Measure and manage portfolio risk Lesson 3 of 14
Portfolio and risk · Measuring return and risk
Maximum drawdown (MDD): the worst fall from a peak, and the way back
How maximum drawdown is measured, how long recoveries took for stocks, bonds and gold since 1993, why a 50% loss needs a 100% gain, and what the number leaves out.
On this page
A drawdown is how far an investment's value is below its highest point so far. The maximum drawdown is the deepest such fall over a period: the loss of someone who bought at the worst possible moment and looked at their account at the worst possible moment after that.
It is the risk measure closest to experience. Nobody feels a standard deviation. Everyone who held stocks in March 2009 remembers an account that had lost half its value. Maximum drawdown records that number, the dates it happened and how long it took to climb back. This lesson measures it on eight funds, works through the arithmetic of recovery, and sets out what a single worst case can and cannot say.
The idea
Follow an account through time and keep a running record of the highest value it has ever reached, the high-water mark. On any date, compare the current value with that mark. If the account is at a new high, the drawdown is zero. If it is 15% below its best level, the drawdown is −15%, whatever the reason and however long ago the high was.
The figure shows both views of SPY since 1993. The upper panel is the growth of $1. The lower panel is the drawdown: it sits at zero whenever the fund is at a record, and hangs below zero the rest of the time. The deepest point of the lower panel is the maximum drawdown.
The lower panel is sometimes called an underwater chart. It makes three things plain that the price chart hides: how deep each fall was, how long each lasted, and how much of the time an investor spent below a level they had already seen.
How it works
Each drawdown has three dates: the peak before the fall, the trough at the bottom, and the recovery, the first date the value is back at the old peak. The time from peak to recovery is the length of the drawdown.
A worked example on SPY
The data: SPY month-end closes with dividends reinvested, February 1993 to October 2026.
- At the end of October 2007 the fund reached a new high. Call the account $100,000.
- Sixteen months later, at the end of February 2009, the same account was worth $49,200.
- Drawdown: 49,200 ÷ 100,000 − 1 = −50.8%. Nothing before or since has been deeper, so this is the maximum drawdown.
- The account did not see $100,000 again until the end of March 2012, 37 months after the trough and 53 months after the peak.
Using the price alone, without dividends, the same fall is 52.2%. Dividends collected during the decline cushion it slightly and bring the recovery date forward.
Every large fall since 1993
| Peak | Trough | Depth | Months falling | Months to recover | Back at the peak |
|---|---|---|---|---|---|
| October 2007 | February 2009 | −50.8% | 16 | 37 | March 2012 |
| August 2000 | September 2002 | −44.7% | 25 | 50 | November 2006 |
| December 2021 | September 2022 | −23.9% | 9 | 15 | December 2023 |
| December 2019 | March 2020 | −19.5% | 3 | 4 | July 2020 |
| June 1998 | August 1998 | −15.3% | 2 | 3 | November 1998 |
| September 2018 | December 2018 | −13.5% | 3 | 4 | April 2019 |
These are the six falls of more than 10% on month-end values. Recoveries took longer than declines in every case, and the two deep ones took years. An investor who put money in at the August 2000 peak waited until November 2006 to be whole, and eleven months later the second fall began.
In the 404 months of the record, SPY was below a previous high in 256 of them, or 63%. It was at least 10% below in 118 months and at least 20% below in 65. Being "down from the peak" is not an emergency. It is where a stock investor spends most of their time.
The arithmetic of recovery
Losses and gains are not symmetric because the gain is earned on a smaller base. Up to about 20% the difference is modest. Beyond that it grows quickly. The time needed grows with it: at a steady 7% a year, recovering from a 20% loss takes about 3.3 years, from 50% about 10.2 years, and from 80% almost 24 years. This is the main argument for caring about the depth of drawdowns and not only about average returns.
Reading it across asset classes
The same calculation on eight funds and a 60/40 mix, January 2005 to October 2026:
| Portfolio | Maximum drawdown | Peak | Trough | Back at the peak | CAGR |
|---|---|---|---|---|---|
| VNQ (real estate) | −68.3% | January 2007 | February 2009 | April 2012 | 6.7% |
| EFA (foreign stocks) | −57.4% | October 2007 | February 2009 | May 2017 | 6.0% |
| IWM (small companies) | −52.5% | May 2007 | February 2009 | February 2011 | 8.4% |
| SPY | −50.8% | October 2007 | February 2009 | March 2012 | 10.7% |
| QQQ | −49.7% | October 2007 | February 2009 | December 2010 | 14.8% |
| TLT (long Treasuries) | −47.3% | July 2020 | October 2023 | Not yet | 3.0% |
| GLD (gold) | −42.9% | August 2011 | December 2015 | July 2020 | 10.1% |
| 60% SPY, 40% AGG | −32.3% | October 2007 | February 2009 | December 2010 | 7.8% |
| AGG (US bonds) | −17.1% | July 2020 | October 2022 | Not yet | 2.9% |
Three observations:
- Different assets have their worst moment at different times. Stocks and real estate bottomed in February 2009. Gold's worst stretch ran from 2011 to 2015. Bonds' came in 2020 to 2023, when interest rates rose from near zero.
