Is My Trading System Broken Or Just Unlucky?

Your system has just printed its tenth loss in a row. You’re staring at the screen thinking: is my trading system broken, or is this just a bad joke.

You’re not wondering about the maths.

You’re wondering if you should hit the big red OFF switch before it eats the rest of your account.

Why the question “is my trading system broken” is backwards

When you ask “is my trading system broken”, what you usually mean is “I’m seeing pain I didn’t emotionally budget for”.

The trades are real now. The backtest was just lines.

The correct question is colder: “Is this drawdown statistically consistent with what I tested, at my current risk per trade?”.

If yes, you’re probably just unlucky.

If no, you might be trading a corpse.

Step 1: Define what “broken” actually means

Most traders call a system broken when it has “too many losses” or “feels different”.

That’s not a definition. That’s a mood.

“Broken” needs a mechanical definition before you ever go live: specific numbers that, if breached, mean you pause or retire the system, no debate, no late-night forum advice.

You set those numbers from your testing.

Not from how brave you feel this week.

Three numbers that decide if it’s luck or decay

You don’t need twenty metrics.

You need three.

Maximum historical drawdown, worst losing streak, and a simple expectation of what an average run of trades looks like for your win rate and risk–reward.

If today’s pain sits inside those bounds, you’re probably fine.

Uncomfortable, but fine.

1. Historical drawdown: your pain budget

Take your backtest and any forward test and find the worst peak-to-trough drop in equity.

Say a system risking 1% per trade showed a worst historical drawdown of 18% across thousands of trades, with a typical range of 8–15% during more common bad patches.

You do not go live and then panic at -7% wondering what’s gone wrong; that’s inside the “normal misery” zone.

Where it gets interesting is when live trading breaches that 18% historical worst, adjusted for your actual risk per trade.

If you doubled the risk to 2% per trade, your drawdown in percentage terms will roughly double as well.

So that 18% backtest number becomes something like 36% in live risk terms before it’s comparable.

A 20% live drawdown on double risk is not evidence of decay.

It’s evidence you liked the equity curve too much and the risk section too little.

2. Losing streak: what your brain hates most

The fastest way to think your system is broken is a long losing streak.

Eight, ten, twelve losses in a row feels like the market is personally offended by you.

But the length of a “normal” losing streak is pure probability: mainly driven by win rate.

Rough guideline (per 1,000 trades or so):

• 70% win rate: expect worst losing streak around 5–7
• 55% win rate: expect worst losing streak around 8–10
• 40% win rate: expect worst losing streak around 10–14

Those are not magic numbers, they’re approximate, but they show the point: even decent systems can have nasty runs.

The mistake is treating the first time you see that streak live as proof the system has died.

Usually it’s proof your expectation setting was optimistic.

3. Expectancy: is the average trade still doing its job?

Suppose your system historically wins 45% of the time, average winner is 2R, average loser 1R.

Your expectancy per trade is:

E = 0.45 × 2R − 0.55 × 1R = 0.35R

If over the last 100 live trades you’re seeing something like 0.30R to 0.40R per trade, you’re broadly in line with the model, even if the sequence hurts.

If you’re at -0.10R per trade over those 100 trades, something has shifted: win rate lower, winners smaller, losers larger, or all three.

You then ask whether that shift is inside the worst few hundred-trade patches you saw in testing, or completely outside anything you’ve seen before.

That’s where Monte Carlo comes in, which we covered at length here: /blog/what-is-monte-carlo-simulation-in-trading.

If your backtest never threw up a run as bad as your current one, even when shuffled thousands of times, that’s strong evidence you’re no longer in “just unlucky” territory.

Backtest bounds vs live pain: how to compare apples with apples

You can’t compare live performance to a backtest without adjusting for two boring details most people skip.

Position sizing and sample size.

Skip either one and you’ll call systems broken that are perfectly intact, or keep trading systems that are clearly bleeding out.

Risk per trade: your invisible amplifier

If your test risked 0.5% per trade and you’re now running 1.5%, you’ve tripled the volatility of your equity curve.

The strategy didn’t get more aggressive; you did.

So when you see a 12% live drawdown and panic, remember: in backtest risk terms that’s only 4%.

This is why basing risk on a percentage of equity, not random lot choices, matters so much; we covered the compounding side of that here: /blog/fixed-fractional-vs-fixed-lot-sizing-that-really-compounds.

Once you normalise risk, you can compare drawdowns honestly.

Sample size: you’re judging a book by three pages

A 200-trade backtest does not give you robust bounds for a 20-year live career.

All it gives you is a vague sense of how the system behaves in those specific conditions.

So when you hit a 10% drawdown in live trading and your worst backtest drawdown was 9%, that “breach” is not as dramatic as it looks if your whole test only had 80 trades in it.

The smaller the sample, the more humility you need around all of these limits.

Bad runs you “never saw in testing” often just didn’t show up in your very short sample.

