Maximum Drawdown Explained: The Real Risk No Average Captures

Maximum drawdown is the worst peak-to-trough decline a portfolio (or strategy, or index) has suffered over a given window. It is the single number that answers the question every investor actually wants answered: how bad has this gotten in the past, and how bad could it get again? Volatility tells you how much returns jiggle around the mean. Maximum drawdown tells you how deep the hole has been – and that is the number that drives behavior at the bottom.

TL;DR: Maximum drawdown (MDD) is the largest cumulative loss from a prior high-water mark to the next trough, measured over a sample window. A 30 percent MDD means the portfolio was, at its worst point, 30 percent below its previous peak. Two portfolios can have the same average return and the same standard deviation, yet very different MDDs – because volatility treats upside and downside the same, and drawdown only counts the downside. Used together with the Calmar ratio, MDD is one of the most honest risk numbers in finance.

The formal definition

Let V(t) be the value of a portfolio at time t. The drawdown at time T is the percentage decline from the running peak up to that point:

D(T) = (max0 ≤ t ≤ T V(t) − V(T)) ÷ max0 ≤ t ≤ T V(t)

The maximum drawdown over the window [0, T] is just the worst (largest) drawdown observed anywhere in that window:

MDD(T) = max0 ≤ τ ≤ T D(τ)

Plain English: walk forward through the time series, keep track of the highest value you have seen so far, and at every step compute how far the current value sits below that running high. The deepest gap is the maximum drawdown. It is always non-negative, expressed as a percentage of the peak, and depends entirely on the window you measure over – a longer history almost always uncovers a deeper drawdown.

Two related quantities matter as much as the depth:

  • Drawdown duration – how long the portfolio stayed below the prior high-water mark before recovering it. The MDD’s duration is sometimes called the “underwater period.”
  • Recovery time – how long the climb back from the trough to the prior peak took. Investors who needed the money during the underwater period never got the recovery.

A worked example with real numbers

Suppose a portfolio starts at $100 and prints these end-of-month values over a year:

Month Value ($) Running peak ($) Drawdown (%)
0 100.00 100.00 0.0
1 108.00 108.00 0.0
2 115.00 115.00 0.0
3 110.00 115.00 4.3
4 95.00 115.00 17.4
5 82.00 115.00 28.7
6 90.00 115.00 21.7
7 100.00 115.00 13.0
8 112.00 115.00 2.6
9 118.00 118.00 0.0
10 122.00 122.00 0.0
11 126.00 126.00 0.0
12 130.00 130.00 0.0
Illustrative path. Drawdown column = (running peak − value) / running peak. MDD = 28.7 percent at month 5. Recovery to a new high happens at month 9, so the underwater period is four months.

The portfolio finished the year up 30 percent. Annualized volatility is roughly 27 percent. Yet at one point it was down 28.7 percent from its peak – which is the experience the investor actually lived through. The headline calendar return and the headline standard deviation each hide that fact in their own way. Maximum drawdown does not.

The underwater equity curve

The cleanest way to visualize MDD is the underwater equity curve: a plot of the running drawdown over time, expressed as a non-positive percentage. Every new equity high prints a zero on the chart; everything else is some amount of red. The bottom of the deepest valley is the MDD; the width of that valley is the drawdown duration.

Equity curve with an underwater drawdown panel A top panel showing portfolio value rising, falling, and recovering, with running peaks marked. A bottom panel shows the drawdown as a shaded area below zero, deepest at the trough and returning to zero at the recovery point.

Equity curve and underwater drawdown

value

Prior peak Trough Recovery (new high)

0% drawdown

Max drawdown (deepest point) Underwater period

Schematic. The top panel shows a typical equity curve; the bottom panel shows the same path expressed as drawdown from the running high. The maximum drawdown is the deepest point of that shaded valley.

Historical maximum drawdowns by asset class

Maximum drawdown is meaningful only in context. The table below shows headline peak-to-trough declines in nominal price terms for major asset classes during named historical episodes. Total-return drawdowns (with dividends reinvested) are usually a few percentage points shallower; the price-index figures are the ones investors saw on the screen at the time.

