Knowledge Hub · Portfolio & Risk ·
Maximum drawdown: how to measure the risk of an investment
Maximum drawdown explained: what the largest peak-to-trough loss shows, how it differs from volatility and what it does not show.
Key points at a glance
- Maximum drawdown is the largest fall from a peak to a later low, measured in per cent.
- Volatility and maximum drawdown measure different things: volatility ignores the order of returns, maximum drawdown does not.
- The deeper a loss, the larger the gain needed to get back to the old peak, because the loss shrinks the base.
- The Sharpe ratio relates the income above a benchmark rate to volatility, yet none of the three figures shows the whole risk.
Maximum drawdown is the largest loss of an investment from its highest level to the lowest level after it. It shows how far down things went, and it complements two other figures: volatility and the Sharpe ratio. None of the three shows the whole risk.
In brief
Peak and low
Maximum drawdown
Volatility
Sharpe ratio
Maximum drawdown: the largest loss from peak to low
Maximum drawdown does not measure the loss over the whole period. It measures the deepest fall from a peak. The order matters: the low must come after the peak. A drawdown only ends when a new peak is reached.
The formula is: (low − peak) ÷ peak. The following example applies it to six values.
Numerical example (illustration)
The values 100, 120, 90, 110, 80 and 130 are freely chosen. At the value 90, the previous peak is 120, and the fall is −25%. At the value 80, it is (80 − 120) ÷ 120 = −33.3%. This is the maximum drawdown of this series. The value 130 is above 120, so the drawdown has ended.
| Level | Peak so far | Fall from peak |
|---|---|---|
| 100 | 100 | 0% |
| 120 | 120 | 0% |
| 90 | 120 | −25% |
| 110 | 120 | −8.3% |
| 80 | 120 | −33.3% |
| 130 | 130 | 0% |
Same volatility, different depth of loss
Volatility asks how far returns deviate from the mean on average. The order in which they occurred plays no role. Maximum drawdown, by contrast, depends on exactly that. How volatility works in option prices is explained in the article on volatility in options.
Numerical example (illustration)
Two investments have the same four annual returns: twice +10% and twice −10%, freely chosen. Investment A achieves them alternately, investment B first the two gains, then the two losses. The mean of the returns is 0%, and the standard deviation is 10 percentage points. It is calculated as the square root of the mean of the squared deviations, treating the four returns as a complete series. Because they are annual returns, no conversion to a yearly figure is needed. Both end at 98.01.
| Value in CHF, start 100 | Investment A | Investment B |
|---|---|---|
| After year 1 | 110 (+10%) | 110 (+10%) |
| After year 2 | 99 (−10%) | 121 (+10%) |
| After year 3 | 108.9 (+10%) | 108.9 (−10%) |
| After year 4 | 98.01 (−10%) | 98.01 (−10%) |
| Volatility | 10 percentage points | 10 percentage points |
| Maximum drawdown | −10% | −19% |
For investment B, the value falls from 121 to 98.01. That is −19%, because 98.01 ÷ 121 = 0.81. For investment A, the value never falls by more than −10%.
Volatility shows how restless the path was. Maximum drawdown shows how deep it went on that path.
Why does a loss of 50% need a gain of 100%?
Anyone who loses 50% has only half left. For the value to reach the old peak again, this half has to double: +100%. In general: required gain = loss ÷ (1 − loss). The loss shrinks the base on which the later gain acts.
Numerical example (illustration) · interactive
After a loss of 50%, a gain of 100% is needed to get back to the peak
Starting point: a loss of 50% from the peak and an assumed return of 5% per year. Both values can be changed. The chart shows the required gain for every loss from 0% to 90%.
−50% → +100.0%
Arithmetic only, excluding costs, taxes and payments. The return is a freely chosen assumption, not a forecast. Whether and when an investment recovers is open; losses can also be permanent.
Editorial's own calculation, as of 13 November 2026: required gain = loss ÷ (1 − loss); duration = ln(1 + gain) ÷ ln(1 + return). The return is freely chosen, not a forecast.
The deeper the loss, the steeper the required gain rises. The table shows it, together with the time that recovery would take at an assumed constant return of 5% per year. How an annual return is calculated is shown in the article on the return after inflation.
| Loss from peak | Required gain | Duration at an assumed 5% per year |
|---|---|---|
| −10% | +11.1% | around 2.2 years |
| −20% | +25.0% | around 4.6 years |
| −50% | +100% | around 14.2 years |
| −70% | around +233% | around 24.7 years |
After −30% it is +42.9%, and after −90% as much as +900% (calculation, as of 13 November 2026). The duration is an assumption, not an expectation. Real paths fluctuate, and whether an investment recovers is open.
