Two closed trades all week, both C+ longs, both stopped inside six minutes of entry. The desk gave back 2R against a +19.60R YTD line, and a third entry was sti
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.
Through Jun 15, 2026, the desk has banked +19.60R YTD across 109 trades from Jan 12 inception. A $100,000 simulated account at 2 percent risk per trade sits at $139,204.12 on the static path, or $144,227.48 on the compounded path. This week the desk gave 2R of that figure back, and it did it in under eleven minutes of combined market exposure. Only two setups cleared the confluence gate in five sessions. Both were longs, both graded C+, and both stopped inside six minutes of entry: a GBPUSD second-chance read at Wednesday's post-CPI VWAP and Fibonacci cluster, and a NAS100 post-reversal continuation on Friday afternoon. A third entry, a GBPUSD pullback long filled at 15:46 UTC Friday, was still open when this report went to press. It is not in this week's numbers; it will be counted, at whatever it resolves to, in the window it closes. What follows is the teardown of the two trades that are.
Monday and Tuesday produced no entries. The Trend Agent flagged candidates on both sessions, but nothing cleared the confluence threshold, and the desk does not trade to fill a calendar. At a pace of roughly five trades a week across the YTD ledger, a two-session drought is unusual but not unprecedented; the gate is doing exactly what it is designed to do when the tape does not offer a read. The cost of a confluence-gated system is quiet stretches like this. The benefit is supposed to be that the entries that do clear are worth taking. This week tested that proposition and the answer was uncomfortable.
The first entry of the week filled at 15:13 UTC on Wednesday: a GBPUSD long at 1.33952, a second-chance read at a VWAP and Fibonacci cluster in the post-CPI session, graded C+ on confluence. The stop sat at 1.3390, just over five pips below entry, at the structural level the read was built on. Five minutes and thirty-six seconds later the stop was hit. The simulated curve stepped from $100,000 to $98,000 on the reference account. Thursday passed without another entry. A flat 1R, taken exactly as sized, and the fastest stop-out on the desk's books in weeks.
Friday at 14:40 UTC the NAS100 trader went long at 29,519 on a post-reversal continuation read, stop at 29,400 at the reversal structure. Four minutes and forty seconds later that stop was hit too, and the window's equity walk printed its 4 percent trough at $96,000. Then, at 15:46 UTC, the GBPUSD trader filled a second long on the pair, a VWAP and EMA pullback read at 1.34112, just over 48 hours after Wednesday's GBPUSD stop, in the same direction. That position was still open when this report went to press. It is not counted in this window's loss line, and we want to be plain about why: an open trade has no realized outcome, and a desk that books unrealized positions into a published weekly record is writing fiction. It will appear in the window it closes, at whatever it closes at.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Jun 10 | 15:13 UTC | GBPUSD | Long | Claude Opus 4.7 | GBPUSD Post-CPI Second-Chance at VWAP/Fib Cluster | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Jun 12 | 14:40 UTC | NAS100 | Long | Claude Opus 4.7 | NAS100 Long - Post-Reversal Continuation | 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.
The pattern this week is entries that never got past hello. Both losses stopped inside six minutes of their fills. That is not a coincidence of two unlucky prints; it says something specific about where the entries sat relative to the tape. A stop that is hit that fast was placed inside the market's immediate noise band, which means either the stop was too close to entry for the session's volatility, or the entry triggered before the move it was betting on had confirmed.
Last week's recap named a meta-pattern we keep coming back to: every winning trade in that window entered after a confirmed second leg with volume, and every losing trade entered into a premise the tape had not yet proven. This week extends the losing half of that observation. The Wednesday cable long was a second-chance read, an entry built on the idea that the first move would re-offer itself; it never did. The Friday NAS100 long was a continuation read taken before the continuation printed. Two trades, both betting on follow-through that had not yet appeared, both invalidated almost immediately. The honest read is that on a two-trade sample we cannot call this signal. But it rhymes with last week, and patterns that rhyme across windows are the ones that earn a line in the tuning queue.
The desk stood down on Monday, Tuesday, and Thursday. Three of five sessions produced no entry because no read cleared the confluence threshold, and nobody overrode the gate to manufacture activity. On a week where both gated entries stopped, it is worth saying plainly that the decision the system got most right was every trade it did not take.
The Risk Agent held flat 1R sizing on Friday's NAS100 entry despite the week already sitting 1R down from Wednesday. No revenge sizing after the first stop, no hesitation discount either: the second trade was sized exactly as the first. That symmetry is the policy working, and it is why the week's give-back is 2R and not something that requires a recovery narrative.
