How Long Can a Losing Streak Last? The Maths of Bad Runs

Most traders accept that losing streaks happen. Far fewer have a number for how long one should be expected to last, so every bad run gets judged on feel, and feel is badly calibrated for this question. The maths is not complicated, and it tends to show that long streaks are an ordinary feature of strategies that work, not a sign that they have stopped working.

The intuition that goes wrong

Take a hypothetical strategy that wins 55 percent of the time, so any single trade has a 45 percent chance of losing. The chance that the next six trades all lose is 0.45 multiplied by itself six times, which is about 0.8 percent, comfortably less than one in a hundred. Six losses in a row feels rare, and for any specific six-trade window, it is.

But nobody trades a single six-trade window. Over 500 trades there are hundreds of overlapping windows, and the question that matters is not whether the next six lose but whether a run of six appears anywhere in the sequence. For a 55 percent strategy over 500 trades, the probability of at least one run of six or more losses is about 90 percent. Both numbers are correct. They answer different questions, and traders tend to feel the first while living through the second.

Expected longest streak, by win rate and trade count

The table below shows the expected length of the longest losing streak for a strategy with a fixed win rate, assuming each trade is independent of the last. The figures are calculated exactly, and they describe the mathematics of independent trades, not any real trader's results.

Win rate	100 trades	200 trades	500 trades	1,000 trades
40%	7.9	9.2	11.0	12.4
50%	6.0	7.0	8.3	9.3
55%	5.3	6.1	7.3	8.1
60%	4.6	5.4	6.4	7.1
65%	4.0	4.7	5.6	6.2

Two patterns stand out. First, the longest streak grows with the number of trades, but slowly. It rises roughly with the logarithm of the trade count, so doubling the number of trades adds roughly one loss to the longest run rather than doubling it: about 0.9 at a 55 percent win rate, 1.4 at 40 percent and 0.7 at 65 percent. A rough rule of thumb puts the longest losing run near log(n × p) divided by log(1 / q), where n is the number of trades, p is the win rate and q is the loss rate. For 55 percent over 500 trades that gives about 7.0, close to the exact 7.3.

Second, win rate matters more than intuition suggests. A strategy that wins 40 percent of the time, which is ordinary for approaches built on a small number of large winners, has an expected longest losing run of about 11 over 500 trades, and roughly a 70 percent chance of a run of ten or more. None of that says anything about whether the strategy makes money, which is the point made in win rate is the most misleading stat in trading.

The expected value is also only the middle of a spread. For the 55 percent strategy over 500 trades, the chance that the longest streak reaches eight or more is about 37 percent, and ten or more about 9 percent. A ten-trade losing run from a strategy that wins more often than it loses is unusual, but it would still turn up in roughly one in eleven such histories.

A hypothetical example

Consider a hypothetical strategy that wins 55 percent of the time, with average winners and average losers both equal to one unit of risk. Its expectancy is 0.55 minus 0.45, or 0.10 units per trade: modest, positive, and worth about 50 units over 500 trades in expectation. The mechanics of that number are covered in what is expectancy in trading.

Now suppose the trader running it holds an informal belief that six losses in a row means something is broken. The strategy has not changed and the edge is exactly what it was, yet the probability that a run of six or more appears somewhere in the first 200 trades is about 60 percent. Within 500 trades it is about 90 percent. Applied to a strategy behaving precisely as designed, the belief would usually be triggered.

When it is, the run itself costs six units, and the drawdown around it is usually deeper, because a losing streak tends to sit inside a broader patch of mixed results. That gap between the streak and the drawdown is the subject of what is maximum drawdown. From the inside, it looks like a positive-expectancy strategy producing nothing but losses, which is exactly when many approaches get dropped.

Why working approaches get abandoned

The pattern is easy to describe. A strategy gets judged by its recent run rather than its full sample, the recent run is the part that hurts, and a losing streak is the most vivid kind of recent run there is. Without a baseline for what variance produces at a given win rate, there is no way to tell a streak that was entirely expected from one that is genuinely unusual, so every streak gets read as evidence. It is a close relative of base rate neglect: the specific, painful sequence crowds out the background frequency that would explain it.

There is an asymmetry too. Winning streaks come from exactly the same arithmetic, and for a strategy that wins more than half the time they run longer than the losing ones. Few traders look at a run of eight winners and conclude it was probably variance. Usually only the losing kind prompts a change of approach, and the replacement then meets its own perfectly normal losing run in due course.

The limits of the maths

Every number above assumes trades are independent and the win rate is constant. Real trading rarely matches that. Win rates drift with market conditions, and losses can cluster when a strategy meets a regime it was not built for, which is the theme of your edge is a moving target. When losses cluster, streaks run longer than the table suggests, and a streak well beyond the expected range for a given win rate and sample size is harder to put down to ordinary variance. The table is a baseline for what independent variance produces, not a test that proves anything on its own.

Small samples deserve particular caution. A six-trade losing run inside a 15-trade sample is 40 percent of all the evidence there is. NoxarQuant gives no verdict at all on a cluster with fewer than 12 trades, and streak maths is a good illustration of why that kind of restraint is reasonable.

Putting your own streaks in context

The useful comparison is between your actual longest losing streak and what your own win rate and trade count would produce through variance alone. Take a hypothetical history of 300 trades at a 52 percent win rate. The expected longest losing run is about seven, a run of ten or more would appear in roughly 9 percent of such histories, and twelve or more in about 2 percent. A trader with those numbers and a nine-trade streak is well inside the normal range. A trader with those numbers and a fifteen-trade streak is looking at something variance struggles to explain.

It also helps to look at where the losses in a streak came from. In NoxarQuant every trade is tagged with its setup, session, volatility bucket, RSI bucket and trend, so a losing run can be broken down by condition. A streak spread evenly across sessions and volatility regimes looks like ordinary variance. A streak concentrated in one session, or in high-volatility conditions, describes something more specific about where the losses sit. Neither says what to do next, but both say more than the streak length alone.

The portfolio analytics show your longest runs of consecutive wins and losses. Monte Carlo resampling then redraws your out-of-sample trades at random, with replacement, into many alternative sequences, to show the spread of drawdowns that trades like yours could have produced, as described in Monte Carlo for traders.

A losing streak is easier to sit through, or to question properly, once you know roughly how long it was likely to be. That starts with measuring your own trades, which is what our case study walks through.