Three Losses, One Habit: Approvals That Ran Ahead of the Candle
Three stops in under a day, exactly 3R given back from a +4.08R Thursday peak. Each approval wrote down, in its own words, that part of its confirmation was sti
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.
This is a short losses report on a week that finished positive. Seven trades, four winners, three losers, and each loser cost exactly the 1R it was sized for. From Tuesday afternoon to Thursday morning the desk took four trades and closed all four as winners. By 15:02 UTC on Thursday the week stood at +4.08R. Then three stops in a row: a Nasdaq short on Thursday afternoon, and a Dow long and a Nasdaq long on Friday morning after a payrolls miss flipped the macro read. The window gave back 3.0R, 6,000 dollars on a simulated 100,000 dollar account at 2% risk, and closed at +1.08R. Against the year to date ledger the scale is plain. The system has banked +38.35R YTD from Jan 12 inception, sealed at the September close, which puts that same simulated account at $176,703.28. The desk gave back 3R this week against +38.35R YTD. What makes the week worth writing up is not the size of the loss. It is that every losing approval described, in writing, a confirmation that was not fully there.
Tuesday to Thursday morning: four for four
The losses only make sense against what came before them. A EURUSD short on Tuesday into a two-month dollar high, a Nasdaq long and a Dow short thirty-four minutes apart on Wednesday, and a second Dow short on Thursday that reached its first target sixteen minutes after it filled.
That last trade is the one to remember. Three evaluations declined it in five minutes because the candle was still pushing into the pocket, and it entered only once the failed reclaim was on the chart. By its first target the simulated account sat at its weekly peak of 108,161.20 dollars.
Thursday afternoon: the short into a bounce
At 15:17 UTC the Nasdaq short book sold NAS100 at 30,402, a minute and seventeen seconds after an evaluation had declined the same trade. Price went 104 points our way, to 30,297.9, then reversed through the 30,495 stop. The position closed at 30,506.8 a little over two hours after it filled.
Friday morning: two longs after payrolls
Payrolls came in at 29K against 89K expected, yields fell, and the macro read turned bullish on the indices. At 14:06 UTC the Dow long book bought US30 at 51,258.6, saw 51,314.5, and was stopped fifty-seven minutes later at 51,160.4. At 14:58 the Nasdaq long book bought NAS100 at 30,929.5, saw 31,009.8, and was stopped at 30,795 fourteen minutes after the fill.
The drawdown curve shows it plainly: four steps up, then three even steps down, each worth 2,000 dollars. A peak to trough of 5.55%, and the balance never fell below where the week began.
Every trade the system took.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Oct 1 | 15:17 UTC | NAS100 | Short | NAS100 SHORT: Pullback to Broken Structure (Primary) | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - | |
| Oct 2 | 14:06 UTC | US30 | Long | US30 LONG reclaim of 51257.8-51288 shelf | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - | |
| Oct 2 | 14:58 UTC | NAS100 | Long | Claude Opus 4.7 | NAS100 LONG: Pullback to Breakout Retest | 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 is not an instrument or a direction. The three losses split across two indices and both sides of the market. The pattern is the approval note.
Every winner this week entered on a trigger that had fully printed. Every loser entered on an approval that named something still missing. The Thursday Nasdaq short was approved with the note that the rejection was not yet confirmed by a clean bearish candle and that the 5-minute MACD histogram had just flipped positive. The Friday Dow long was approved on its first evaluation, with the note that the ideal confirmation candle had not fully closed and that the entry was tradeable but not clean. The Friday Nasdaq long did get its candle close above 30,930, but entered with the 5-minute MACD histogram negative and falling and price slightly below its 5-minute EMA9.
Last week's losses report landed on the same observation, when two of three Thursday shorts were approved ahead of conditions their own plans had set. Two weeks is still a small sample. But five of the last six losses now carry an approval note describing something not yet in place, and the trades that waited are the ones that got paid.
Decision highlights
The Friday Dow long is the decision we would take back. Its plan asked for a 5-minute close back above the 51,257.8 to 51,288 shelf, and the approval came on the very first evaluation with the note that the ideal confirmation candle had not fully closed and that the entry should be sized lighter. The desk sizes every position at 1R, so a note asking for a lighter size had nowhere to go.