Asset / index Episode Peak Trough MDD
S&P 500 (price) Global financial crisis Oct 9, 2007 Mar 9, 2009 −56.8%
Nasdaq Composite (price) Dot-com bust Mar 10, 2000 Oct 9, 2002 −77.9%
S&P Composite (price) Great Crash Sep 1929 Jun 1932 ≈ −86%
S&P 500 (price) COVID-19 shock Feb 19, 2020 Mar 23, 2020 −33.9%
S&P 500 (price) 2022 rate shock Jan 3, 2022 Oct 12, 2022 −25.4%
Long-term Treasuries (TLT ETF) 2020-2023 bond rout Aug 2020 Oct 2023 ≈ −52%
Gold (London PM fix) Post-1980 bear Jan 1980 Aug 1999 ≈ −70%
Bitcoin 2021-2022 cycle Nov 2021 Nov 2022 ≈ −77%
Sources: FRED S&P 500, FRED Nasdaq Composite, Damodaran NYU Stern historical returns, LBMA gold price history. Index levels are price-only and rounded; total-return drawdowns are typically a few points shallower.

The S&P 500 in the global financial crisis

The single most useful drawdown to internalize is the one investors lived through most recently in scale: the S&P 500’s 56.8 percent peak-to-trough decline from October 9, 2007 to March 9, 2009. The price index closed at 1,565.15 on the peak day and at 676.53 at the trough – a 56.78 percent decline that took roughly four years to fully recover on a total-return basis.

S&P 500 price index, January 2007 to June 2009 A line chart showing the S&P 500 climbing to a 1565 peak in October 2007, falling to a 676 trough in March 2009, and beginning to rebound by June 2009, with the peak and trough labeled.

S&P 500 price index: Oct 2007 peak to Mar 2009 trough

1600 1300 1000 700 400

Jan 07 Oct 07 Jul 08 Mar 09 Jun 09

Peak 1,565 (Oct 9, 2007)

Trough 677 (Mar 9, 2009)

Maximum drawdown −56.8%
Source: FRED – S&P 500 daily close. Peak and trough values from the S&P Dow Jones Indices closing values, verified against Federal Reserve Economic Data.

The 56.8 percent figure is what gets quoted – but it is only one of three numbers that matter. The drawdown itself took 17 months to develop. The full recovery to the prior nominal price high (back through 1,565.15) was not achieved until late March 2013, a five-and-a-half-year underwater period in price terms. An investor who started withdrawing during the underwater period – retirees, endowments, anyone with mandatory distributions – locked in losses that the eventual rebound never replaced. This is what is meant by sequence-of-returns risk, and it is why drawdown matters more than the same-sized loss spread out as gentler volatility.

Why drawdown beats volatility as a risk number

Standard deviation – the volatility number – is symmetric. It treats a 4 percent up week and a 4 percent down week as equally risky, because both are equally far from the mean. For an investor with no behavioral or liquidity constraints, that is a defensible view of risk. For everyone else, it is misleading. Three reasons:

  • Path matters. A portfolio that loses 50 percent and then doubles ends the year flat in arithmetic terms but lived through a 50 percent drawdown. Volatility does not capture how the year actually unfolded.
  • Investors capitulate at troughs. The empirical drag from investors selling at the bottom is well documented in the annual DALBAR QAIB study: the gap between fund returns and investor returns is largest in the wake of deep drawdowns.
  • Withdrawals turn drawdown into permanent loss. For anyone drawing from a portfolio (retirees, foundations, leveraged funds facing margin calls), a deep drawdown sells units at the worst price and removes them from the rebound. Volatility blind to direction misses this entirely.

This is why hedge funds, CTAs, and risk-parity managers all report MDD alongside Sharpe and volatility. It is also why the Sharpe ratio is often paired with a downside-aware cousin (Sortino, Calmar, or maximum drawdown itself) in any serious due-diligence report.

