Ulcer Index: Depth and Duration Together
The ulcer index is the root mean square of an equity curve's drawdown series, so it rises with both the depth of declines and the time spent in them. A maximum drawdown describes one moment; the ulcer index describes every bar.
How it works
The ulcer index answers a question a maximum cannot. Two accounts can share an identical worst decline and be completely different to live with, and this is the figure that separates them.
Compute the drawdown at every bar, square each one, average the squares, take the root. Squaring makes deep readings count disproportionately; averaging across every bar makes duration count at all.
A maximum drawdown is a single observation. It reports the worst moment and says nothing about whether the account spent a week below its peak or a year.
What the number looks like
Take two ten-bar curves that share a 10% maximum drawdown. One touches 10% on a single bar and sits at its peak for the other nine, for an ulcer index of 3.16. The other also reads 5% below its peak on the remaining nine bars, and its index is 5.70.
Divide each index by the maximum and the ratio is the informative part — about 0.32 for the first curve and 0.57 for the second. A curve that spends most of its time near its peak and dips once produces a low ratio; one that spends most of its time underwater produces a high one.
Duration enters through the averaging. Every bar spent below a peak contributes a positive number, so a long shallow decline can score above a brief violent one — which is the behavior the measure was built for.
That makes it a genuine comparison tool. Given two methods with the same maximum drawdown and the same return, the one with the lower ulcer index recovered faster and spent less time underwater — and that difference is invisible in every standard performance table.
In practice
Costs raise it in a way they may not raise a maximum. Fees slow every recovery, which adds bars underwater without necessarily deepening the worst point — so the ulcer index catches an effect the headline figure misses entirely.
It reads the account, not the chart. Volume, price and structure are all irrelevant to it — the input is an equity curve and nothing else.
Comparisons need the same bar frequency. A position-trading method holds through declines that a scalper never sees, so the two produce different figures for reasons that have nothing to do with quality.
A gap loss hits both components at once — it deepens the decline and lengthens the climb back, which is why single events show up so strongly in this measure.
No stop touches it. Stops cap individual trades; this measures what a run of them does to the curve, and the levers that move it are position size and trade frequency.
Frequency feeds it directly. Every round trip pays the spread and commission, so an active method pays a steady toll that shows up as extra time below every peak.
Using it alongside a return figure
The original purpose was a ratio: excess return divided by the ulcer index, giving a reward-per-unit-of- pain figure in the same spirit as a Sharpe ratio but penalizing downside persistence rather than variability in both directions.
That distinction matters because upside variability is not a problem. A method that occasionally jumps higher is penalized by any measure based on deviation and not by this one. If you are comparing two methods with similar returns, the ulcer index is a better tiebreaker than the maximum drawdown — it reflects what holding each one actually felt like, which is the thing that decides whether either gets stuck with.
What the ulcer index is not
It is not a maximum drawdown. It uses every bar, not the worst one.
It is not a volatility measure. Upside movement does not enter it.
It is not comparable across bar frequencies. Same interval or nothing.
And it is not a forecast. It describes a curve that already exists.
How this connects
Its closest relatives are the other downside ratios. The Sortino ratio divides return by downside deviation, which also ignores upside moves but says nothing about time underwater, and the Calmar ratio uses only the worst decline.
Position sizing is the main lever that moves it. Smaller positions shrink every drawdown reading in proportion, and so the index, without changing the method itself.
A losing streak appears here as a long shallow stretch, which is exactly the part of the curve where risk tolerance is tested and plans get abandoned.
Backtesting reports should include it. A method with good expectancy and a high reading may be impossible to hold in practice, and a return figure alone hides that.
It also makes overtrading visible. Extra marginal trades rarely produce a dramatic loss, so they hide from a worst-case figure while adding time below the peak.
It measures persistence, not tail risk. An extreme event the sample never contained is invisible to it, which is why it complements a risk of ruin estimate rather than replacing one.
And investors meet the same idea in a balanced portfolio, where the bond side is held largely to make declines shallower and shorter, and rebalancing is what keeps that mix in place.
When it fails
In a range the index rises with nothing in the trade log to explain it. Small losses and costs keep the curve just below its peak for months, which is precisely the pattern this measure is designed to catch and the one a maximum drawdown reports as unremarkable.
The second failure is comparing across periods of different length. A longer sample includes more declines and usually scores higher, so the periods have to match.
A third is treating a low reading as safety. It is a description of the past, on the sample it was computed from.
A fourth is computing it on closed trades. Open positions are where the time underwater actually happens.
And a fifth is using it alone. It says nothing about return, which is the other half of every comparison.
The original data
On the S&P 500 from 1950 to 2026, 93% of trading days closed below the prior record close. The longest spell ran 1,897 trading days, from 11 Jan 1973 to 17 Jul 1980. The deepest fall, 56.8% from 9 Oct 2007 to 9 Mar 2009, was not recovered until the first close above the old peak on 28 Mar 2013.
The question: Same maximum, double the ulcer. Which system?
Those two records are different kinds of pain, and an ulcer index weighs both. The 1973 spell was the longest and the 2007 fall the deepest; a maximum drawdown reports only the second, while an index built from every day’s reading also counts each of those 1,897 days. Compute the ratio as well as the index: the index alone tells you how much pain there was, and the ratio tells you whether it arrived all at once or never went away.
Related
Drawdown is the underlying measurement this summarizes. Risk management is where the response to a rising reading belongs. And Sharpe ratio is the deviation-based alternative and what it misses.
I track this alongside maximum drawdown, because the maximum can look fine while the year feels awful. One bad two-week stretch sets the maximum; the ulcer figure catches eight months of grinding sideways, which is usually what does the damage.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.