- Length can matter more than depth. Foreign stocks took more than eight years from the 2009 trough to regain their 2007 level. Gold was under water for almost nine years. A shallower drawdown that lasts a decade may be harder to live with than a deep one that recovers in two.
- Bonds are not exempt. Long Treasuries lost 47.3% from July 2020 to October 2023, nearly as much as stocks in 2008, and were still below their peak in October 2026. The broad bond fund's 17.1% loss is small beside stocks' but large relative to its 2.9% yearly return.
Drawdown beside volatility
In this sample each portfolio's maximum drawdown was between about 2.5 and 3.8 times its annual volatility: SPY's volatility was 14.8% and its worst fall 50.8%; the 60/40 mix's figures were 9.4% and 32.3%. That is a rough regularity from one period, not a law, but it is a useful warning. A fund with 15% volatility should not be expected to lose "about 15%" in a bad stretch. Losses accumulate over many months, and a year or two of bad months can add up to several times the annual figure.
The Calmar ratio
Dividing the compound annual return by the maximum drawdown gives the Calmar ratio, a return-per-unit-of-pain measure used mainly for trading strategies.
SPY's is 10.7 ÷ 50.8 = 0.21 over 2005 to 2026; the 60/40 mix's is 0.24 and QQQ's 0.30. The original definition uses the most recent 36 months; over longer periods the ratio falls, because the worst drawdown can only get deeper as more history is included. Risk-adjusted return compared: Sharpe, Sortino, Treynor, Calmar and M² compares it with the Sharpe and Sortino ratios.
Where it fails or misleads
It is a single event
Maximum drawdown is determined by one peak and one trough. Two strategies with the same maximum drawdown may differ completely in how often they fall 20%. And one extreme episode in the sample fixes the number for as long as that episode stays in the window: every five-year figure that included 2008 looked alarming, and those starting in mid-2009 looked mild.
Longer records have deeper drawdowns
The maximum can only stay the same or get worse as time passes. A fund with three years of history has a small maximum drawdown mostly because it has not existed for long. Comparing a new fund's worst fall with a 30-year-old fund's is comparing sample sizes. The fair comparison is over identical dates.
The data interval hides the bottom
Month-end data sees only twelve points a year. SPY's fall in early 2020 is 19.5% on month-end values. On daily closing prices the fund fell about 34% between February 19 and March 23, 2020. An intraday low would be lower again. Any drawdown figure should state the bar size it was measured on, and finer data never makes it smaller.
It describes the asset, not the investor
The calculation assumes one purchase and no further activity. An investor adding money every month through a decline has a different, usually shallower, personal drawdown in percentage terms. One who is withdrawing money has a deeper one, which is the subject of Sequence of returns risk: why the order of returns matters.
The worst is always still to come
A maximum drawdown is the worst that has happened in the data. In 2007, the worst post-war fall in US stocks was the 48% of 1973 to 1974 and the similar fall of 2000 to 2002; the next one exceeded both. Planning on the historical maximum as a limit is planning to be surprised.
Compared with its neighbours
| Measure | What it captures | Uses the order of returns? | Weakness |
|---|---|---|---|
| Maximum drawdown | The worst cumulative loss and its dates | Yes | One event; grows with the window |
| Volatility | Typical size of periodic swings | No | Blind to runs of losses |
| Downside deviation | Typical size of shortfalls | No | Same blindness to sequence |
| Value at risk | A loss threshold for one period | No | Says nothing beyond the threshold |
| Time under water | How long the investor waited for a new high | Yes | Depends heavily on the start date |
How Finabase does it
Finabase does not compute a maximum drawdown for a symbol, a portfolio or a backtest. It shows the ingredients of the current drawdown:
- Symbol pages show the 52-week range. Last price ÷ 52-week high − 1 is the drawdown from the past year's high. A stock at $80 with a 52-week high of $100 is in a 20% drawdown, as far as one year of history can tell.
- The screener has columns for the 52-week high and low and a filter for symbols at 52-week extremes.
- The chart at its longest range shows the whole history. Switching the price scale to logarithmic makes equal percentage falls the same height, so the 2000 and 2008 declines can be compared by eye.
- A portfolio's history chart plots its market value or account value over a chosen range. The largest dip on that line is the portfolio's drawdown for the range, to be read off by eye. Deposits and withdrawals move the line too.
- Alerts include trailing-stop types, which trigger when a price falls a set amount or percent from its high since the alert was created: a live drawdown monitor for one symbol.
The Backtest and Strategy tools report gains, trade counts and win rates and do not report a drawdown. The lessons in the strategies section compute it separately for that reason.
Frequently asked questions
What is a normal maximum drawdown for stocks?