The Friday GBPUSD re-entry at 15:46 UTC is the decision we cannot fully grade yet. It filled just over 48 hours after Wednesday's GBPUSD stop, in the same direction, which is precisely the shape the 72-hour cooldown gate proposed in last week's report is designed to scrutinize. That gate is still in back-test and was not live at entry. The position was open at press time; whichever way it resolves, it lands in next week's books and we will write it up against the gate's logic, not around it.
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.
EURUSD: no losses this window. The EURUSD trader did not size an entry; nothing on the pair cleared the confluence threshold in five sessions.
All EURUSD this week →GBPUSD: one loss. Wednesday's post-CPI second-chance long at the VWAP and Fibonacci cluster stopped in five minutes and thirty-six seconds for a flat 1R. A second GBPUSD long, filled Friday at 15:46 UTC, was still open at press time and is not counted in this window's numbers.
All GBPUSD this week →US30: no losses this window. The US30 trader flagged no setup that cleared the gate; the index sat outside our entry criteria all week.
All US30 this week →NAS100: one loss. Friday's post-reversal continuation long stopped in four minutes and forty seconds for a flat 1R, printing the window's 4 percent equity trough. The continuation the read was priced on never confirmed.
All NAS100 this week →USDJPY: no losses this window. After last week's stopped pullback long, the pair offered no read that cleared the threshold, and the trader stood down.
All USDJPY this week →US500: no losses this window. No entry was sized; the index never fit the setup criteria this week.
All US500 this week →Loss of the week: NAS100 Long · -1R
The location. Wednesday's GBPUSD long at 15:13 UTC was built on a VWAP and Fibonacci cluster in the post-CPI session, and a C+ grade priced the conviction honestly: a structural level worth a flat 1R, nothing more. Sizing followed the grade.
The stop distance. The stop sat at 1.3390, just over five pips below the 1.33952 entry, in a session still digesting a CPI print. Five pips on cable in post-data tape is inside the noise band, and the market took it in five minutes and thirty-six seconds without ever testing the actual thesis. The read was never really in the trade.
Take a C+ second-chance read at flat 1R when it clears the gate. The grade and the size were right.
Enforce a stop-width floor keyed to the session's realized volatility. If the structural stop has to sit five pips away in post-CPI tape, the entry waits for the level to be retested at calmer volatility, or the trade passes.
The structure. Friday's NAS100 long at 29,519 placed its stop at 29,400, at the reversal structure the read was built from, roughly 0.4 percent below entry. That is a defensible structural stop, not a crowded one, and the C+ grade matched a continuation premise that had confluence but not confirmation.
The timing. A continuation read sized before the continuation printed. The entry triggered into what turned out to be the next leg of the pullback, and the stop was hit in four minutes and forty seconds. Last week's window taught that our winners enter after a confirmed second leg with volume; this entry did not wait for one.
The stop placement and the size. Structure-based stops that fail fast are cheap information; this loss cost 1R and ten minutes, not a slow bleed.
Require the confirmed second leg before sizing continuation entries. The premise can be right and the trade still early, and early is just wrong with better grammar.
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 | -2R | −$4,000 |
A $100,000 simulated account at 2 percent risk per trade entered this week at $143,204.12 on the static path, or $150,174.39 on the compounded path. It exits at $139,204.12 static and $144,227.48 compounded. This week's give-back is exactly $4,000 on the static figure and approximately $5,947 on the compounded one. The compounded path loses more dollars in a drawdown because it risks 2 percent of a larger balance; that is the same property that has it sitting $5,023.36 above the static path through Jun 15, 2026. Compounding cuts both ways, and a desk that only shows you the way it cuts upward is not showing you the math.
We also owe the reader plainness about the position this report does not count. A GBPUSD pullback long filled Friday at 15:46 UTC and was still open when we went to press. It risks the same flat 1R as everything else on the books. If it stops, next week's window opens 1R down and we will write that. If it pays, it pays into next week's recap and we will write that instead. What we will not do is book an unrealized position into a published record, in either direction, because the whole value of this series is that the numbers are settled facts.
Two trades, ten minutes of exposure, 2R back to the market. The system's edge has never lived in any single week, and a window this thin is mostly a test of whether the process holds when there is nothing to do. It held: the gate stayed shut for three sessions, the sizing stayed flat, the stops were honored at the prices they were set. The reads themselves were early, and the tuning section says what we are doing about it. That is the whole story, and it is enough of one to publish, because the weeks that are tempting to skip are exactly the ones a public ledger exists for.
