A gold strategy recovered its peak. How long was it underwater?

One loss percentage leaves the timeline unfinished

A strategy summary can report its largest decline from a previous high and still conceal how long the recovery took. Suppose two entirely hypothetical equity indices both fall from 100 to 90. One returns to 100 after five trading sessions; the other needs sixty. Their maximum percentage drawdowns are both 10%. They would make very different demands on a researcher deciding how much observation time is needed to judge whether a method has changed.

This distinction has a formal research basis. Vedernikov, Liesiö and Seppälä study drawdown duration separately from drawdown magnitude. Zhang and Hadjiliadis examine the time between a preceding maximum and a specified drawdown threshold. Those papers motivate separate clocks; their model results are not evidence about a CyGuru strategy. The method below is an original reporting design for a gold research notebook, not a measured performance record.

Source notes: [1] Vedernikov, Liesiö & Seppälä · Portfolio models for optimizing drawdown duration · [2] Zhang & Hadjiliadis · Drawdowns and the Speed of Market Crash

Choose the value series before computing the maximum

For a strictly positive, consistently valued index V at time t, keep the running peak H as the largest earlier or current value. The percentage below that peak is 100 × (1 − V/H). A maximum drawdown is the largest such reading inside the selected window. Starting the window later can remove an earlier peak or loss, so two reports with different starting dates need not describe the same risk experience.

For portfolio work I would first build a cash-flow-adjusted return index. An investor deposit must not masquerade as recovery and a withdrawal must not masquerade as a trading loss. State whether values include open positions and which prices value them. A closed-trades-only balance can remain flat while a large open loss develops. If the record contains only proposed entries and target messages, leave portfolio drawdown unavailable: position sizes, simultaneous exposure, executable marks and costs are still missing.

Keep three dates and name the clock

My episode record would retain the last occurrence of the prior high, the lowest value before recovery, and the first later observation reaching or exceeding that high. The first-to-second interval describes the descent; the second-to-third describes recovery from the trough; the first-to-third describes the complete underwater episode. Repeated equal highs need a consistent convention. I use the last equal high before the decline so a flat period at the peak does not inflate time underwater.

Count elapsed calendar time as well as sampled trading intervals. A Friday-to-Monday interval can occupy only one step in a daily trading series while requiring capital to remain committed over a weekend. Daily marks also miss any recovery and renewed decline occurring between observations. Label the resulting measure daily-observed drawdown, not an intraday maximum. Do not silently compare a minute-sampled strategy with a daily-sampled benchmark.

A recovery percentage uses a different denominator

In the invented 100-to-90 example, a 10% gain from the trough reaches only 99. Reaching 100 requires 100/90 − 1, or about 11.11%. This arithmetic does not predict a rebound; it explains why the percentage needed to recover exceeds the percentage previously lost. A method that reports a loss against the peak and a gain against the trough must retain those different denominators.

The longest episode is not necessarily the deepest. I would publish the largest depth and longest duration as separate records, each with its own dates. A moderate persistent decline may dominate the time statistic while a brief shock dominates the depth statistic. Collapsing them into one attractive ratio can erase the distinction that matters to an investor with a fixed liquidity deadline.

An unfinished recovery remains an open observation

If the series ends below its previous high, recovery time is not zero and is not known. Report time underwater through the cutoff with recovery still open. Excluding such episodes from an average creates a favourable sample of declines that have already ended. Before comparing strategies, agree on the valuation frequency, cost treatment, cash-flow adjustment and common observation window, then retain the unfinished episodes beside the completed ones.

For the Research Lab, the practical next step is to preserve the full marked-value series and the rules that created it. Overlapping gold setups must be represented as simultaneous exposure where applicable; counting their target hits cannot reconstruct the account path. This article supplies no historical equity curve, broker fills, commissions, spread, slippage or P&L. Its image is conceptual. The separate weekly report examines gold’s macro backdrop; this method examines what would make a strategy recovery claim reproducible.