Sharpe ratio: income relative to volatility
The Sharpe ratio subtracts a benchmark rate, such as that of a low-risk investment, from the average income. The result is divided by volatility. The question is therefore how much extra income an investment delivered per unit of volatility. Why investors expect a return for taking risks at all is shown in the article on the risk premium.
Numerical example (illustration)
The benchmark rate is 1% per year, and all values are freely chosen. Investment X earns 6% per year on average with a volatility of 10 percentage points: (6 − 1) ÷ 10 = 0.5. Investment Y earns 4% with a volatility of 4 percentage points: (4 − 1) ÷ 4 = 0.75. Y delivers the larger extra income per unit of volatility. X, however, earns the higher income in absolute terms.
The calculation uses only the mean and volatility. Upward and downward moves therefore count equally, and the depth of a loss cannot be seen in it.
Risks and limitations
Each of the three figures answers only one question:
| Metric | Shows | Does not show |
|---|---|---|
| Volatility | how widely returns scatter | how deep a loss went and how long it lasted |
| Maximum drawdown | the deepest fall in the period | the duration, the frequency and future losses |
| Sharpe ratio | extra income per unit of volatility | the depth of a loss; gains and losses count equally |
- Past: Maximum drawdown is a single event from the chosen period. Past performance is not a reliable indicator of future results.
- Period: A longer period can produce a deeper value. In a mathematical model with random price movements, studied by Magdon-Ismail and others in 2004, the expected maximum drawdown grows with the length of the period. This applies to the model, not to real prices.
- No upper limit: According to MathWorks, the expected maximum drawdown is not an upper bound but an estimate of the average maximum loss. Deeper losses remain possible, up to a total loss.
- Recovery: The duration in the table rests on an assumed constant return. A loss can also be permanent.
- Sharpe ratio: The value depends on the chosen period and on the benchmark rate.
- Examples and calculator: They are calculated without costs and taxes, using freely chosen values, and are not a forecast.
The metrics in the factsheet
Whether a provider states these metrics varies from product to product. Where they are given, three details should be checked: the label (for example «maximum loss»), the period and the reference date. Without the period, two values cannot be compared. How a data sheet is structured is explained in the article How to read a factsheet.
Conclusion
Maximum drawdown shows the deepest fall from a peak within a period. Volatility shows the restlessness, and the Sharpe ratio shows the extra income per unit of volatility. In the example, two investments have the same volatility but a maximum drawdown of −10% and −19%. After a loss of 50%, a gain of 100% is needed, which takes around 14.2 years at an assumed 5% per year. None of the figures shows the whole risk, and deeper losses remain possible.
Questions & Answers
Common questions around this topic, briefly answered.
What is a good maximum drawdown?
There is no fixed benchmark. Which loss is bearable depends on the investment, the period and personal loss tolerance. A small past figure does not rule out deeper losses.
Does maximum drawdown say anything about future losses?
No, it describes the deepest fall in the chosen, past period. According to the MathWorks documentation, even the expected maximum drawdown is not an upper limit for the maximum loss from a peak. Deeper losses remain possible.
How long does it take to recover a loss?
That depends on the loss and the path. With an assumed constant return of 5% per year, it takes around 2.2 years after −10% and around 14.2 years after −50% (own calculation). Real paths fluctuate, and a loss can also be permanent.
Where do I find these figures for an investment?
If the provider states them, they appear in the factsheet or data sheet. The label (for example «maximum loss»), the period and the reference date should be checked, because without these details two figures cannot be compared.
Related insights
All insights
Important information Information only, not investment advice or a solicitation. Investments carry risks up to total loss; past performance is not a reliable indicator.
This article is for general information and educational purposes only. It does not constitute an offer, solicitation or recommendation to buy or sell any financial instrument, nor does it constitute investment, legal or tax advice.
The content does not take into account your personal circumstances, objectives or risk tolerance. Please consult a qualified professional before making any investment decision.
Investing in financial instruments involves risks, up to and including the total loss of the capital invested. Past performance is not a reliable indicator of future results. Numerical examples are for illustration only and are not a forecast.
Third-party information has been obtained from sources believed to be reliable; no guarantee is given as to its accuracy or completeness. Market data refers to the stated data date and may have changed since.