Seven losses this week, and why they are normal even for an AI
The second losing week in a row. At a 59% win rate, a run of losses like this is not a malfunction, it is a statistically expected event. The year still stands at +31.43R.
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 the honest one. For the second week running, the desk lost money: seven trades stopped out between August 10 and 16, the losses briefly ran three in a row, and the system's drawdown from its peak deepened to about 9%. We are publishing all seven losses, because the weeks that test a system are the weeks worth reading, and because no trading operation, no matter how sophisticated its AI, gets to pretend drawdowns do not happen. Here is the context before the teardowns. Since the January 12 inception, the system has banked +31.43R, and a simulated $100,000 account at 2% risk sits at $162,869 on a static basis. This week's seven losses cost 7R gross. Two down weeks in August have pulled the drawdown to 9%, and that number, uncomfortable as it feels, is squarely inside the range of normal for a system that wins 59% of its trades. Drawdowns of 10% or more are a routine feature of any real edge, not evidence that the edge is gone. The losses were not random, and they were not sloppy. Five of the seven were long entries into a choppy, mean-reverting tape that would not deliver the follow-through those setups need. The other two were shorts that also failed to extend. Every one of them entered on a qualifying read and closed at its predefined stop for exactly minus 1R. The teardowns below show three of these losses in full, and the statistics section explains why a run like this, in a system with our win rate, is not just survivable but expected. Markets move in cycles. So does any system that trades them honestly.
The shorts that would not extend
The week's losses opened with a pair of failures on the short side. A US30 short on Monday and a EURUSD short on Tuesday both entered on reasonable reads and both stopped when the market refused to extend lower. These were not the classic losing-week culprits; they were shorts in a market that, it turned out, did not want to trend in either direction. That was the first hint of the regime we were in: a chop that punished conviction on both sides.
The longs took over the losses
By midweek the losing trades had shifted to the long side, and that is where most of the damage came from. A GBPUSD long on Wednesday stopped, and then Thursday brought the cluster: three longs, a GBPUSD, a NAS100, and a EURUSD, all entered and all stopped in a single session. Five of the week's seven losses were longs that failed to hold their levels. The tape was mean-reverting, and every entry built on continuation was on the wrong side of it.
The drawdown deepened
Thursday's three-loss session pushed the week firmly negative and deepened the system's drawdown from its peak to about 9%, up from 7% the week before. Two consecutive down weeks will do that. But the shape of the drawdown matters as much as its depth: seven losses, each capped at exactly minus 1R, with no single trade allowed to run past its stop. A 9% drawdown built out of clean, uniform 1R losses is a controlled retreat, not a rout. That distinction is the whole difference between a survivable drawdown and a damaging one.
Every trade the system took.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Aug 10 | 14:22 UTC | US30 | Short | GPT-5.5 | US30 SHORT | B | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 10 | 14:37 UTC | NAS100 | Long | Claude Opus 4.7 | NAS100 LONG — VWAP Reclaim Continuation | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 11 | 15:48 UTC | EURUSD | Short | Claude Opus 4.7 | Short EURUSD — VWAP Rejection | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 13 | 14:03 UTC | GBPUSD | Long | GPT-5.5 | GBPUSD buy-the-retracement / reclaim long | B | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 14 | 14:02 UTC | GBPUSD | Long | GPT-5.5 | GBPUSD Pullback Continuation Long | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 14 | 14:17 UTC | NAS100 | Long | Claude Opus 4.7 | NAS100 LONG — Bullish Pullback to Fibonacci/VWAP Support | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 14 | 14:34 UTC | EURUSD | Long | Claude Opus 4.7 | EURUSD LONG — Pullback Buy into Consolidation 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
A tape that punished conviction
The pattern was a mean-reverting market that would not reward follow-through in either direction, though it hurt the long side most. Five of the seven losses were longs that entered on valid continuation or pullback setups and stopped when the expected extension never came. The two shorts that lost failed the same way in reverse: they could not push lower. This was not a trending market that we fought; it was a choppy one that faded strength wherever it appeared.
The instructive detail is the US30 short that stopped on Monday. Three days later, the same instrument produced one of the week's cleanest winning longs. That is not a contradiction; it is the signature of a regime with no sustained direction, where the same instrument can stop you short and then pay you long inside a single week. When the tape is like that, a system built to catch continuation will have a low hit rate almost by definition, because continuation is exactly what a mean-reverting market withholds. The correct response is not to re-engineer the setups. It is to keep taking the valid ones at the same small risk and let the hit rate normalize when the regime turns, which it always eventually does.
Decision highlights
We kept every one of seven losses at exactly minus 1R. Through a second consecutive down week, not a single stop was widened, no position was averaged down, and no loss was held past its level. Seven clean 1R stops is what kept the drawdown at a controlled 9% rather than letting it spiral, and it is the single most important thing a system does in a losing stretch.
