A report says maximum drawdown, followed by a percentage sign. What was divided by what? Without that answer, even a perfectly copied number can tell the reader the wrong thing about loss.
This research note is an accounting note. The example below is a deliberately constructed balance path, not market data or an archived strategy result.
Two denominators, two questions
For positive net asset value, peak-relative drawdown compares current NAV with the highest NAV observed so far. Expressed as a negative percentage, it is 100 × (current NAV / running peak NAV − 1). The PerformanceAnalytics implementation constructs a wealth path, includes starting wealth in its running maximum, and divides that path by the running maximum.[4]
A fixed-capital P&L curve answers a different question. If cumulative P&L is expressed as a percentage of starting capital, subtracting its running maximum gives a decline in percentage points of starting capital. It does not divide by the wealth actually reached at the peak. Both measures can be useful, but neither is a substitute label for the other.
Sign conventions also vary. Some reports display drawdown as a negative return; others report a positive loss magnitude. PerformanceAnalytics explicitly supports that presentation choice.[5] State the sign, denominator and valuation frequency together instead of leaving readers to reverse-engineer them.
Conceptual example
Take an account with starting NAV of 100 units, rising to 150 and then falling to 120, with no external cash flows. Its cumulative P&L goes from zero to 50% of starting capital, then to 20%. The decline is 30 percentage points of starting capital. The loss from peak NAV is 20%, because the 30-unit decline is measured against 150, not 100.
Those are two descriptions of the same path, not conflicting observations. Calling the first a 30% peak-NAV drawdown would overstate this particular loss. The direction of a reporting error matters: a pessimistic error can reject a useful idea just as an optimistic error can promote a bad one.
Cheap check
Use that three-point fixture to test the reporting function independently of any strategy. Include starting wealth in the running peak. Require the function to return the declared sign convention, denominator and units alongside the value.
Pass: the fixture produces a 30-percentage-point fixed-capital decline and a 20% peak-NAV loss magnitude. Multiplying every NAV by the same positive constant leaves the peak-relative percentage unchanged.
Fail: both definitions return the same figure, the opening wealth is omitted, or the output supplies a percentage sign without identifying the denominator. Correct the reporting contract before comparing strategies.
Limitations
The clock defines which losses are visible. On the same valuation path, observing only a subset of timestamps can miss peaks or troughs; adding intermediate observations cannot reduce the worst peak-relative loss magnitude. That statement assumes nested samples, identical marks and identical cash-flow treatment. It does not license comparisons between differently constructed curves.
Closed-trade balances omit unrealised losses. Daily marks can miss intraday excursions. Deposits and withdrawals can also masquerade as performance unless the wealth series treats external flows explicitly. None of these historical measures establishes the maximum possible future loss.
Continue from benchmark construction to availability clocks. The reporting guide explains why measurement conventions must travel with evidence.
Sources
[4] https://raw.githubusercontent.com/braverock/PerformanceAnalytics/master/R/Drawdowns.R — PerformanceAnalytics — Drawdowns implementation [5] https://raw.githubusercontent.com/braverock/PerformanceAnalytics/master/R/maxDrawdown.R — PerformanceAnalytics — maxDrawdown documentation and implementation