Five losing trades, and why a three-loss streak is normal
Every loss this week was a pullback long that did not hold. Five stops, 5R gone, a three-trade losing streak. Here is the full teardown, and the statistics that
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 part of the record most systems hide. Between August 3 and 9, five of our trades stopped out, the losses briefly ran three in a row, and the desk gave back ground. We are publishing the whole thing, every loss, because a trading system that only shows you its winning weeks is not showing you a system at all. It is showing you a brochure. Here is the context that matters 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 five losses cost 5R gross. Set against a +31.43R year, that is the give-back of one ordinary week inside a strongly positive record, and the system's drawdown from its peak is a routine 7%. The losses were not random. Every one of them was a long, and nearly every one was the same idea: buy a pullback into support and hold for the bounce. This week, the bounces did not come. When a single setup family goes cold for a few days, the losses cluster, and clustering is exactly what produces a short losing streak. The teardowns below show three of these losses in full, and the statistics section explains why a three-loss streak at our win rate is not a malfunction. It is the arithmetic working as designed.
The pullback that would not bounce
The week's losses shared a single origin: pullback-into-support longs that failed to hold. It started Monday with a US30 long stopped inside two hours, and it set the template. Buy the dip into a level, place the stop below the level, and wait for the market to reward the support. All week, across instruments, that reward did not arrive. The support levels that had been holding for weeks gave way, one after another.
The streak arrives midweek
By Thursday the losses had begun to cluster. A GBPUSD long stopped Thursday, a second EURUSD long entered Thursday and would run for days before stopping, and a second GBPUSD long stopped as well. Three losses in a row is the sequence that tests conviction, because it is exactly where the temptation to change the system peaks. This is where an undisciplined book starts widening stops or doubling size to "make it back." We did neither. Each trade took its predefined minus 1R and closed.
The five-day loss resolves
The longest-running loss of the week was a EURUSD long entered Thursday the 7th. It did not resolve inside the window. It held open for five days, through the weekend and into this week, before finally stopping on the 12th. We could have cut it early to tidy the week's numbers. We did not, because the trade was still inside its original parameters and the stop had not been hit. When it did hit, we took the minus 1R. Managing a trade on its plan rather than on the calendar is the same discipline that lets our winners run when the resolution goes the other way.
Every trade the system took.
| Date | Time | Instrument | Dir | Model | Setup | Grade | R | $ Sim | Result | Details |
|---|---|---|---|---|---|---|---|---|---|---|
| Aug 3 | 14:45 UTC | US30 | Long | GPT-5.5 | US30 Pullback Reclaim Long | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 5 | 14:20 UTC | NAS100 | Long | Claude Opus 4.7 | NAS100 Long — Pullback to Fibonacci 61.8% / VWAP Reversion | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 6 | 14:26 UTC | GBPUSD | Long | GPT-5.5 | Buy pullback into support | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 7 | 14:29 UTC | EURUSD | Long | Claude Opus 4.7 | EURUSD Pullback Buy at Session Support | C+ | -1.0R(SL) | -$2,000(SL) | Stop hit | - |
| Aug 7 | 14:29 UTC | GBPUSD | Long | GPT-5.5 | GBPUSD LONG (preferred pullback 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.
Pattern of the week
When one setup family goes cold
The pattern this week was uncomfortable in its uniformity: every loss was a long, and nearly every loss was a pullback-into-support entry. The GBPUSD losses were both pullback-continuation longs. The EURUSD loss was a pullback buy at session support. The US30 and NAS100 losses were the index versions of the same idea, buying a dip into a level and waiting for the bounce.
That uniformity is the explanation for the streak. A trading system does not lose at random across a week; it loses when the market regime turns against the setups it is currently finding. This week the market was not rewarding dip-buying. Support levels that had held were breaking, and every entry built on "this level will hold" was on the wrong side of that shift. When a single setup family is what the market is offering and that family stops working, the losses naturally cluster into a streak. The alternative, forcing different setups to avoid the clustering, would mean trading setups we have no edge in. We would rather take the honest streak.
Decision highlights
We let the three-loss streak run without touching the system. The single most important decision of a losing week is the one not to make: no widened stops, no doubled size, no switch to a different strategy mid-drawdown. Three consecutive losses is precisely where discipline is tested, and holding the process steady through it is what keeps a normal streak from becoming a self-inflicted disaster.
