What does a drawdown actually measure?

Volatility is an average of wobbles. Drawdown is a specific, dated, remembered event: how far below its own high-water mark the portfolio fell. It is the number people actually experience, and it behaves quite differently from the ones in the last unit.

The arithmetic

At every point in time, compare the value to the highest value reached so far:

Drawdown at time t = (value at t ÷ running peak up to t) − 1

The running peak only ever ratchets upward. When the portfolio is at a new high the drawdown is exactly 0%; otherwise it is negative. The maximum drawdown is the most negative value that expression ever took over the window.

Worked example. A portfolio runs 100 → 130 → 78 → 130 over three years.

• At 130, running peak is 130, drawdown = 0%.
• At 78, running peak is still 130 → 78 ÷ 130 − 1 = −40%. That is the maximum drawdown.
• Back at 130, drawdown returns to 0%.

Note that the portfolio ended exactly where its peak was, with a total return of +30% from the start. The 40% hole is invisible in the start-and-end numbers and dominated three years of lived experience.

The recovery asymmetry

Falls and recoveries are not symmetric, because they act on different bases:

Gain required to recover = x ÷ (1 − x), where x is the drawdown depth.

Check the worked example: 0.40 ÷ 0.60 = 66.7%, and indeed 78 × 1.667 = 130. The curve steepens brutally at the deep end — which is why deep drawdowns are structurally different events from shallow ones, not merely larger ones.

Why it disagrees with volatility

Volatility is symmetric and averaged across every period; drawdown is one-sided, path-dependent, and an extreme rather than an average. Two portfolios can share an identical annualised volatility while one drifts down in a long grinding sequence and the other bounces around a rising line. The first accumulates a far deeper drawdown from the same "risk" number.

This is why serious performance reports carry both. Neither is the risk; each is a different projection of it.

In the data

A drawdown series is a running maximum over closing prices, and the close is the boundary of what it can see. Monday 24 August 2015, the S&P 500 fund:

Close to close the fund fell about 4%. At its low that morning it had been down almost 8% from Friday's close. A drawdown built from closes records the 4% and never sees the 8% that a holder with a stop order or a margin call actually met.

Try it now

1. The chart below is twenty years of a broad fund, drawn on adjusted closes — the only line a drawdown can honestly be computed on. The traded price is unadjusted, and mixing the two across a split or a dividend breaks the running maximum in a way that is very hard to see afterwards.

2. Find the deepest valley on it and Measure from the peak that preceded it down to the bottom. That percentage is the maximum drawdown for the window, and the date it happened is worth noting. Then build the same thing arithmetically on paper if you want to prove it: a running-maximum column, then value ÷ running maximum − 1, and its minimum should land where your measurement did.
3. Apply the recovery formula to that depth. How large a rise was needed to get back to even? Write it down as a fact about that period — not as an expectation about any future one.