 this week →USDCAD accounts for the third loss, a short on Thursday for -1R. It is the book's only September trade and one stop-out is not a sample.
All USDCAD this week →Loss of the week: NAS100 Long · -1R
Losses worth learning from
NAS100 Short, Monday, -1R, grade B
The highest-graded loss of the week and the one we would take again unchanged.
The entry used the same VWAP rejection structure that had produced multiple winners the previous week, in a macro configuration that still leaned bearish going into the session. The gate cleared, the Risk Agent sized against structure, and the stop sat where the idea would be invalidated rather than at a convenient distance.
Price traded through it. The position closed at -1R exactly as designed.
What makes this a grade B loss rather than a lower one is that every step of it was correct except the outcome. There is no threshold to tighten here, no filter that would have caught it without also filtering out the winners the same pattern produced days earlier. The honest teardown of a well-formed losing trade is that it was well-formed and it lost, and a track record that never contains one of these is a track record that is hiding something.
NAS100 Long, Friday, -1R, grade C+
The lower grade reflects a genuinely weaker setup, and the interesting part is the twenty-four hours before it.
Thursday morning the same NAS100 long book entered on a Fibonacci pullback continuation and ran to its third target for +2.84R, taken over the Macro Agent's objection because the live rate tape contradicted the macro read. Friday's entry was a different configuration on the same book and it stopped out for -1R.
Nothing about the book changed between those two trades. The tape did. The grade difference, B to C+, is the system's own record of that: Friday's confluence was thinner going in, and it is scored that way in the log before the outcome was known rather than after.
The lesson is not that the Friday trade should have been skipped. It is that a book which wins on Thursday and loses on Friday is behaving normally, and the correct response to the pair is nothing at all.
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 | -3R | −$6,000 |
All six agents combined.
From the desk
From the desk
We publish the losses on the same schedule as the wins, in the same detail, with the same numbers attached. This week that means three stop-outs for a combined -3.00R, and a maximum drawdown of 3.61 percent against a simulated $100,000 account.
Through Sep 7, 2026, the desk stands at +28.75R across 220 trades since the January 12 inception, 125 winners against 95 losers. On a $100,000 account at 2% risk, static sizing puts that at $157,519, and compounding the same sequence puts it at $169,638. Ninety-five of those 220 trades cost exactly 1R each, and that is the number most performance marketing leaves out.
The reason to publish this report every week is not transparency as a virtue signal. It is that a reader deciding whether this system is usable needs the drawdown path, not just the endpoint. A week that finishes at +3.43R after being 3.61 percent below its peak is a different experience from a week that climbs steadily to the same total, and only one of those two descriptions tells you what you would have had to sit through.
One position remains open from Thursday, a EURUSD short at 1.1616 with a stop at 1.1629. It is in none of these figures. We will report it in the week it closes, in whichever direction it goes.
What we're tuning
Nothing changes on the back of this week's losses, and that is the finding rather than a failure to produce one. Three losses at exactly 1R each, spread across two books that also won, in a week that finished up, is the system operating inside its designed distribution. The threshold for acting is a repeated one-sided failure across a meaningful sample, which is what the August GBPUSD long book produced and what nothing here resembles.
The one item under observation is the EURUSD short still open from Thursday, entered at 1.1616 against a stop at 1.1629. It is excluded from every figure in this report. When it resolves it will be counted in the week it closes, and if it changes anything about how that book is handled we will publish that reasoning then rather than speculating about it now.
What the numbers actually mean
A 56.82% win rate across 220 trades means roughly 95 losses so far this year, and it means losing streaks are not a possibility to be managed but a certainty to be planned for.
The arithmetic is unforgiving and worth stating explicitly. At a 56.82% strike rate, the probability of two consecutive losses on any given pair of trades is about 18.6 percent. Three in a row is about 8 percent. Across a 220-trade year, streaks of four and five are not anomalies, they are expected occurrences that will happen several times. This week produced a streak of two, which at this win rate should occur roughly once every five or six trades. It arrived exactly on schedule.
This is the point Van Tharp built his work on R-multiples around: the size of the individual loss is a design decision, not a market outcome. Once every loss is fixed at 1R, the profitability question reduces to whether the average winner exceeds the average loser by enough to cover the strike rate. Our current R target sits at 1.29, meaning the average winning trade needs to return about 1.29R for the system to hold its edge at this win rate. Anything above that compounds; anything below it bleeds regardless of how often the desk is right.
Jack Schwager's interviews across three decades of Market Wizards keep returning to the same observation from otherwise unrelated traders: the drawdown distribution, not the return, is what determines whether a strategy survives contact with a real account. A 3.61 percent drawdown of the kind this week produced is inside the ordinary range for a system with these parameters. It is also, importantly, the number a reader needs in order to judge whether they could actually hold the position through it. A return figure quoted without its drawdown is a number without a cost attached, and the cost is the part that determines whether anyone can stay in the seat long enough to collect the return.
At a Glance
Drawdown questions
Why publish a losses report at all?
Because a track record that only contains winners is marketing rather than a record. Ninety-five of the 220 trades this year have lost money, each costing exactly 1R, and a reader deciding whether the system is usable needs that distribution and the drawdown path that comes with it, not only the closing total.
Why did every loss cost exactly 1R?
Because position size is calculated so that the distance from entry to stop equals a fixed 2% of the account, whatever the instrument. A NAS100 stop 78.9 points wide and a USDCAD stop in a different price scale represent identical risk. The system does not try to make losses small, it makes them the same, so the only remaining variable is what the winners return.
Is a 3.61 percent drawdown normal?
For a system with these parameters, yes. At a 56.82% win rate, two consecutive losses should occur roughly once every five or six trades and streaks of four or five are expected several times across a 220-trade year. This week produced a streak of two. The drawdown that came with it is inside the ordinary range rather than an outlier.
The same NAS100 book won twice and lost twice. Should it be switched off?
No. A book winning and losing in both directions across one week is ordinary variance, and the sample is four trades. The direction split exists to catch a book losing repeatedly on one side, which is what the GBPUSD long book did across ten August trades for -6.18R. Nothing in the NAS100 record resembles that.
What happens to the open EURUSD trade?
It is excluded from every number in this report and will be counted in full in the week it resolves, whichever way it goes. Counting an open position at its current mark would let an unrealized figure flatter the week and would require restating the total if it reversed.
Trade with the system that publishes its drawdowns.
Subscribers receive every signal, winners and losers alike, 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: +28.75R YTD across 220 trades, see stats strip.
From the SkyAnalyst Journal
Five Straight Wins, Two Losses Back, and +3.43R on the Week
Eight trades, five winners, +3.43R. The week opened with a loss, ran five in a row, then gave two back. NAS100 traded four times in both directions and finished the week almost exactly flat.
Twelve of Thirteen: The Long Book Answers Back at TP3
The desk has now won 12 of its last 13 trades. This one was a NAS100 long, taken against its own Macro Agent, and it ran the full distance to the third target in 53 minutes for +2.84R.
The Cable Short That Entered on Its Lowest Confidence Reading
Four refusals at 79, 81, 82 and 84 percent. Then the system entered at 62, the weakest score of the sequence, and took TP1 for +1.02R. The number was never the trigger.