The Thursday Nasdaq short changed its mind in seventy-seven seconds. At 15:16 UTC the evaluation declined at 42% confidence because the last closed candle was a strong bullish bounce from 30,285. At 15:17 it approved at 68% with no new closed candle in between, and the stop was hit after the bounce carried on.
The Thursday Dow short is the decision we would repeat, and it is the counterweight to the three losses. Three evaluations in five minutes declined it because the 5-minute candle was still pushing up into the pocket. It entered only after the failed reclaim printed and reached its first target sixteen minutes later.
Six instruments, six stories.
No losses this week. The euro book's only trade was Tuesday's short, which reached its second target.
All EURUSD this week →No losses this week because there were no trades. The Cable book did not find a setup that cleared its gates.
All GBPUSD this week →One loss, the Friday long at 51,258.6 after payrolls. Approved on its first evaluation before the confirmation candle closed, stopped fifty-seven minutes later.
All US30 this week →Two losses, a Thursday short and a Friday long. The short went 104 points our way before stopping; the long lasted fourteen minutes.
All NAS100 this week →No losses this week because there were no trades. The yen book did not produce a qualifying setup.
All USDJPY this week →No losses this week because there were no trades. The S&P book did not produce a fill.
All US500 this week →No losses this week because there were no trades. USDCAD remains the newest book and has not yet produced a qualifying setup.
All USDCAD this week →Loss of the week: NAS100 Long · -1R
Losses worth learning from
NAS100 Long, Friday, -1R, grade C+
This is the loss of the week, and the closest call of the three on the trigger itself.
What was right
The macro read had turned. Payrolls missed, yields fell, and the Trend Agent read the Nasdaq as bullish at 78%. The plan wanted a pullback into 30,895 to 30,930 and a 5-minute close back above 30,930. The completed candle dipped to 30,912.5 and closed at 30,947.4. On the letter of the plan, the trigger was met.
What was wrong
Two and a half minutes earlier the same check had declined, at 52%, because the 5-minute MACD histogram was strongly negative and price was in active selloff mode. At approval the histogram was still negative and still falling, and price sat just below its 5-minute EMA9. The approval wrote both of those down. The position filled at 30,929.5, saw 31,009.8, and stopped at 30,795 fourteen minutes later.
What we'd do the same
The direction and the location. We would take a post-payrolls Nasdaq long at a breakout retest again, with momentum turned.
US30 Long, Friday, -1R, grade C+
What was right
The setup read the morning correctly. Payrolls had missed, yields were falling, and the Dow had tested below a 51,257.8 to 51,288 shelf near VWAP and the prior-day high, then closed a 5-minute candle back above it at 51,290.3. A reclaim long there is a reasonable trade.
What was wrong
The approval came on the first evaluation, with live price already back inside the zone at 51,269. Its own note said the ideal confirmation candle had not fully closed, that 5-minute and 15-minute MACD were still cooling, and that this was a tradeable but not clean entry that should be sized lighter. The fill came at 51,258.6, the stop at 51,174 sat 84.6 points away, and price saw 51,314.5 before stopping fifty-seven minutes later.
What we'd do the same
Reading the payrolls reversal as a long setup. The level was right. The approval was early.
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
The week gave back 3.0R, or 6,000 dollars on the simulated 100,000 dollar account, and still closed at +1.08R, about 2,160 dollars, because of the four winners in front of it. The peak to trough on the week's own equity curve was 5.55%.
The year to date ledger reads +38.35R across 257 trades at a 57.59% win rate, sealed at the September close. On a static 100,000 dollar account at 2% risk that is $176,703.28. Compounding the same trades, where each position is sized off the balance it actually had, gives $203,855.48. The gap between the two, a little over 27,000 dollars, is not extra edge. It is the same R banked through fixed fractional sizing, and it holds only because no single position is ever allowed to be bigger than the rest. This week's 6,000 dollar give-back is under 4% of the static figure, and that sizing is the reason it stays that small.
One note on scope. October is still open, so it is not in the year to date figure. All three of this week's losses fell on Oct 1 and Oct 2, and they join the year to date number when the month closes.