For a broad US stock fund, falls of around half have occurred twice since 2000, and falls of 20% to 25% twice more since 2020. Over a holding period of decades, a drawdown of 50% in a stock index is a historical fact to plan around, not a remote scenario.
Is drawdown the same as a loss?
Not exactly. Drawdown is measured from the highest value reached, not from the purchase price. An investor who bought years ago can be in a 20% drawdown and still hold a large gain.
How long do drawdowns last?
For SPY since 1993 the two deep ones lasted 53 and 75 months from peak to full recovery, and the others between 5 and 24 months. Foreign stocks took almost ten years to recover their 2007 high.
Does diversification reduce maximum drawdown?
It has reduced it when the added assets did not fall at the same time. The 60/40 mix's worst fall was 32.3% against 50.8% for stocks alone, with bonds rising in 2008. In 2022 stocks and bonds fell together and the cushion was much thinner.
Why do people compare drawdown with volatility?
Because volatility understates how losses accumulate. A portfolio's worst fall has typically been a multiple of its annual volatility, so the two together give a better picture than either alone.
Summary
Maximum drawdown is the largest fall from a previous high: value over running peak, minus one, at its lowest. For SPY since 1993 it is 50.8% on month-end values with dividends, a fall that took 16 months and a recovery that took 37 more. The measure matters because recovery gains are larger than the losses that precede them and because long periods under water are when plans are abandoned. It is also a single worst case that depends on the window, the bar size and whether money is being added or withdrawn. Read it with its dates and its recovery time, on the same period as whatever it is being compared with.
Check your understanding
5 questions on this lesson. Each answer comes with the reason, and nothing is scored.
An account rises from $100,000 to $150,000, falls to $105,000, then climbs to $140,000. What is its maximum drawdown?
- 0%, because it never fell below the $100,000 starting value
- −30%
- −45%
Show the answer
B. −30% Drawdown is measured from the highest value so far, not from the purchase price. The peak was $150,000 and the low after it $105,000: 105 ÷ 150 − 1 = −30%. The account is still in a drawdown at $140,000.
A portfolio falls 60%. What gain does it need to return to its previous high?
- 60%
- 120%
- 150%
Show the answer
C. 150% Gain needed = 1 ÷ (1 − 0.60) − 1 = 150%. The gain is earned on a base that is only 40% of the old value, so it must be much larger than the loss.
Fund A, launched three years ago, shows a maximum drawdown of 12%. Fund B, launched in 2000, shows 50%. What can be concluded?
- Fund A is much safer
- Nothing yet: the two figures cover different periods, and a longer record can only show a deeper worst fall
- Fund B is badly managed
Show the answer
B. Nothing yet: the two figures cover different periods, and a longer record can only show a deeper worst fall Maximum drawdown grows with the length of the record and depends on whether a severe decline falls inside it. The fair comparison uses identical dates.
On month-end data SPY’s early-2020 drawdown is 19.5%. On daily closes it is about 34%. Why the difference?
- Dividends were paid in March 2020
- Month-end data only sees the last day of each month and misses the low reached within the month
- Daily data double-counts losses
Show the answer
B. Month-end data only sees the last day of each month and misses the low reached within the month The fall bottomed on March 23 and the market had partly recovered by March 31. Coarser data can only understate a drawdown, so the bar size should always be stated.
SPY has been below a previous high in 63% of the months since 1993. What does that imply?
- The fund has mostly lost money
- Being in a drawdown is the usual condition of a rising investment, not a sign that something is wrong
- The drawdown calculation is too strict
Show the answer
B. Being in a drawdown is the usual condition of a rising investment, not a sign that something is wrong A growing asset sets new highs only occasionally and spends the time between them below its record. The fund still compounded at 10.8% a year.
Exercise
One task to do in Finabase. Nothing is graded or saved.
Open SPY’s page and read the last price and the upper end of the 52-week range. Work out the current drawdown: last price ÷ 52-week high − 1. Then put that loss, as a positive decimal, into the recovery formula: gain needed = 1 ÷ (1 − loss) − 1. Which is the larger percentage, the fall or the gain needed to get back to the high?
Open the SPY pageWhat you should find
The gain needed is always the larger one, because it is earned on a smaller base: about 11% after a 10% fall and 25% after a 20% fall. If SPY is at a new high, both are zero. This is the drawdown as far as one year can tell. If the all-time high is more than 52 weeks old, the true drawdown is deeper than this figure.
Sources and further reading
- Malik Magdon-Ismail and Amir F. Atiya, “Maximum Drawdown,” Risk Magazine, vol. 17, no. 10 (2004)
- Terry W. Young, “Calmar Ratio: A Smoother Tool,” Futures (October 1991)
- U.S. Securities and Exchange Commission, Investor.gov — What is risk?
- Finabase chart data — monthly bars and dividends of SPY, QQQ, EFA, IWM, AGG, TLT, GLD and VNQ through October 2026; drawdowns computed by Finabase
