Max DD: how to measure maximum drawdown and judge a strategy

Contents

1. What is max DD and what does it tell you about a strategy?
2. How do you calculate maximum drawdown step by step?
3. Closed-trade drawdown or intrabar drawdown: which one does your report show?
4. How much drawdown is acceptable?
5. How long does a drawdown last, and how does max DD relate to Sharpe and Calmar?
6. Why will the live max drawdown probably be larger than the backtest?
7. Frequently asked questions
8. References

What is max DD and what does it tell you about a strategy?

Max DD (maximum drawdown) is the largest fall of an account's equity from a peak to the following low before a new peak is reached, measured in money or in percent. It is the one backtest number that describes the worst stretch you would have had to sit through, and it is judged against account size, not against net profit.

A drawdown is any fall of equity below its earlier high point. Think of a hiker on a mountain ridge: every time the path drops below the highest point reached so far, the hiker is in a drawdown, and the deepest of those drops along the whole walk is the max DD. The walk can end at the summit and the deepest drop is still recorded. Three things max DD is not: it is not the fall measured from the starting balance, it is not the biggest single losing trade, and it is not the drop from the final peak only. If you have an equity series or a list of trades, the free drawdown calculator finds the peak, the low and the recovery for you.

Max DD matters because it is the number that decides whether you keep trading. A strategy that finishes its futures backtest with a profit is worthless if, halfway through, the loss was deep enough to empty the account or to make you switch the strategy off. Net profit tells you where the path ended; max DD tells you how far it dropped on the way.

How do you calculate maximum drawdown step by step?

Maximum drawdown is calculated by walking forward through the equity curve, keeping the highest equity seen so far, measuring the gap below it at every point, and taking the largest gap. The running peak is the highest equity value reached from the start up to the current point. Because the peak is carried forward, the method never compares two arbitrary points; it only ever compares the present with the best past.

1. Record equity after each trade (or after each bar, see the next section) in time order.
2. Keep the running peak. At each point, the running peak is the larger of the previous running peak and the current equity.
3. Compute the gap at each point. Gap = running peak − current equity, where the gap is the money you are below the best level so far.
4. Take the largest gap. That number is the max DD in money. Divide it by the running peak at that moment for the max DD in percent.

Two choices change the result and should be written down with it. First, money or percent: a fixed-contract backtest is usually read in money, a compounding one in percent. Second, the sampling point: a drawdown measured only at trade exits can miss a deeper low that happened while a trade was open. That second choice is the subject of the next section.

Closed-trade drawdown or intrabar drawdown: which one does your report show?

The same strategy shows a smaller max DD when drawdown is measured from closed trades only and a larger one when open losses inside each trade are counted, and the second is the one your account statement shows. Closed-trade drawdown is a drawdown series built only from equity after each trade has closed. Intrabar drawdown (also called open-equity or mark-to-market drawdown) is a drawdown series that includes the unrealized loss of the open position at every bar.

The difference is largest for strategies that hold through adverse moves without a stop. A trade can be deep under water for weeks and still close near break-even. The closed-trade view records a small step; the broker records the full dip, and the margin call, if there was one, happened on the dip. For a daily strategy, the intrabar figure taken on daily closes is already much closer to reality than the closed-trade figure; the true low can be slightly deeper still, because prices inside the day can go lower than the close.

Question	Closed-trade drawdown	Intrabar drawdown
What is sampled	Equity at each trade exit	Equity at every bar, open position marked to market
Open losses inside a trade	Not counted	Counted
Size of the max DD	Smaller or equal	Larger or equal
Matches the account statement	No	Yes, up to the bar size used
Use it for	Comparing trade lists quickly	Sizing the account and margin

Before you compare two reports, check which series each one used; a report that shows a flat equity line while a position is open is a closed-trade report. The Performance Report of AlphaPrime (Valynox) draws the equity curve together with its drawdown, so the location and the depth of the worst dip can be read off the same chart as the trades that caused it.

How much drawdown is acceptable?

An acceptable max DD is one that leaves the account above the margin it must hold and below the loss at which you would switch the strategy off, so the judgment depends on account size, not on the drawdown alone. The same dollar drawdown is a nuisance on a large account and a forced exit on a small one. For futures this is sharper than for stocks: after a drawdown, the remaining balance must still cover the margin that futures leverage requires per contract, or the position is closed at the worst moment.

