How Many Losing Months Are Normal for a Trading Strategy?
Three losing months in a row can be normal even for a good strategy
Under independent Normal monthly excess returns, a strategy with a true annualized Sharpe ratio of 1.0 has a 90.8% chance of at least one three-month losing streak over five years. A four-month streak appears at least once with probability 56.8%. Even six losing months in a row have a 21.3% chance of appearing somewhere in a ten-year record.
What a Sharpe 1 strategy can look like over five years
The longer the required streak, the less likely it becomes. But the probabilities remain much larger than intuition often suggests. In this benchmark, a five-year Sharpe 1 record is almost certain to contain two losing months in a row at least once, usually contains a three-month streak, and is more likely than not to contain a four-month streak.
Chance of at least one losing streak over five years
Probability of losing-month streaks for a Sharpe ratio of one over five years
How common is a three-month losing streak?
A three-month streak is a useful stress point because it is long enough to feel meaningful to an investor but short enough to arise surprisingly often. The table shows the probability of at least one such streak somewhere in the stated track record.
A longer track record creates more opportunities for a three-month losing streak
For a true Sharpe ratio of 1.0, the probability of at least one three-month losing streak rises from 34.6% over one year to 90.8% over five years. Even Sharpe 2 reaches 63.0% over five years.
| True annualized Sharpe | 1 year | 3 years | 5 years | 10 years |
|---|---|---|---|---|
| 0.5 | 46.2% | 87.1% | 96.9% | 99.9% |
| 1.0 | 34.6% | 75.4% | 90.8% | 99.2% |
| 1.5 | 24.3% | 60.4% | 79.3% | 95.9% |
| 2.0 | 16.1% | 44.3% | 63.0% | 86.7% |
The horizon matters as much as the headline Sharpe. A Sharpe 2 strategy has only about a 16.1% chance of a three-month losing streak within a single year, but that rises to 63.0% over five years and about 86.7% over ten years because the longer record provides many more places for a streak to occur.
Sharpe 1 streak probabilities by track-record length
| Losing streak | 1 year | 3 years | 5 years | 10 years |
|---|---|---|---|---|
| 2 months | 75.5% | 98.8% | 99.9% | 100.0% |
| 3 months | 34.6% | 75.4% | 90.8% | 99.2% |
| 4 months | 12.9% | 38.7% | 56.8% | 82.0% |
| 6 months | 1.6% | 6.3% | 10.9% | 21.3% |
The six-month row is a useful reminder not to overcorrect the lesson. Six straight losing months are far less routine than three. For Sharpe 1, the probability is only about 1.6% within one year, about 10.9% over five years, and 21.3% over ten years.
How the streak probability is calculated
Let S be the true annualized Sharpe ratio. Under independent Normal monthly excess returns, the monthly Sharpe is S / √12. The probability that any one month has negative excess return is therefore:
Calculating a streak is not the same as raising q to a power. A long record contains many overlapping places where a streak can begin. The exact calculation keeps track of the current run length from zero through one less than the target streak. A positive month resets the run to zero. A negative month extends it by one. Once the target run length is reached, the streak has occurred.
After all months are processed, the probability of at least one target streak is one minus the probability that the process remained in the no-streak states for the entire track record. This finite-state recurrence is exact under the stated monthly independence model; no Monte Carlo approximation is required.
What this benchmark leaves out
Monthly independence is the biggest simplification. Andrew Lo's work on Sharpe-ratio statistics shows why serial dependence changes square-root-of-time scaling. It also changes streak behavior directly. Positively autocorrelated losses may cluster more than this table suggests, while mean-reverting returns can do the opposite.
Normality is not a promise about real strategies. Skewness, fat tails, volatility clustering, changing regimes, stale pricing, and overlapping exposures can all alter the observed pattern of losing months.
The true Sharpe is assumed known. Real investors only have an estimate. A high backtested Sharpe may be noisy or selected from many tested alternatives. The Probabilistic Sharpe Ratio, Deflated Sharpe Ratio, and backtest-length study address those separate inference and selection problems.
The related Sharpe probability study focuses on the chance of negative aggregate excess return over an entire horizon. This study asks a different path-dependent question: whether at least one consecutive run of losing months appears somewhere inside the horizon.
Download the full losing-streak matrix
The public dataset contains 64 exact streak scenarios across four Sharpe ratios, four track-record lengths, and four streak lengths, plus the one-month loss probability for each Sharpe ratio.
Research data
- Download CSVCSVCSV downloadTabular public study data for spreadsheet analysis, independent checks, and new charts.Data snapshot September 17, 2026 · Reuse with attribution to the canonical study.
- Download JSONJSONJSON downloadStructured public study data for programmatic verification while preserving the published field names and research context.Data snapshot September 17, 2026 · Reuse with attribution to the canonical study.
Sources and further reading
When citing a streak probability, preserve the true Sharpe ratio, track-record length, target streak length, and the independent-Normal monthly-return assumption.