We held the five-day EURUSD long on its original plan rather than closing it to clean up the week. The trade was still inside its parameters and its stop had not been hit, so cutting it early would have been managing the calendar instead of the trade. When the stop finally hit this week, we took the minus 1R, the same way we would want to hold a winner that took five days to mature.
We did not abandon GBPUSD after its first loss, nor press it after its second. The first stop was one data point; the second confirmed the pair's setups were not working this week. The correct response to that confirmation is to stand down on the instrument, not to revenge-trade it, and that is what the following days reflected.
Six instruments, six stories.
EURUSD: One loss here, a pullback buy at session support that held five days before stopping. Counting its two wins elsewhere on the week, the euro was actually net positive overall; this was its single failed entry.
All EURUSD this week →GBPUSD: The week's heaviest contributor at minus 2R across two losing longs. Both were pullback-continuation entries on a pair that kept failing its support, and the second loss confirmed the setup was not working.
All GBPUSD this week →US30: One loss, a Monday pullback long stopped within two hours for minus 1R. The index version of the week's failed dip-buy template, closed cleanly at its stop.
All US30 this week →NAS100: One loss, a midweek long that would not hold its level, stopped for minus 1R. No follow-through, no argument, damage capped at the agreed risk.
All NAS100 this week →USDJPY: No losses this window. The yen sat outside our setup criteria, so it contributed nothing to the week in either direction.
All USDJPY this week →US500: No losses this window. The S&P book found no qualifying setup and stayed flat, which on a losing week was its own quiet form of contribution.
All US500 this week →Loss of the week: EURUSD Long · -1R
Losses worth learning from
EURUSD long, the five-day stop
What was right. The entry followed the process exactly: a pullback buy at session support, entered at 1.15575 with the stop placed below support at 1.1530, a defined 27.5 pips of risk. The setup was a legitimate read of a level that had held before, and the position size was correct for a 1R risk.
What was wrong. The support did not hold, and the euro spent five days chopping before resolving lower. Nothing in the entry was an error of process; it was a correct-process trade that the market did not reward. If there is a critique, it is that the level was a C-plus setup, adequate rather than premium, in a week when only premium dip-buys were paying.
What we'd do the same. Hold on plan. The trade stayed inside its parameters the entire time, so we let it run to its stop rather than cutting it early to tidy the numbers. Taking the clean minus 1R when the stop finally hit is the correct outcome.
GBPUSD long, the pullback continuation
What was right. A preferred pullback-continuation long entered at 1.34935 with a tight, well-defined stop at 1.34795, just 14 pips of risk. The read was that Cable would continue after its pullback, and the stop was placed exactly where that thesis would be proven wrong.
What was wrong. The continuation never came; the pair failed its level and stopped us in three evaluations. This was the second GBPUSD long to fail the same way in two days, which in hindsight was the market telling us the pair's pullbacks were not being bought.
What we'd do the same. Keep the stop tight and honor it. Fourteen pips of risk on a failed continuation is a cheap way to be wrong, and the tight stop is why this loss cost exactly 1R and no more.
GBPUSD long, the pullback into support
What was right. A buy-pullback-into-support long entered at 1.34684 with the stop below support at 1.34527, about 15.7 pips of risk. The system spent ten evaluations confirming the level before committing, which is the patience we want to see.
What was wrong. Despite the patient read, the support gave way and the trade stopped inside an hour. Ten evaluations of confirmation did not change the fact that the level itself was in a market that had stopped respecting support. The confirmation process validated the entry mechanics, not the regime.
What we'd do the same. Confirm before entering and cap the loss at 1R. The patience was correct even though the outcome was not; the alternative, entering faster, would only have produced the same loss sooner.
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 | -5R | −$10,000 |
All six agents combined.
From the desk
From the desk
If you are evaluating this system, this is the report to read twice, because it is the one that tells you the truth about how the losses behave. Five trades stopped this week, the losses ran three in a row, and the desk gave back 5R gross. Every one of those stops held at its predefined level. There was no trade that ran away, no position that got doubled, no stop that got moved. The losses cost exactly what we agreed they would cost.