What automation changes – and what it absolutely doesn’t

Automation solves a big chunk of one problem: you overriding the rules in the middle of a losing streak.

It does not magically remove variance, drawdown or the possibility that your edge decays.

You still need definitions of “normal pain” and “stop here” baked in, or you’ve just built faster execution for your panic.

Automated systems can also run across multiple markets at once – FX, gold, indices – which smooths some risk, but creates another issue.

You now have several systems, each with their own drawdown profile, correlations, and historical bounds.

So you don’t just ask “is my trading system broken”, you ask it for each component and for the whole portfolio.

And yes, watching an XAUUSD strategy swing your equity around because 0.01 lot is proportionally large on a small account is still your risk decision, automated or not.

The robot didn’t do that.

Variance vs system decay: the line in the sand

So when is a bad run “just noise”, and when are you in system decay territory?

Think in layers.

Each layer breached raises the probability that you’re looking at real edge erosion, not just bad cards.

Layer 1: Inside historical drawdown and streaks

If your current drawdown and losing streak fit comfortably inside your historical worsts (properly risk-adjusted), and expectancy over a reasonable chunk of trades is near the tested range, you are looking at variance.

Annoying, confidence-sapping variance.

But variance.

What do you do here?

Nothing dramatic.

Maybe you trim risk slightly if the account size drop makes the swings uncomfortable, but you do not re-optimise, add filters, or turn it off because you “don’t like” the last ten trades.

Layer 2: Breaching historical streaks, but not drawdown

Say your worst tested losing streak was 9 and you’re now at 12, but total drawdown is still pretty much in-line with the backtest (because losers are similar size and winners are still there, just poorly sequenced).

This is a yellow flag, not a fire alarm.

You may simply be seeing a tail event the backtest didn’t contain, especially if your sample was small.

Here you log it, maybe run a fresh Monte Carlo shuffle on your trade data, and carry on.

But you don’t start fiddling with entries because of a once-in-a-blue-moon cluster.

Layer 3: Breaching both drawdown and streaks, expectancy slumping

Now you’re in the danger zone.

Your live drawdown is beyond your historical worst even after risk adjustment, your losing streak is longer, and your per-trade expectancy over, say, the last 200 trades is significantly lower than almost any 200-trade window in your test.

That’s when you move from “just unlucky” to “probable system decay”.

Which we’ve written about here: /blog/what-is-system-decay-and-when-to-retire-a-strategy-2.

Note the word probable – markets don’t give you certainty, just odds.

Why your psychology makes everything look broken sooner

You can have all these numbers written down and still bail out early.

Because loss aversion makes a -10% drawdown feel twice as bad as a +10% upswing feels good.

So your brain screams “broken” long before the maths does.

That’s the core of why traders abandon otherwise sound systems at exactly the wrong time, which we covered here: /blog/why-traders-abandon-winning-systems-at-the-worst-time.

And why delegating execution to code often beats trying to “be strong” in front of the screen: /blog/why-trading-psychology-automation-beats-willpower.

But psychology doesn’t change the core rule.

A bad run inside your tested bounds is variance.

A clear, risk-adjusted breach of those bounds is evidence to investigate decay.

How to set your kill-switch before you ever go live

Let’s make this practical.

Here’s a simple pre-flight checklist you can write for every system so you’re not making it up during the storm.

• Document tested stats: worst historical drawdown (%), worst losing streak, and 10–20 of the worst 100-trade expectancy windows.
• Define risk level: your live risk per trade as a % of equity and how it compares to the test (e.g. “live is 2× test risk”).
• Set yellow flag rules: e.g. “If live drawdown > 0.75 × adjusted max DD, review but do not stop; if losing streak > tested max by 20%, log and continue”.
• Set red flag rules: e.g. “If live drawdown > 1.2 × adjusted max DD and 200-trade expectancy < worst tested 200-trade window, pause and investigate”.
• Pre-commit action: what you’ll actually do on a pause: analyse trade distribution, slippage, spreads, any structural market changes, not randomly “optimise” because someone on a forum likes a different moving average.

This turns “is my trading system broken” from a feeling into a checklist.

Your future self will thank you, quietly, instead of rage-typing on Telegram.

One last boring but essential line

Trading leveraged products carries a high level of risk and you can lose money; no system or level of automation removes that.

What to do with this tomorrow

Pick one system you’re running or planning.

Write down its tested maximum drawdown, worst losing streak, and a realistic expectancy range, all in risk-adjusted terms.

Then write the exact numbers that mean “variance, carry on” and the ones that mean “stop, review for decay”.

If you can’t do that from your current testing, the system isn’t ready.

It’s not broken or unlucky.

It’s just undefined.

Start your free 14-day ArcisTrade demo →

See live automated systems – across FX, gold and indices – running side by side, including the drawdowns. Watch winners and losers, no martingale, no grid, mirrored directly to your own broker. P.S. Use the demo to stress-test your tolerance for real-world variance before you put actual money behind any system.