Drawdown

A drawdown is the decline in the value of an investment, a portfolio or a trading program from a previous high point to a subsequent low point, stated as a percentage of that high point. The drawdown is not over when prices stop falling; it ends only when the previous peak is recovered. The Commodity Futures Trading Commission's rules define the base term broadly, at 17 CFR 4.10(k), as "losses experienced by a pool or account over a specified period", and then define the headline version precisely.

That headline version is the maximum drawdown, which the same regulation calls the "worst peak-to-valley draw-down" and defines at 17 CFR 4.10(l) as "the greatest cumulative percentage decline in month-end net asset value due to losses sustained by a pool, account or trading program during any period in which the initial month-end net asset value is not equaled or exceeded by a subsequent month-end net asset value." The clause after "during any period" is the high-water-mark idea written out in full.

Drawdown answers a different question from the dispersion statistics it sits beside. Volatility and standard deviation describe how much returns scatter around their own average, counting moves in both directions. A drawdown counts only the path from a peak to a low, and it is a statement about a specific historical episode rather than about a distribution. Two funds can share a standard deviation and have very different worst falls, because the same amount of scatter can arrive as noise or as one sustained decline.

The duration half is the part most summaries drop, and for a person spending from a portfolio it is often the part that matters. Any drawdown has a fall and a recovery, and the recovery is usually much the longer of the two. The fall is what people remember; the years of being below the old high are what they actually have to live through. A track record that reports only a maximum drawdown of 35 percent, with no indication of whether the recovery took eight months or eight years, has answered half the question.

Federal regulation takes the statistic seriously enough to mandate it. Under 17 CFR 4.25(a)(7), the performance capsule in a commodity pool's disclosure document must show both "the largest monthly draw-down during the most recent five calendar years and year-to-date" and "the worst peak-to-valley draw-down" over the same span, each expressed as a percentage of the pool's net asset value and each identified by the month or months and the year in which it occurred. A registered mutual fund's own standardized risk and return summary is built around a bar chart of annual total returns, average annual returns over one, five and ten years, and the fund's highest and lowest return for a quarter, so an investor comparing two funds usually has to derive a drawdown figure rather than look one up.

Two measurement choices change the answer, and neither is signposted. The first is frequency: the CFTC definition works on month-end values, so a fall that happened and reversed inside a month does not appear at all, while a daily-data version of the same statistic will report a deeper worst case. The second is the window: a maximum drawdown computed over five years and one computed over twenty are different numbers, and the longer window will almost always be worse simply because it contains more chances to be bad. A drawdown figure without its frequency and its period is not comparable to another one.

A last caution runs the other way. Because a maximum drawdown is the single worst thing that has already happened, it is often read as a floor. It is not. It is the worst outcome in the sample, and a sample that has not contained a severe episode will report a comforting number for that reason alone.

Track the running maximum of the value. At every point, the drawdown is the current value divided by that running maximum, minus one. The largest such figure over the period is the maximum drawdown, and the stretch from the peak to the day the peak is finally exceeded is the underwater period.

A hypothetical illustration. A portfolio peaks at $500,000, falls over eleven months to $300,000, and then climbs back. The drawdown is $200,000 divided by $500,000, which is 40 percent.

The recovery is where the arithmetic stops being intuitive. Getting back to $500,000 from $300,000 requires a gain of $200,000 on a base of $300,000, which is 66.7 percent, not 40 percent. The general rule is that a decline of d requires a gain of d divided by one minus d. A 10 percent fall needs 11.1 percent, a 20 percent fall needs 25 percent, a 33 percent fall needs about 49 percent, and a 50 percent fall needs 100 percent. The gap widens fast, which is the whole reason a deep drawdown takes so much longer to repair than it took to create.

Suppose the climb back takes a further thirty-one months. The full drawdown then ran from the old peak to the new one: eleven months down and thirty-one months up, or forty-two months underwater. Only at month forty-two does the drawdown close and a new high-water mark begin.

What the measure is good for

• It is stated in the units investors actually experience, which is how much the balance fell, rather than in units of variance.
• It captures sustained one-directional declines that dispersion statistics smooth over.
• Paired with its duration it describes the whole episode, including the part that tests whether someone can hold on.
• It is directly relevant to anyone who may have to sell during the decline, because a paper drawdown becomes permanent at the moment shares are sold.

What it does not do

• It is entirely backward-looking. The worst fall in the record is not a limit on the next one.
• It depends on the data frequency and the window measured, and both are often unstated, so two published figures may not be comparable.
• A short track record produces a flattering figure by construction, because a fund that has not lived through a severe market cannot report one.
• It says nothing about how the decline was distributed among holdings, so it cannot substitute for looking at what the portfolio owns.