forex-scalping.com

The chance of eight losses in a row over 200 trades at a 45% win rate

This question decides whether a risk size is survivable.

Take a system that wins 45% of the time, so each trade has a 55% chance of losing. Over 200 trades, what's the chance of at least one run of eight or more consecutive losses?

It's about 53%. More likely than not.

For comparison, at least one run of six or more is about 93%, close to certain. A run of ten or more is about 20%, so one in five.

These are calculated with a simple recursive count on the run length, assuming trades are independent. Real trading isn't perfectly independent, since bad conditions tend to cluster, which makes long streaks more likely, not less.

Now apply it to risk. At 2% per trade, eight losses in a row is about a 15% drawdown. At 1% it's about 7.7%. At 0.5% about 3.9%.

So a trader at 2% risk with a 45% win rate system should expect, more likely than not, to see a 15% drawdown from a single streak within 200 trades. That's before counting the normal ups and downs around it.

If that drawdown would make you abandon the system or break a prop limit, the risk size is too high for that system, regardless of how good the edge is. Working out the likely streak first, then the risk, puts things in the right order.

If you want a quick rule of thumb, the typical longest losing run over n trades is roughly the log of (n times the win rate) divided by the log of 1 over the loss rate. For 200 trades at 45% wins that's the log of 90 over the log of 1.82, about 7.5. So the longest streak usually lands around seven or eight, which matches the probabilities above. Use your own win rate and trade count, since small changes in either move the answer.

The numbers are a good reality check. Most traders feel that eight losses in a row means something is broken, but at that win rate it is close to a coin flip over 200 trades. One way to use this is to set the risk per trade so that a run of ten losses would still leave the account within its limits. At 1 percent risk, ten losses is roughly a 10 percent drawdown, which many traders and prop firms can absorb. At 2 percent it is closer to 18 percent, which is much harder.

If you want to check your own system, the inputs are your win rate and the number of trades in your period. The calculation is easy in Python: keep an array of probabilities for each current run length, update it trade by trade, and accumulate the probability of reaching the target length. A few lines of code give you the exact number. A spreadsheet simulation with random numbers gets close enough if you run a few thousand trials. The exact calculation is better because it's fast and repeatable, and you can run it for every win rate and streak length you care about.

Higher win rates change things dramatically. At 60% wins, so 40% losses, the chance of eight in a row over 200 trades drops to about 7%. That's one reason high win rate systems feel smoother, even when their expectancy isn't better. Low win rate trend systems with big winners need much smaller risk per trade to survive the streaks they're guaranteed to have. At 50% wins the chance is around 30%, so the shift between 45% and 60% win rates is large. That's part of why traders on low win rate systems need more patience than the numbers alone suggest. The emotional cost of streaks is real, and risk size should account for it.

Streaks aren't the only measure of pain. A long run of alternating small wins and losses with a slight negative drift can produce a bigger drawdown than one streak, and it feels worse because there's no clear point where you know things are bad. Looking at the maximum drawdown in a Monte Carlo of your trades gives a fuller picture than streak length alone. Running a few thousand resamples of your actual trade results, in random order, shows the distribution of drawdowns your system could produce. The 95th percentile drawdown is a good number to compare with your limits.

One practical rule I've seen work: decide in advance that a streak of a certain length triggers a review, not a stop. For a 45% system, eight losses in a row is within normal range, so it shouldn't trigger panic. Twelve in a row is roughly a 6% event over 200 trades, so that's a reasonable point to stop and check whether something about the market or the execution has changed. Writing down both the review trigger and the stop trigger before you start removes the decision from the bad moment. That way a streak leads to a planned check instead of a change made in frustration.

Blunt version: 200 trades is a sample large enough to feel serious and small enough to be mostly noise, particularly for a method with a modest edge. Take a system that wins 55 percent of the time with equal wins and losses. Over 200 trades, runs of results that look like a 48 or a 62 percent win rate come up purely from chance. A losing streak of eight or nine in a row is well within normal for such a system. None of that says the edge is gone or confirmed.

The curve itself misleads too. Scalps often cluster around the same sessions and similar conditions, so 200 trades might represent a few dozen distinct market situations, not 200 independent tests. Correlated trades make the effective sample smaller than the count.

What it can tell you: gross problems. If execution costs eat half the average win, or the method behaves nothing like the test, 200 trades will show it. What it cannot do is separate a small true edge from luck.

A more honest read uses ranges instead of single numbers, such as the plausible span of win rate and expectancy given the count, and a comparison against what the backtest predicted for the same market conditions.

How many trades do you wait for before deciding a scalping method is behaving as tested?