Losing-Streak Probability: Even a 55% Win-Rate Strategy Can Lose Eight in a Row

Most people change strategies after six consecutive losses. Yet a strategy winning 55% of the time has well over a one-half chance of at least six straight losses across 200 trades. Streaks are calculable beforehand, not automatic evidence of failure. Knowing this before starting changes the experience.

Start with the Probability of One Streak

Let p be win probability and q = 1 − p be loss probability. With independent outcomes, the probability of k consecutive losses is q multiplied k times.

Consecutive Loss Probabilities at a 55% Win Rate, q = 0.45

3 losses: 0.45³ = 9.1%
5 losses: 0.45⁵ = 1.8%
8 losses: 0.45⁸ = 0.17%
10 losses: 0.45¹⁰ = 0.034%

At a 40% win rate, q = 0.60:
3 losses: 21.6% · 5 losses: 7.8%
8 losses: 1.7% · 10 losses: 0.60%

An eight-loss streak at 0.17% appears almost impossible, encouraging the conclusion that the strategy must be broken. That conclusion is wrong because the calculation considers only one starting point.

The Real Question: At Least One Eight-Loss Streak in 200 Trades

What matters is not the chance of losing the next eight trades, but the chance of an eight-loss streak somewhere during 500 trades this year. With 500 potential starting points, there are many opportunities and the probability changes completely.

An approximate probability of at least one streak of length k or longer in N trades is 1 − (1 − qᵏ)^(N−k+1). Overlapping windows make it inexact, but it is close enough for practical judgment.

55% Win Rate, 200 Trades: At Least One Streak of k Losses

4 or more: About 99%, effectively certain
5 or more: About 97%
6 or more: About 81%
7 or more: About 52%
8 or more: About 28%
10 or more: About 6%

Eight-loss example:
0.45⁸ = 0.00168
0.00168 × 193 ≈ 0.325
1 − e^(−0.325) ≈ 0.28

A functioning 55% strategy traded 200 times in a year can therefore experience six losses about eight times in ten, and eight losses roughly one time in four. Without knowing this, you may discard a normally functioning strategy at the sixth loss.

A quicker estimate is Expected longest losing streak ≈ ln(N) ÷ ln(1 ÷ q). At 55% over 200 trades: ln(200) ÷ ln(1÷0.45) = 5.30 ÷ 0.80 ≈ 6.6. For trend following with a 40% win rate: 5.30 ÷ 0.51 ≈ 10.4. Long streaks are part of the design of lower-win-rate strategies.

The Drawdown a Streak Leaves Behind

Streak length is only half the issue. The other half is risk per trade. The same eight losses produce very different outcomes at different risk fractions.

Account Drawdown from Eight Losses — Risk to Capital at the Stop

1% risk per trade:
0.99⁸ = 0.9227 → −7.7%
Recovery requires +8.4%.

2% risk per trade:
0.98⁸ = 0.8508 → −14.9%
Recovery requires +17.5%.

5% risk per trade:
0.95⁸ = 0.6634 → −33.7%
Recovery requires +50.8%.

10% risk per trade:
0.90⁸ = 0.4305 → −57.0%
Recovery requires +132%.

If an event with 28% probability halves the account, the sizing is gambling rather than a strategy. At 1% risk, the same streak loses only 7.7%. The priority is budgeting for streaks in advance, not increasing the win rate.

The practical sequence is: ① Estimate the strategy's win rate from a sample → ② Calculate its expected longest streak → ③ Set a drawdown you could tolerate through the whole streak → ④ Work backward to per-trade risk. See position sizing for quantities and drawdown for recovery requirements.

Working Backward

Assume 45% wins and 300 trades/year.
q = 0.55 → Expected longest streak:
ln(300) ÷ ln(1÷0.55) = 5.70 ÷ 0.60 ≈ 9.5 losses

Tolerable drawdown = 20%.
To keep 10 losses within 20%:
(1−r)¹⁰ ≥ 0.80 → r ≤ about 2.2%

→ Fix per-trade risk at 2% or less.

Normal Streak or Broken Strategy?

A normal streak does not mean continuing regardless of circumstances. Strategies can fail. The distinction depends on sample size, not intuition.

For a strategy with win probability p observed n times, the standard deviation of the measured win rate is √(p·q ÷ n). A low observed rate means little when this uncertainty is large.

Sampling Error for a 55% Strategy — Standard Deviation

n = 20 → √(0.55×0.45÷20) = 11.1 percentage points
2σ range = 33%–77%
→ A 35% observed win rate over 20 trades is still normal.

n = 50 → 7.0 points · 2σ range 41%–69%

n = 100 → 5.0 points · 2σ range 45%–65%
→ Begin investigating below 45% over 100 trades.

n = 400 → 2.5 points · 2σ range 50%–60%

Recent results from 20–30 trades cannot establish a conclusion. A 35% win rate in the latest 20 trades commonly occurs under a true 55% strategy. Changing strategies then destroys the sample needed to evaluate it.

More credible signs of failure include: ① win rate remaining below 2σ with at least 100 observations; ② payoff deterioration turning expectancy negative; ③ changed loss behavior, such as gaps and slippage repeatedly exceeding stops that previously held; and ④ structural changes to the trend, range, or volatility conditions assumed by the strategy. Judge several signs together rather than one number.

Experiencing Streaks Before Trading

Tables do not fully convey their force. Before trading, use Monte Carlo simulation to experience possible paths. Keep the historical trades unchanged and randomly reorder them across 1,000 replays.

The same set of trades can produce very different longest streaks and maximum drawdowns in different orders. If the worst 5% of 1,000 runs exceed your account limit, the sizing is already dangerous. Judging tolerability from one backtest MDD is a common error: it represents only the one sequence that actually occurred.

Mark streak intervals separately in your journal. Later review often shows that major damage came from your response during the streak, not the streak itself. Doubling size to recover quickly is typical revenge trading. A streak that would have lost 7.7% can become −20% when the final two trades are tripled.

For mathematical bet sizing, see the Kelly criterion. It depends on accurate win-rate and payoff estimates. Because actual estimates are uncertain, traders commonly use no more than half the calculated amount, or half Kelly.

Three Key Points

① A specific k-loss streak has probability qᵏ, but the practical question is at least one streak in N trades: 1 − (1 − qᵏ)^(N−k+1). At 55% wins over 200 trades, the approximate probabilities are 81% for six losses and 28% for eight.
② Expected longest streak ≈ ln(N) ÷ ln(1 ÷ q). Work backward from tolerable drawdown through that streak to per-trade risk before trying to improve win rate.
③ Recent 20–30-trade results cannot diagnose failure. Investigate when at least 100 trades show a win rate below 2σ and negative expectancy.

Notice

Win rates, probabilities, and drawdowns are illustrative calculations, not measured performance of an account or strategy. The calculations assume independent outcomes. Actual losses often cluster within persistent conditions, so real streaks can be longer. Probability estimates help assess risk in advance and do not guarantee profit. Leveraged trading can lose all principal. Investment decisions and responsibility are yours.

NOONOO TRADING invites you to follow live trading in our free chat.

Start in the bot

📈 OKX trading fee discount for new registrations