Losing streak probability

2026-10-24

A run of consecutive losses is not evidence that something has broken. At a 45% win rate, a streak of six losses has roughly a 2.8% chance of occurring on any given sequence of six trades — which means it shows up several times in a few hundred trades, by construction.

Knowing the number in advance changes what you do when it happens.

The probability of a run

For independent trades with win probability p, the chance of k consecutive losses is:

P(k losses in a row) = (1 − p)^k

At a 45% win rate, so a 55% loss rate:

Streak	Probability of that exact run
3 losses	16.6%
5 losses	5.0%
6 losses	2.8%
8 losses	0.84%
10 losses	0.25%

Those look small. They are not, because you do not get one attempt.

Over many trades, the small numbers arrive

Across n trades there are many overlapping opportunities for a run to start. A useful approximation for the expected longest losing streak in n trades is:

longest streak ≈ log(n) ÷ log(1 ÷ (1 − p))

At a 55% loss rate, log(1/0.55) ≈ 0.598:

Trades	Expected longest losing streak
50	~6.5
100	~7.7
250	~9.2
500	~10.4
1000	~11.6

So a trader taking 250 trades a year at a 45% win rate should expect to see a run of nine losses at some point, as a normal feature of a working strategy. Not as a warning sign. As the arithmetic.

The practical consequence is that the streak which makes you doubt everything is usually the one you should have written down in advance.

What a streak does to the account

Streak probability only matters through what it costs. At a fixed risk fraction f per trade, after k consecutive losses the balance is:

remaining = (1 − f)^k

Risk per trade	After 6 losses	After 10 losses
0.5%	−3.0%	−4.9%
1%	−5.9%	−9.6%
2%	−11.4%	−18.3%
5%	−26.5%	−40.1%
10%	−46.9%	−65.1%

Read the last two rows against the drawdown recovery calculator. A 10% risk per trade and a ten-loss streak — which the table above says to expect within a few hundred trades — leaves you needing a 187% gain to get back. That is not a bad run. That is an account that no longer functions.

Note also that fixed-fractional sizing shrinks the loss as the balance falls, which is why the numbers are better than k × f would suggest. That is the mechanism working. It does not save you at large f.

This is the case for a cooldown

A cooldown after consecutive losses is usually framed as emotional discipline — step away, clear your head. That framing makes it sound optional.

The arithmetic case is narrower and harder to argue with:

The probability of a loss does not change during a streak. Your behaviour does. The streak is not informative about the next trade — but by the sixth loss, position sizes tend to drift, stops tend to move, and setups that would have been skipped start looking acceptable. The damage comes from the response, not from the run.

A cooldown enforced as code removes the window in which that happens. It does not need to believe anything about the market. It only needs to be true that you are a worse decision-maker six losses in than you were at the start, which is both obviously true and impossible to notice at the time.

Same reasoning as a daily loss limit, and the same reason it should not be a note in a prompt or an intention — see why MCP needs an approval layer for the version of this argument that applies when a model is drafting the trades.

The independence assumption

The formulas above assume trades are independent. They are not.

Losses cluster, because the conditions producing one produce the next. A ranging market takes out breakout entries repeatedly; a trending market punishes mean-reversion entries in a row. Correlated losses make long streaks more likely than the table says, not less.

So treat these numbers as a floor. If the arithmetic says expect nine and your setups are correlated, expect worse.

Using this

Before you need it, work out two numbers:

The longest streak you should expect over your annual trade count, from your actual win rate.
What that streak costs at your risk per trade.

If the second number is one you could not tolerate, the fix is the risk percentage — it is the only term you control. The win rate is what it is, and the streak is going to happen.

Writing both down in advance converts the experience from “something is wrong” into “this is the six-loss case, I expected it around now”, which is the whole value.

FAQ

How likely is a 10-trade losing streak?

For any specific run of ten trades at a 55% loss rate, about 0.25%. Across a few hundred trades the chance of seeing one somewhere is much higher — the expected longest streak in 500 trades at that loss rate is around ten. Rare per attempt is not rare over a career.

Does a losing streak mean my strategy stopped working?

Not on its own. Streaks of the length predicted by your win rate are a normal feature of a working approach, and mistaking them for a signal is how people abandon positive-expectancy strategies. What is informative is whether your realised win rate over a large sample has moved, which a run of six trades cannot tell you.

What risk per trade survives a long streak?

That depends on what drawdown you can tolerate and recover from. The useful exercise is to take your expected longest streak, compute (1 − f)^k for your risk fraction, and check the recovery required against the drawdown recovery calculator. If the answer is uncomfortable, the risk percentage is the term to change.

Why have a cooldown if the odds are the same after a streak?

Because the odds on the next trade are unchanged but your decision-making is not. Sizing drifts, stops get moved, and marginal setups start looking acceptable. A cooldown enforced in code removes the window in which those things happen, and it does so without requiring you to notice that they are happening.