Two items this cycle. First, a status update we owe from last week: the per-instrument cooldown gate, which downgrades a same-instrument, same-direction re-entry inside 72 hours of a stop by half a grade, is still in back-test against the YTD ledger. This week handed it a live specimen: Friday's GBPUSD pullback long filled 48 hours after Wednesday's GBPUSD stop, in the same direction, at a C+ grade the gate would have cut to C. Whether a C read still clears the sizing threshold is exactly the question the back-test answers. It ships, or we publish why it does not, next week, against the criteria already on the record.
Second, a new item from this week's tape: a stop-width floor keyed to session volatility. Both of this week's stops were hit inside six minutes, and the Wednesday cable stop sat five pips from entry in post-CPI conditions. A stop placed inside the session's noise band converts a structural read into a coin flip on the next tick cluster. The proposed check is mechanical: if the structural stop distance falls below a floor derived from the session's realized volatility, the entry either waits for a retest at calmer tape or passes. It is testable on the historical ledger the same way the cooldown gate is, and it goes into the same queue.
Start with what a two-trade week can and cannot tell you. It cannot tell you anything about expectancy. Van Tharp's framework in `Trade Your Way to Financial Freedom` grades a system on expectancy, win rate times average win minus loss rate times average loss, and expectancy is a property of a distribution, not of a week. The distribution that matters here is the YTD sample: 109 trades, a 57.8 percent win rate, +19.60R net. Two stops at flat 1R move that ledger by exactly 2R and change nothing about the underlying math.
What a short week can tell you is whether the system behaves under scarcity, and the data says it did: flat sizing on both entries, stops taken where they were placed, no override of the gate to generate activity. A zero-winner week is also less rare than it feels. Back-of-envelope, at a 57.8 percent win rate and assuming independence, two trades both losing is roughly an 18 percent event, something like one week in five at this cadence. Jack Schwager's `Market Wizards` series catalogs streak distributions for systems with win rates in this range, and runs far longer than two appear on the historical math several times a year. A two-loss streak is not a streak; it is Tuesday.
A note on the panel above. Several of its fields derive from the window itself when the rolling snapshot is missing a value, and a two-trade, zero-winner window produces degenerate readings: the R target reads zero because there were no window winners to average, and the drawdown figure is the 4 percent equity walk of two consecutive 2 percent stops on the reference account. Read those cells as a description of one quiet week, not of the system. On sizing, the Kelly criterion sets the conceptual ceiling for risk per trade. We do not publish a computed Kelly fraction, because the inputs drift on a system this young, but flat 2 percent risk sits well below any reasonable estimate of it. That is why a 2R week subtracts $4,000 from the simulated static path instead of becoming something that needs a recovery plan.
Because the publish rule is mechanical, not editorial: any week with at least one realized loss gets a losses article. The alternative is a desk that quietly decides which losing weeks are interesting enough to disclose, and that is a marketing filter wearing a research costume. Two trades is a thin week. Thin weeks are part of the record.
Because it has no realized outcome. An open position is a risk, not a result, and booking unrealized trades into a published weekly ledger would let us flatter or punish any given week depending on when we hit publish. The trade will be counted in the window it actually closes, at whatever R it closes at.
Not at this sample size. At the YTD win rate of 57.8 percent, and assuming trade independence as a rough approximation, both of two trades losing is about an 18 percent outcome, roughly one week in five at this cadence. The number that would constitute a flag is a drift in the win rate or average R across hundreds of trades, and the YTD ledger at 109 trades and +19.60R shows no such drift.
It is still in back-test against the YTD ledger and was not live this week. Friday's GBPUSD re-entry, filled 48 hours after Wednesday's stop on the same pair in the same direction, is exactly the shape the gate targets, and it will be part of the evidence either way. The ship-or-kill criteria have not changed: if the gate avoids more same-pair re-entry losses than the winners it would have vetoed, it ships.
For Wednesday's cable trade, probably yes: five pips in a post-CPI session is inside the market's noise band, and the thesis never got tested. For Friday's NAS100 trade the stop sat at real structure 119 points away, and the speed says more about entry timing than stop placement. The distinction matters, which is why the tuning queue has a volatility-keyed stop-width floor rather than a blanket "wider stops" rule.
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: +19.60R YTD across 109 trades, see stats strip.
The gate cleared two setups in five sessions and both stopped inside six minutes of entry. Net minus 2R against +19.60R YTD, and a GBPUSD pullback long entered Friday afternoon was still open at press time.
Seven trades. Three losses inside the dollar bid that broke our Monday open and Wednesday cable. One bounce-rejection short on NAS100 that paid the week. A 57.1 percent win rate that owes most of its margin to one outlier.
Three losses, two of them on the same instrument, same direction, forty-six hours apart. The desk gave back 2.5R against a +21.60R YTD line, then closed the week at a fresh equity peak.