Is My Strategy Broken or Is This a Normal Drawdown?

A losing stretch is only evidence of a broken strategy if it falls outside the distribution your own statistics predict. Inside that distribution, a red run is not a malfunction — it is the cost of the method, arriving on schedule. The test is arithmetic: model the streaks your win rate produces, then check whether the current one is unusual or merely unpleasant.

Most traders never run that test, so the decision to abandon a system gets made on how the drawdown feels rather than on whether it is statistically remarkable. That is how profitable strategies get switched off during their normal operating range, usually a few weeks before the distribution turns.

Do trading strategies stop working for a few months and then come back?

Both things happen, and they look identical while you are inside them. A strategy in a normal variance trough and a strategy whose edge has genuinely decayed produce the same experience: signals firing, trades losing, equity falling. The difference is not visible in the feeling. It is visible in whether the current run exceeds what the strategy has already demonstrated it can do.

Three genuinely different situations get confused with each other:

Normal variance. Any edge with a win rate below 100% produces losing runs. At a 40% win rate, a run of six consecutive losses appears often enough that it should be treated as a scheduled event rather than an anomaly. Nothing has changed; you are simply in the part of the distribution nobody screenshots.

Regime mismatch. The strategy still works in the conditions it was built for, and those conditions are absent. A breakout system in a compressed range is not broken — it is being asked to operate outside its domain. This one is temporary by definition, and it does come back.

Edge decay. The relationship the strategy exploited has changed structurally. This is the only case where switching off is correct, and it is the rarest of the three.

The difference between losing streaks and winning streaks covers how runs distribute. The question here is narrower: what do you do about it.

What is the numerical test for a broken strategy?

Compare the current drawdown against the distribution of drawdowns your own trade history produces, not against your expectations. Three numbers make the call.

1. Your expected maximum losing streak. From your win rate and sample size, calculate the longest run of losses that appears in normal operation. A 45% win rate over 200 trades produces longer runs than most traders assume. If your current streak is inside that range, you have no evidence of anything. The losing streak calculator returns this directly.

2. Your expected maximum drawdown. Take your historical trades and reshuffle their order several thousand times. Each reshuffle produces a different equity path from identical trades, and the distribution of maximum drawdowns across those paths tells you what the strategy can do without changing at all. Your single historical drawdown is one draw from that distribution — usually a lucky one, since a worse draw would have made you abandon the strategy earlier. The Monte Carlo simulator does this reshuffling.

3. Where the current drawdown sits in that distribution. If it lands inside the range the reshuffles produce, the strategy is behaving normally by its own measure. If it exceeds every path in several thousand reshuffles, that is a genuine signal — not proof of decay, but the first evidence that would justify one.

This is the whole test. It converts "this feels wrong" into a percentile.

When should you actually turn a strategy off?

When the current performance sits outside the modeled distribution and you can identify what changed. Both conditions matter, because the first alone produces false positives at exactly the rate you would expect — run enough strategies and some will hit their tail without anything being wrong.

Reasonable grounds for switching off:

• The drawdown exceeds the worst path in several thousand reshuffles of the strategy's own trades
• Win rate has moved outside its historical range across a sample large enough to be meaningful — 30 trades is not that sample, 150 might be
• The specific market structure the strategy depends on has measurably changed, and you can name it
• Costs have risen enough to eliminate the edge, which happens quietly to high-frequency methods

Not reasonable grounds:

• A losing streak that your own statistics predict
• A red month, in isolation
• A different strategy performing better over the same window
• Discomfort

The asymmetry is worth stating plainly. Turning off a working strategy during normal variance costs you the entire remaining edge. Leaving a genuinely decayed strategy running costs you the drawdown until you notice. On a funded account with a hard floor the second error is more expensive in the short run, which is why the test needs to be run in advance rather than improvised during the drawdown.

How does this decision change on a funded account?

The floor removes the option of waiting. On your own capital, a drawdown that sits inside your expected distribution can be waited out at full size, because the account survives regardless. On a prop account with a trailing drawdown, the same statistically normal stretch can breach before the distribution turns.

That changes the practical response. The answer is not switching strategies — the strategy is fine, by its own numbers. The answer is sizing such that the expected worst path fits inside the drawdown from the start. If your modeled maximum drawdown is larger than the account's floor, the strategy is not wrong; the size is. That calculation happens before the attempt, not during.

There is also a recovery asymmetry that punishes the wrong response. A 20% drawdown requires a 25% gain to break even, and a 33% drawdown requires 50%. Increasing size during a drawdown to recover faster raises the probability of breaching before the recovery arrives — the drawdown recovery calculator makes the arithmetic explicit.

The structural answer is running strategies whose bad stretches do not coincide. Two systems with genuinely uncorrelated returns produce a combined equity curve with shallower troughs than either alone, which means the size that survives the floor is larger for the pair than for a single strategy. That is the reasoning behind our portfolio compositions — not more strategies for their own sake, but bad months landing in different places.

What to do while you are in one

Keep trading the plan at the size you already calculated, and collect the data that would answer the question. Specifically: log every trade normally, and once the streak resolves, re-run the reshuffle including the new trades. Either the drawdown stays inside the distribution, in which case you learned that your worst case is worse than you thought and your sizing needs to reflect it, or it exits the distribution, in which case you have your first real evidence.

What you should not do is change the strategy mid-drawdown. Every modification during a losing run resets the sample and destroys the only thing that could have answered the question. A strategy modified three times in a bad quarter has no statistics at all — it is three strategies with 30 trades each, none of which can tell you anything.

FAQ

How do I know if my trading strategy stopped working?

Compare the current drawdown to the distribution your own trades produce. Reshuffle your historical trade sequence several thousand times and look at the range of maximum drawdowns across those paths. If the current drawdown falls inside that range, the strategy is behaving normally. Only a drawdown exceeding every simulated path is evidence of decay.

Do trading strategies stop working for a few months and then come back?

Yes, and this is usually regime mismatch rather than edge decay. A strategy built for trending conditions produces losses in a compressed range without anything being wrong with it. Genuine edge decay, where the exploited relationship changes structurally, is much rarer than temporary condition mismatch.

How long a losing streak is normal?

It depends on win rate and sample size. At a 40% win rate, runs of six consecutive losses occur often enough to be treated as scheduled rather than exceptional. Calculate the expected maximum streak from your own win rate before assuming a run is abnormal.

Should I stop trading a strategy during a drawdown?

Only if the drawdown exceeds what several thousand reshuffles of your own trades produce, and you can identify what changed structurally. Turning off a working strategy during normal variance forfeits the remaining edge. On a funded account the response is usually reducing size so the expected worst path fits the drawdown, not switching strategies.

Performance figures are a combination of live-tracked and modeled results. Past performance does not guarantee future results. Not financial advice.