Skip to content

Drawdown

Drawdown measures the distance below a preceding peak. Distinguish current loss from maximum drawdown, calculate recovery, and keep cash flows out of performance.

Drawdown is the decline of an investment’s value below its previous highest value. It measures how much ground the portfolio must regain to return to that high, rather than how much it has earned since an arbitrary starting date. A profitable strategy can still be in drawdown: the relevant reference is its own preceding peak.
Here, drawdown is a signed percentage: zero at a high and negative below it. A reported “10% drawdown” often means the positive loss magnitude corresponding to −10%. Both conventions are used; the sign and denominator should be stated.

The peak stays put during a rebound

For positive total-return wealth observations W0,…,WTW_0,\ldots,W_T, including the initial observation, define

Ht=max⁡0≤s≤tWs,dt=WtHt−1.H_t=\max_{0\le s\le t}W_s, \qquad d_t=\frac{W_t}{H_t}-1.
HtH_t is the running high-water mark. It uses only values observed up to time tt, and includes the current value. Therefore dt≤0d_t\le0; reaching an equal or higher peak makes the current drawdown zero. The absolute shortfall Ht−WtH_t-W_t is measured in currency or index points, whereas dtd_t divides that shortfall by the preceding peak.

Consider this synthetic investment wealth index. Observations are daily closes at days 0–5, income is reinvested, and there are no external deposits or withdrawals.

DayWealth indexRunning highSigned drawdown
01001000%
11201200%
290120−25%
3108120−10%
480120−33⅓%
51241240%
At day 3, the portfolio has rebounded from 90 to 108. It is 8% above its initial 100, yet 10% below its high of 120:
108100−1=8%,108120−1=−10%.\frac{108}{100}-1=8\%, \qquad \frac{108}{120}-1=-10\%.

The rebound reduces the shortfall to 12 index points. It does not reset the high-water mark to 108. Nor can the later high of 124 be used to measure the earlier drawdown at 80: a drawdown’s reference peak must already have occurred.

Three snapshots of the synthetic path. At day 2 wealth is 90 against a peak of 120: current drawdown −25%, worst loss so far 25%. At day 3 wealth is 108: current drawdown −10%, worst loss still 25%. At day 5 wealth and peak are 124: current drawdown zero, worst loss still 33.33%.
Three snapshots of the synthetic path. At day 2 wealth is 90 against a peak of 120: current drawdown −25%, worst loss so far 25%. At day 3 wealth is 108: current drawdown −10%, worst loss still 25%. At day 5 wealth and peak are 124: current drawdown zero, worst loss still 33.33%.
Current drawdown is a state that can improve. The historical worst remains in the record after the state has recovered. Percentages shown to two decimal places where needed.
This distinguishes drawdown from maximum drawdown. On the full path, the positive maximum loss magnitude is
MDD=−min⁡0≤t≤Tdt=13≈33.33%.\mathrm{MDD}=-\min_{0\le t\le T}d_t=\frac13\approx33.33\%.

At day 5, current drawdown is zero and MDD is still one third. A new high closes the underwater episode; it does not erase the earlier loss.

What “recovered” means

The day-4 trough of 80 requires a 50% gain to regain the old peak of 120:
12080−1=50%.\frac{120}{80}-1=50\%.

Losing a third and gaining a third do not cancel because the gain starts from a smaller base. For a positive loss fraction L<1L<1, the required recovery return is L/(1−L)L/(1-L).

The first observed recovery is day 5, when wealth reaches 124. The recovery period is four elapsed days if measured from peak to recovery, or one day from trough to recovery. State which interval is intended. If the record ended at day 4, recovery would be unfinished; the elapsed time under water would not be a completed recovery duration. Daily closes also do not identify the exact intraday crossing time.

A cash deposit is not an investment recovery

A drawdown calculation needs a clearly defined wealth series. For strategy performance, use a unitised or otherwise flow-adjusted total-return index. A raw account balance can rise because new money arrives, or fall because its owner withdraws capital.

Suppose a synthetic USD account holds two units initially worth USD 100 each. An investment loss reduces each unit to USD 80, so the account falls from USD 200 to USD 160. A USD 40 subscription then buys 0.5 additional units at the prevailing USD 80 unit value, with no investment price movement at that instant.

Unitisation ledger: initially account value USD 200, two units, USD 100 per unit and zero drawdown. After an investment loss: USD 160, two units, USD 80 per unit and −20% drawdown. After a USD 40 subscription: USD 200, 2.5 units, USD 80 per unit and still −20% drawdown.
Unitisation ledger: initially account value USD 200, two units, USD 100 per unit and zero drawdown. After an investment loss: USD 160, two units, USD 80 per unit and −20% drawdown. After a USD 40 subscription: USD 200, 2.5 units, USD 80 per unit and still −20% drawdown.
An external contribution restores the USD 200 account balance while the unit value stays 20% below its high. Cash-flow timing is explicit; the subscription itself earns no investment return.
The investor may care about both the account balance and investment performance, but they answer different questions. The GIPS Standards Handbook’s discussion of time-weighted returns explains how subperiod returns neutralise external flows and are linked geometrically. That return construction supplies a suitable index on which to measure investment drawdown.
For the accounting and reporting workflow, Backtest Evaluation works through linked returns, drawdown and unfinished recoveries. The full lesson requires membership, with a public preview. It uses positive drawdown magnitudes, so translate the sign when comparing its formulas with this page.

Read the measurement conditions

A useful drawdown number identifies its window, observation frequency and return basis. Resetting the high-water mark at the start of a reporting window differs from carrying in a peak established earlier. Monthly marks can miss a fall and rebound between observations. Fees, financing, income and reporting currency also affect the measured path.

The positive-wealth formula assumes a meaningful positive peak and a wealth series to which relative losses can be applied. Zero wealth corresponds to a 100% loss from a positive peak and cannot recover through a finite percentage investment gain. Negative net equity requires a separately specified treatment; mechanically continuing the usual percentage formula can mislead.

Finally, historical drawdown describes what happened in the measured record. It is not a forecast, a probability of loss or a guaranteed floor. A stop-loss response is a separate trading policy, with execution, funding and opportunity costs. Paleologo’s treatment of high-watermark loss policies makes that distinction practical: reducing exposure changes participation in any subsequent recovery.

Sources

  • Lisa R. Goldberg and Ola Mahmoud, Drawdown: From Practice to Theory and Back Again, version 5 (2016), §2.1 — running maxima, chronological drawdown and observation frequency. Its process drawdown is an absolute positive difference; this page explicitly uses a signed percentage on positive wealth.
  • CFA Institute, GIPS Standards Handbook for Firms (November 2020), discussion of Provision 2.A.24 — external cash flows, subperiod valuation and geometric linking of time-weighted returns. The cash-flow ledger above is an original illustration, not a claim of GIPS compliance.
  • Giuseppe A. Paleologo, Advanced Portfolio Management: A Quant’s Guide for Fundamental Investors (2021), §§9.1–9.2, pp. 150–154 — high-watermark loss baselines and costs of stop-loss policies. All numerical examples and figures here are synthetic.
Last updated
  • drawdown
  • high water mark
  • path dependence
  • performance measurement
  • recovery
← Back to the glossaryBack to top ↑