The Calmar ratio: return per unit of drawdown

The most common way to fold MDD into a risk-adjusted return number is the Calmar ratio, introduced by hedge-fund analyst Terry Young in 1991. Its definition is just:

Calmar ratio = Compound annual return ÷ |Maximum drawdown|

A strategy that compounds at 12 percent a year with a 30 percent MDD has a Calmar ratio of 0.4. One that compounds at 8 percent with a 10 percent MDD scores 0.8 – the same edge per unit of pain, twice as palatable. Calmar is usually computed over a trailing 36-month window to keep the MDD figure relevant to the current regime.

Two cousins of Calmar use the same idea with slightly different denominators:

  • Sterling ratio – return divided by the average of the worst few annual drawdowns. Smoother but more arbitrary.
  • Burke ratio – return divided by the square-root of the sum of squared drawdowns. Penalizes many small drawdowns less than one giant one.

Common mistakes when interpreting MDD

1. Treating the historical MDD as the worst case

The MDD is the worst observed loss in the sample, not the worst possible loss. Extending the window almost always uncovers a deeper drawdown – the S&P 500’s “global financial crisis” 56.8 percent figure looks gentle next to the roughly 86 percent decline of 1929-1932. Always quote the sample window with the number.

2. Confusing daily, weekly, and monthly MDDs

The same series can show very different MDDs depending on the sampling frequency. Daily lows are deeper than monthly closes; intraday troughs are deeper still. A 30 percent monthly MDD on a strategy can become 45 percent on daily marks – meaningful for anyone with stop-losses, margin lines, or behavioral limits.

3. Ignoring the recovery duration

A 20 percent drawdown that recovers in two months is materially different from a 20 percent drawdown that takes seven years to recover (the Nasdaq Composite, post-2000). Drawdown depth is one axis; drawdown duration is the other. Both belong in any risk report.

4. Forgetting that MDD scales with leverage

Leveraging a strategy 2x roughly doubles both the volatility and the MDD – more so if the strategy uses path-dependent products like daily-rebalanced leveraged ETFs, where the path makes the drawdown worse than the multiplier alone implies. This is the volatility-drag math that crushes long-dated returns on 3x ETFs in choppy markets.

5. Comparing across strategies with mismatched windows

If one fund reports a 12 percent MDD over a calm three-year window and another reports a 28 percent MDD over a window that included 2008, the right comparison is not “fund A is half as risky” – it is “the two numbers are not comparable.” Always reset windows to a common period before drawing conclusions.

Related concepts and what to learn next

  • Sharpe ratio – the standard risk-adjusted return number, using volatility as the denominator. See the explainer on what it captures and what it misses.
  • Sortino ratio – Sharpe but with the denominator restricted to downside deviation. A natural step between Sharpe and Calmar.
  • Value at Risk (VaR) – a quantile-based loss number (“we are 99 percent confident the day’s loss will be no worse than X”). VaR misses the tail; MDD captures one specific realized tail.
  • Sequence-of-returns risk – the retirement-finance corollary: when withdrawals coincide with drawdowns, the order in which returns arrive can decide whether a portfolio survives.
  • Stress testing – the supervisory practice of imposing hypothetical drawdowns on a portfolio or balance sheet. The Fed’s CCAR/DFAST framework for banks is the most visible example.

The short version

Maximum drawdown is the simplest, most honest answer to “how bad has it gotten” – the worst peak-to-trough loss in a return series, expressed as a percentage of the prior high. It is asymmetric (down-only), path-dependent (the recovery duration matters as much as the depth), and almost always more behaviorally relevant than the same series’ standard deviation. Combine it with the Calmar ratio for a clean return-per-unit-of-pain score, quote it with its sample window, and use it as the gut-check on any strategy whose Sharpe looks too good to be true.

Sources & further reading

Disclosure: This article was produced with AI assistance and reviewed before publication. It is for informational purposes only and is not investment advice.

Leave a Comment