We did not stop taking long setups after the longs began failing. After five losing longs it would have been tempting to simply refuse the next one, but that is how a system misses the trade that signals the regime has turned. Each long was valid on its own read; the market disagreed with most of them this week. Honoring the process through the drawdown is what keeps us positioned for the recovery.
We took the same US30 setup that lost as a short and later won as a long, without letting the first outcome bias the second. A system that flinches from an instrument after one loss on it would have skipped the winning long three days later. Reading each setup on its own merits, independent of the last result, is the discipline that a directionless regime demands.
Six instruments, six stories.
EURUSD: Minus 2R across two losing trades, a failed short and a failed long. The euro would not extend in either direction, which is the defining trait of the week's mean-reverting tape.
All EURUSD this week →GBPUSD: Minus 2R across two losing longs, both stopped when the expected continuation failed to arrive. Cable's pullback and reclaim setups did not hold in a market that faded strength.
All GBPUSD this week →US30: A single minus 1R short, stopped Monday when the Dow refused to extend lower. Notably, the same instrument produced a clean winning long three days later, a marker of the directionless regime.
All US30 this week →NAS100: Minus 2R across two losing longs. Its two shorts elsewhere in the week were winners, so on the long side specifically the index was on the wrong side of a mean-reverting tape.
All NAS100 this week →USDJPY: No losses this window. The yen sat outside our setup criteria and contributed nothing in either direction.
All USDJPY this week →US500: No losses this window. The S&P book found no qualifying setup and stayed flat, its own quiet contribution on a losing week.
All US500 this week →Loss of the week: EURUSD Long · -1R
Losses worth learning from
US30 short, the failure to extend
What was right. A grade-B short entered at 54,002.5 with a defined stop at 54,086, about 83 points of risk, on a read that the Dow would roll lower. The entry followed the process and the stop was placed at the level that would prove the thesis wrong.
What was wrong. The market did not extend down. US30 refused to trend lower, chopped, and stopped the short at its predefined level. In hindsight this was the first sign of a directionless regime, the same instrument would produce a clean winning long three days later.
What we'd do the same. Honor the stop and keep the loss at 1R. The read was reasonable and the risk was defined; the market simply did not cooperate. Taking the clean minus 1R and moving on is the correct outcome for a setup the tape declined to reward.
GBPUSD long, the reclaim that did not hold
What was right. A grade-B buy-the-retracement and reclaim long entered at 1.3499 with a tight stop at 1.3483, just 16 pips of risk. The setup was a legitimate read of a reclaim, and the tight stop kept the cost of being wrong small.
What was wrong. The reclaim failed to hold. Price could not sustain above the level and stopped the long inside its evaluation. It was the first of the week's long failures, the pattern that would repeat through Thursday as the mean-reverting tape faded every continuation attempt.
What we'd do the same. Keep the stop tight and take the loss cleanly. Sixteen pips of risk on a failed reclaim is a cheap way to be wrong, and the tight stop is exactly why this loss cost 1R and nothing more.
EURUSD long, Thursday's stop
What was right. A pullback-buy into consolidation support, entered at 1.15797 with a stop at 1.1564, about 16 pips of risk. A valid setup on its own read, sized correctly for a 1R loss, taken on a level that had structure behind it.
What was wrong. The support did not hold, and this was the third long to stop on Thursday alone. The setup was grade C-plus, adequate rather than premium, in a week when the mean-reverting tape was rewarding almost no long continuation. The read was fine; the regime was against it.
What we'd do the same. Take the valid setup and cap the loss at 1R. We do not skip qualifying setups because the last several failed; that is how you miss the turn. The loss was clean and small, which is all we ask of a trade the market votes against.
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 | -7R | −$14,000 |
All six agents combined.
From the desk
From the desk
This is the report that tells you whether a track record is real, so read it closely. We are down for the second week in a row, the drawdown has deepened to 9%, and we are stating that plainly rather than burying it. No trading system avoids stretches like this. The marketing version of an AI trading desk never draws down; the honest version draws down, reports it in full, and shows you that every loss stayed capped at exactly its unit of risk. Not even the most sophisticated system, human or machine, gets to opt out of drawdowns. They are the toll for being in the market at all.
The context holds it together. A $100,000 simulated account at 2% risk sits at $162,869 on a static basis since inception. Compound those same returns, letting each 2% risk grow with the account, and the figure is $180,579 instead, a gap of more than $17,000 that comes entirely from disciplined, consistent sizing rather than from bigger bets. Two down weeks in August have not touched the +31.43R year those figures rest on. That is the entire point of a cycle: the winning stretches build the number, the losing stretches test whether you will hold your discipline, and holding it is what keeps the number from being given back. The year is +31.43R. The drawdown is 9%. Both are true, both are normal, and reporting both honestly is the job.