Put in dollars, the context holds. 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. This week's give-back is a small fraction of that cumulative figure, and the +31.43R year it sits inside is intact. Drawdowns are the toll you pay to keep an edge in the market. Ours arrived on schedule, cost what it was supposed to, and changed nothing about the process that produced the +31.43R in the first place.
What we're tuning
There is nothing structural to tune out of this week, and it would be dishonest to invent a fix to look responsive. Five clean stops on qualifying pullback setups is variance driven by a regime that briefly stopped rewarding dip-buying, not a flaw in the setups themselves. Every loss cost exactly 1R, every stop held, and no single trade did outsized damage. When the losses share no error of execution, there is no execution error to correct.
The one thing worth watching is setup concentration. When a single setup family, here pullback-into-support longs, is producing most of the week's entries, a regime shift against that family clusters the losses into a streak. We do not force setup diversity for its own sake, but a week this concentrated is a reminder to stay patient when only one kind of setup is on offer and it stops working. The tuning is to conviction and patience, not to a parameter.
What the numbers actually mean
The instinct on seeing three losses in a row is to assume something has broken. The statistics say otherwise, and understanding why is the difference between reacting to a drawdown and riding through it.
Our system wins about 59% of its trades, which means it loses about 41% of them. When you lose 41% of the time, the probability of stringing three losses together is roughly 41% times 41% times 41%, a little under 7%. That sounds small until you remember how many trades we take. Across the 176 trades in our record, an event with a 7% chance per sequence will occur many times over. As Van Tharp lays out in his work on R-multiples, the individual trade is close to random; it is the distribution of many trades that carries the edge. A losing streak is not the edge failing. It is a normal cluster inside a distribution that still has positive expectancy.
Jack Schwager's interviews with professional traders return to this point repeatedly: the best track records in the world contain losing streaks that would frighten an amateur into quitting. A run of three, four, even five losses is a routine feature of any system that wins between half and two-thirds of its trades, and expecting to avoid them is expecting the math to behave differently than it does. The correct question is never "why did we lose three in a row," it is "did each loss cost only what it was supposed to." Ours did. Five losses, five R, each one capped at the predefined stop.
The system's drawdown from its peak currently sits at about 7%, with an average R target of 0.74 on winners and a 59% win rate across 176 trades. Those are the numbers of a healthy system in a normal drawdown, not a broken one in trouble. A week is a sample of five to fifteen trades; it is far too small to tell you anything about an edge that plays out over hundreds. This week was a small, expected dip inside a much longer and much more positive line.
At a Glance
Drawdown questions
Why publish a report about your losing trades?
Because reporting drawdowns is what every legitimate trading fund does, and hiding them is what marketing does. A track record is only meaningful if it includes the losing weeks. This report shows all five of this week's losses, how each one was sized, and the statistical context that explains the streak. It is more useful for evaluating the system than another winning week would be.
Is a three-loss streak a sign the system is breaking?
No. At a 59% win rate, the system loses about 41% of trades, so three losses in a row has roughly a 7% chance of occurring in any given sequence. Across 176 trades, that happens many times. Losing streaks of three to five are a normal, expected feature of any system that wins between half and two-thirds of its trades, not evidence of a malfunction.
How did every loss cost exactly 1R?
Each trade is entered with a predefined stop representing 1R, one unit of risk. When price hits that stop, the position closes. We do not widen stops, average down, or hold past the stop hoping for a reversal. That discipline is why five losses cost exactly 5R gross rather than some larger, uncontrolled figure, and why the drawdown from peak sits at a routine 7%.
Why were all the losses pullback-into-support longs?
Because that was the setup the market was offering, and this week it stopped working. Support levels that had been holding gave way, so every entry built on "this level will hold" ended up on the wrong side of a regime shift. When one setup family produces most of a week's trades and that family goes cold, the losses cluster. We would rather take the honest streak than force setups we have no edge in.
Does this week change the year-to-date track record?
This week's losses are part of the ongoing record, and they will show in the numbers as the month completes and is recorded. The headline year-to-date figure, +31.43R since inception, reflects the system through the last closed month, which is why it does not bounce on an open week. The transparency is in publishing the down weeks in full, which is exactly what this report does.
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.