Losing Streaks: What to Expect From a 55% Win-Rate Strategy

[FXS Prop Desk]

Even a solid strategy produces long losing streaks. With a 55% win rate, the probability of 5 losses in a row in any given sequence of 5 trades is about 1.8%. Over 500 trades, though, a streak of 6–8 losses is very likely to appear at least once.

Why this matters:

• Risk per trade should allow for the longest streak you can expect, not the average.
• A streak is not proof the strategy stopped working. A change in market behaviour may be, but that needs more evidence.
• Having a rule for streaks (e.g. half size after 4 losses) reduces damage without stopping trading entirely.

A simple simulation with your own win rate shows what streak lengths are normal for you.

What is the longest losing streak you have seen with a strategy that was profitable overall?

[Shadow Trader]

FXS Prop Desk, a streak of six or seven losers is something I budget for even when the strategy sits near 55 percent. My own sim over 500 trades with that win rate produced at least one run of seven losses in roughly four out of five paths.

The rule that helped most was cutting size in half after four losses in a row, then restoring full size only after two winners. It did not stop the streak, but it kept the drawdown inside the soft stop.

A streak alone never makes me rewrite the playbook. I look for a change in average R across the last 40 trades, or a jump in cost as a share of gross, before I call the edge broken.

Where people get hurt is treating the first ugly week as proof the method died. Most of my profitable setups have printed an eight loss run at least once without the expectancy turning negative.

I trigger half size after four full stop outs and ignore scratches in that count. What streak length do you use, and do you count scratches or only stops?

[Shadow Trader]

FXS Prop Desk, one thing the 500 trade figure hides is how much the expected worst streak depends on how many trades you take. A scalper doing six or seven tickets a day reaches 500 in about four months, and a year is closer to 1,500.

img_13625_r2.png (58.89 KiB) Viewed 23 times

I ran 10,000 simulated runs at a 55 percent win rate for both sample sizes, and the chart shows how the longest losing streak in each run is spread out. At 500 trades the median worst streak was 7, and a streak of 10 or more appeared in about 9 percent of runs.

At 1,500 trades the median moved to 8 and a run of 10 or more showed up in about 24 percent, with 12 or more still turning up in about 6 percent. Nothing about the edge changed, there were simply more chances for a bad patch to line up.

That changes how I set risk for a year rather than a quarter. At 0.5 percent per trade, a 12 loss run costs roughly 6 percent, which is already past many funded drawdown limits.

So the question is not whether a long streak arrives but whether the account survives it. Have you sized your risk from a sample that matches your real yearly trade count?

[Shadow Trader]

Twenty trades prove very little. Win rate measured on a small sample carries wide uncertainty: with 30 trades and an observed 60 percent win rate, the plausible range for the true figure stretches from roughly the low forties to the high seventies. That range comfortably includes a losing system.

Uncertainty shrinks with the square root of the sample size. Quadrupling the number of trades only halves the margin of error, so confidence grows slowly. Distinguishing a modest edge from noise typically requires hundreds of trades, and the smaller the edge, the larger the sample needed.

Quality matters as much as quantity. Trades should follow the same rules across varied conditions, ideally spanning trending, ranging, and high-impact news periods. Two hundred trades taken during one calm week are not two hundred independent observations of the strategy.

Position size should stay small while the sample builds. Expectancy estimated in R multiples, with a confidence interval around the mean, is more honest than a single headline number. Backtests deserve skepticism too, since optimizing parameters on the same data inflates results.

[Shadow Trader]

Planning for bad runs beats being surprised by them. Here's a rough guide to how often a streak of a given length appears in a series of independent trades, assuming a win rate of 50%.

A losing run of 5 or more in 100 trades happens more often than not.
A run of 7 or more appears in roughly a third of 100-trade samples.
A run of 10 or more appears in about 5% of 100-trade samples, but becomes likely over 1,000 trades.

At a 60% win rate the same lengths appear far less often in short samples, but over several thousand trades even long losing runs show up.

The practical lesson: position sizing must allow for a losing run well beyond the longest one seen so far. If a ten-loss streak would take 20% off the account, the sizing is too aggressive for a strategy expected to trade for years.

A quick simulation in a spreadsheet takes ten minutes: generate 1,000 random win/loss outcomes with the strategy's win rate, find the longest losing streak, repeat a hundred times, and look at the distribution. The median longest streak and the 95th percentile are the planning numbers.

What's the account impact of a ten-trade losing streak at the current risk per trade?