What we're tuning
There is nothing structural to tune out of this week, and pretending otherwise to look responsive would be dishonest. Seven clean stops on qualifying setups, in a mean-reverting tape that punished continuation, is variance and regime, not a flaw. Every loss cost exactly 1R, no single trade did outsized damage, and the drawdown stayed controlled at 9%. When the losses share no error of execution, there is no execution error to fix.
What we are watching is the regime. Two losing weeks with continuation setups failing is the signature of a choppy, directionless market, and while we will not stop taking valid setups, we will stay patient and keep sizing honestly until the tape starts rewarding follow-through again. The tuning is to conviction and patience through the cycle, not to any parameter. Cycles turn. The discipline is to still be standing, intact and small in our losses, when this one does.
What the numbers actually mean
The number that feels alarming this week is the drawdown: about 9% from the system's peak, deepened by a second consecutive losing week. The statistics say it is nothing of the sort, and understanding why is the difference between panicking through a drawdown and riding it out.
Our system wins about 59% of its trades, which means it loses about 41% of them. A win rate like that produces losing streaks and drawdowns as a matter of simple probability, not as a symptom of anything breaking. As Van Tharp explains in his work on R-multiples, the individual trade is close to a coin flip; the edge lives in the distribution of many trades, and that distribution guarantees uncomfortable stretches along the way. A system that never drew down 10% would be a system taking too little risk to compound at all. The drawdowns are not the price of a flawed edge. They are the price of having one.
Jack Schwager's interviews with the best traders alive return to this point again and again: world-class track records are riddled with drawdowns that would frighten an amateur out of the game. A 9% drawdown, a three-loss streak, a 36% hit rate over a single week, these are routine features of any system that wins between half and two-thirds of its trades. The mistake is never the drawdown itself; it is changing a working process in the middle of one. The correct question is not "why are we down 9%," it is "did each loss cost only what it was supposed to." Ours did: seven losses, seven R, each capped at the predefined stop.
The system's win rate across 176 trades remains 59%, with an average winning trade around 1.05R. Those are the numbers of a healthy system in a normal drawdown, not a broken one in trouble. A week is five to fifteen trades, far too small a sample to say anything about an edge that plays out over hundreds. This drawdown is a dip inside a much longer, much more positive line, and the line is what matters.
At a Glance
Drawdown questions
Is a 9% drawdown a sign the system is failing?
No. A drawdown of 10% or more is a routine feature of any trading system that wins between half and two-thirds of its trades. At a 59% win rate, losing streaks and drawdowns are guaranteed by simple probability, not caused by a flaw. What matters is that every loss stayed capped at minus 1R and the year remains at +31.43R. A drawdown inside a strongly positive year is the edge working through a normal cycle, not breaking.
Can an AI trading system avoid losing weeks and drawdowns?
No, and any system claiming to is not being honest. An AI can hold discipline more consistently and process more data than a human, but it cannot turn a losing trade into a winner or force the market to reward a setup it is fading. Losses and drawdowns are permanent features of professional trading. The edge is never in avoiding them; it is in keeping each loss small and letting the winners, over hundreds of trades, add up to more.
Why were most of the losses long trades?
Because the market was mean-reverting and choppy, which is exactly the regime that punishes long continuation and pullback setups. Five of the seven losses were longs that entered on valid reads and stopped when the expected follow-through never came. The same US30 instrument that stopped a short on Monday paid a clean long three days later, a marker of a directionless tape. The setups were sound; the regime was against them.
Why not stop trading until the drawdown ends?
Because you cannot know in advance which trade signals the regime has turned. Stopping after a run of losses is how a system misses the exact winner that begins the recovery. Each of this week's setups was valid on its own read; the market disagreed with most of them. The disciplined response is to keep taking valid setups at the same small risk and wait for the cycle to turn, not to sit out and guess when it will.
How does this drawdown compare to the year?
The drawdown is about 9% from the system's peak, built from two consecutive down weeks in August. Against that, the system has banked +31.43R since inception, with a simulated $100,000 account at 2% risk sitting at $162,869 static. The drawdown is a real but shallow dip inside a strongly positive year. The August give-back will show in the headline numbers as the month closes and is recorded; we freeze the year-to-date to closed months so it never bounces on an open week.
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: +31.43R YTD across 176 trades, see stats strip.
From the SkyAnalyst Journal
Eight losses, five on one pair, and why that is normal trading
Most of this week's losses came from a single instrument that would not cooperate. GBPUSD alone accounted for five of eight stops. Concentrated losing streaks in one instrument are a normal feature of professional trading, not a fault.
US30 went three for three, and the week still finished red
Six winning case studies and a perfect week from US30 were not enough. GBPUSD gave back more than the rest of the desk earned, and the week closed down 3.04R, the third losing week in a row.
NAS100 Short: Selling the Failed Bounce When Yields Turned Hostile
A ~223 point gap down, the 10Y at 4.698%, and a dead-cat bounce into Fibonacci resistance. One evaluation at 75% confidence, one entry, and a clean +0.69R (TP1) into the session low.