Recovery period measures the time needed to regain an earlier wealth peak. Distinguish peak-to-recovery from trough-to-recovery and report unfinished episodes honestly.
Recovery period
A recovery period measures the elapsed time needed for an investment to regain an earlier wealth peak after a drawdown. It adds a time dimension to a loss: an investor may care about how long capital remains below its previous best value as well as how far it falls.
The starting date needs a label. Peak-to-recovery measures the whole episode, including the decline. Trough-to-recovery measures the rebound from its lowest point. Both are useful; an unlabeled “recovery period” leaves a material ambiguity. This article uses peak-to-recovery as its headline duration and reports the trough clock separately.
One endpoint, two clocks
Consider six equally spaced daily observations of a synthetic total-return wealth index, with no external deposits or withdrawals:
Day
0
1
2
3
4
5
Wealth index
100
120
90
108
80
124
The reference peak is 120 on day 1. The rebound to 108 on day 3 does not recover it. The first subsequent observed value at or above 120 is 124 on day 5.
Peak-to-recovery: day 5 minus day 1 = 4 elapsed days.
Trough-to-recovery: day 5 minus day 4 = 1 elapsed day.
Daily wealth 100, 120, 90, 108, 80, 124. Recovery at day 5 is 4 days after the peak on day 1 and 1 day after the trough on day 4. A day 4 cutoff leaves recovery unfinished.
Original synthetic total-return wealth index with no external flows. The two bars use different starting dates but the same first observed recovery. Lines connect daily marks; they do not establish the exact between-observation crossing time.
These are elapsed durations, not inclusive counts of dated marks. There are three strictly underwater observations—days 2, 3 and 4—but the completed peak-to-recovery episode spans four daily intervals. Time under water is closely related; check whether a report means elapsed duration, an observation count or the current age of an unfinished episode.
More formally, select an episode's peak time p, followed by a decline. Its first observed recovery time is
r=min{t>p:Wt≥Wp}.
For a recovered episode, let q be the first time the lowest wealth value between p and r is observed. The two durations are
Tpeak=r−p,Ttrough=r−q.
Use dates or timestamps for irregular observations; subtracting array positions then measures records, not days. Here each step is stipulated to be one day. Equality with the old peak counts as recovery. For repeated equal peaks, choose the last occurrence before the decline; if the trough is tied, the first occurrence is used for the trough clock. These choices should accompany a comparison across reports.
An unfinished episode has no completed recovery time
Now stop the same record at day 4. Wealth is 80 and the old peak remains 120. Three elapsed days have passed since day 1, but recovery has not been observed. Report “unrecovered; 3 days elapsed since peak”, or “at least 3 days” with a censoring flag. Do not record zero, invent a recovery date or quietly remove the episode from an average.
This is right-censoring: observation ends before the event of interest. Averaging only completed episodes can make recovery look shorter because the long, still-open episodes are excluded. A report should identify the measurement window, the number of completed episodes and the open episodes separately. A historical recovery also does not establish that the next drawdown will recover, or how quickly.
Read duration beside depth and the investment policy
Maximum drawdown answers a depth question. Recovery duration answers a timing question. A deep loss can be followed by a quick rebound; a shallower loss can persist. Neither statistic determines the other.
Measure the same investment object throughout. A capital contribution can restore an account balance without recovering an investment loss. For strategy evaluation, use a total-return index adjusted for external flows and disclose fees, currency and valuation frequency. Daily marks identify the first observed recovery; they cannot rule out an earlier crossing and renewed decline between observations.
Recovery is also conditional on the exposure actually held. Cutting a position during a decline changes participation in the subsequent rebound. Paleologo discusses the forgone recovery opportunity alongside the survival purpose and trading costs of stop-loss policies. The recovery time of an unchanged benchmark is therefore not automatically the recovery time of an investor who reduced exposure.
For the broader reporting workflow, Backtest Evaluation connects recovery duration and unfinished episodes to flow-adjusted returns, drawdown depth, costs and research uncertainty. The full lesson requires membership, with a public preview.
PerformanceAnalytics, table.Drawdowns documentation — reports peak, trough and recovery dates alongside separate duration fields. Software record-count conventions should be checked before treating a duration as elapsed calendar time.
Giuseppe A. Paleologo, Advanced Portfolio Management: A Quant’s Guide for Fundamental Investors (Wiley, 2021), §§9.1–9.2, pp. 150 and 154 — high-watermark loss baselines and the opportunity cost of reducing exposure before a recovery.