What we're tuning
Last week we wrote that we were looking at making the timing and candle-close conditions in a session plan binding at the approval step, rather than inputs the approver can weigh against a strong directional read. This week added three more cases. All three losing approvals wrote down that part of the confirmation was still missing, and the one trade that sat through repeated declines got paid in sixteen minutes.
We are also looking at two narrower questions. An approval that follows a decline within a couple of minutes, with no new closed candle in between, should have to say what changed. And an approval that asks for a lighter size is asking for something the desk does not do, since every position is sized at 1R. Neither change ships on one week, but both now have two weeks of evidence behind them.
What the numbers actually mean
Four winners followed by three losers feels like a turn. The arithmetic says it is ordinary.
Across 257 trades since inception the system has won 57.59% of the time, for +38.35R, roughly 0.15R per trade. At that win rate about 42% of trades lose, so the chance that any three particular trades in a row all lose is a little under 8%. Over hundreds of trades, three-loss runs are expected many times a year. Last week produced one of the same length.
Van Tharp built the R-multiple around the idea that a system is a distribution of outcomes measured in units of risk, not a win rate. The average winner in this window's panel is 1.02R, drawn from the week's four winners at their first targets. A 1.02R average winner needs a win rate a little under 50% to break even, which is why the system's 57.59% matters more than any single week's split. It is also why a loss that runs past its stop matters: the Nasdaq short closed at 30,506.8 against a 30,495 stop, and slippage of that kind raises the win rate the system needs to stay positive.
Jack Schwager's Market Wizards interviews return to the same point from the other side. The traders who last are not the ones who avoid losing runs, but the ones whose losing runs are uniform and small. Three losses at 1R each, and a 5.55% peak to trough on the simulated account, is that kind of run.
One caveat on sample size. A single week is seven trades, and the drawdown figure in the panel comes from this window's equity walk alone. The 257-trade record and the 57.59% win rate are the numbers with weight behind them. This week is one slice of that record, and it looks like the record.
At a Glance
Drawdown questions
Four winners and then three losers. Did something break on Thursday?
No. The setups, the risk model and the sizing were the same as earlier in the week. What changed was how strictly the approvals held to their own triggers. The winners entered on triggers that had printed. The losers entered on approvals that said part of the confirmation was still missing. Every loss was exactly 1R, the designed size.
Why was the desk short the Nasdaq on Thursday and long on Friday?
Because the inputs changed overnight. Payrolls came in at 29K against 89K expected, yields fell, and the Macro Agent and Trend Agent both turned bullish on the indices. Each book reads that morning's tape, not the previous day's position. The direction on Friday was a fair read. The entries were not as patient as the read deserved.
The Nasdaq short went 104 points our way. Why is it still a full loss?
Because a trade only counts as a win once its first target is actually reached, and the broker closes the whole position there. The short reversed before that and went through its stop. Nearly reaching a target pays nothing, so we score it as the 1R loss it was.
Why does this report say minus 3R when the weekly recap says +1.08R?
They measure different things. This report covers only the losing trades, so it totals 3.0R across three losses. The recap covers all seven trades in the window and nets those losses against four winners. Neither number is the other one adjusted.
Why publish a losses report for a week that finished positive?
Because we publish it whenever a week contains a loss, not only when a week goes badly. The quality gate on this report tried to skip it, since every loss graded C+ and low-grade losses are assumed to teach less. We overrode it, because three approvals that described their own missing confirmation are exactly the kind of loss worth showing.
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: +38.35R YTD across 257 trades, see stats strip.
From the SkyAnalyst Journal
Up 4R by Thursday, Then Three Entries That Did Not Wait
Seven trades, +1.08R. The desk banked four winners from Tuesday to Thursday morning, then gave back 3R on three entries approved on less than full confirmation, two of them on a payrolls Friday.
Waited at 88, Entered at 72: Why SkyAnalyst Sold the Dow on Its Lowest Score
Three reads scored 84, 88 and 82 percent and all said wait. The fourth scored 72 and sold US30 at 50,785. TP1 printed 16 minutes after the fill for +0.87R (TP1).
2026 So Far: +38.35R, and What Nine Months Taught Us About Risk
Two hundred and fifty-seven trades since the January 12 inception, 148 winners, +38.35R net. Three threads ran through the year: how the edge was built, how we learned to see it, and what it is still exposed to.