A practical way to set the limit is to write two numbers before looking at any backtest: the largest fall in money you would sit through without changing the rules, and the balance below which margin would force you out. The acceptable max DD is the smaller of the two. Many traders keep the planned max DD at a modest share of the account, often somewhere between a fifth and a third, because a deeper hole takes a disproportionately large gain to climb out of: a 50 percent loss needs a 100 percent gain to recover.

The run below shows how one real drawdown figure reads at different account sizes. It tested ES daily bars from 2019-01-02 to 2025-06-30, long and short, one contract, with $20 commission and $25 slippage per trade. The rule is a stand-in built from a standard indicator, a Stochastic oscillator rule as the only entry and exit, with no stop, target or filter; it was not built to study drawdown itself. The strategy shown is the one candidate the search kept out of 1,024 it tried, chosen for its return against its drawdown. Picking the kept one after the fact flatters it, so read the numbers as a description of one past path, not as what the rule earns or what a reader can expect.

The max drawdown of $60,292.5 is about 64 percent of the net profit, which already says the path was rough. Now read it against an account instead of against the profit. On a $100,000 account the drawdown is about 60 percent of the equity, and the balance left at the low would have to cover one ES contract's margin. On an account twice that size it is about 30 percent; on three times that size, about 20 percent. Same strategy, same trades, three very different experiences. Only one of those three sizes would most traders describe as acceptable, and the backtest cannot tell you which one; your written limit does.

The report shows where that drawdown sits in time and how long it lasted, which is the next question, because a drawdown you can survive in money can still be too long to survive in patience.

How long does a drawdown last, and how does max DD relate to Sharpe and Calmar?

Drawdown duration is a separate pain from drawdown depth, and of the two common ratios only Calmar reads the depth, while Sharpe reads variability. Drawdown duration is the time from an equity peak until equity next makes a new peak, so it includes both the fall and the climb back. Two strategies with the same max DD can differ greatly in how long they stayed under water, and most traders give up on the long one well before it recovers.

The Sharpe ratio is a strategy's return above a reference rate, usually the yield of a short-term government bill, divided by the variability (standard deviation) of its returns; the free Sharpe ratio calculator computes the ratio from a list of daily or monthly returns. The Calmar ratio is a strategy's yearly return divided by its max DD. Sharpe does not see the order of returns: shuffle the same monthly returns into a different order and Sharpe is unchanged, but the drawdown can be far deeper, because losses that cluster together dig a deeper hole than the same losses spread out. Magdon-Ismail and Atiya (2004) show how the expected maximum drawdown grows with volatility and with the length of the period, which is why a longer backtest of the same strategy tends to show a larger max DD. Lo (2002) shows that Sharpe itself depends on the return interval and on how returns are correlated in time, so neither ratio replaces reading the drawdown directly.

When you read a report, note four things together: max DD in money and percent, the date of the peak and the date of the recovery, and the longest flat stretch anywhere in the curve, which is often not the same as the deepest one. The article on how to read Sharpe, Sortino, CAGR and max DD in a backtest report explains how the other ratios fit around these.

Why will the live max drawdown probably be larger than the backtest?

The backtest max DD is one number drawn from one ordering of trades and from one candidate chosen after the fact out of many, so the live figure will probably be larger, and a drawdown plan should assume it. Three effects push in the same direction.

1. Selection. When a search tries many variants and keeps the one with the best return against drawdown, the kept drawdown is small partly because it was selected for being small. Bailey, Borwein, López de Prado and Zhu (2014) show that the best result among many candidates on the same data is inflated, and the flattering applies to drawdown as much as to return. The ES run above kept one of 1,024 candidates, which is why its drawdown figure should be treated as a floor rather than a ceiling.
2. Ordering. The historical trades came in one order. The same trades in a different order, with the losers bunched together, produce a deeper drawdown with the same net profit. The history you saw is one path among many that the same strategy could have produced.
3. Future conditions. The next few years contain market moves that were not in the data. More time means more chances for a new deepest hole, as the previous section noted.

Monte Carlo simulation is a method that reshuffles or resamples the recorded trades many times and records the max DD of each new sequence, giving a spread of possible drawdowns instead of one number; the method itself goes back to Metropolis and Ulam (1949). The result is read as a range: if the single backtest figure sits near the middle of the spread, half of the equally plausible paths were worse. You can run the reshuffle yourself on your own trade list with the free Monte Carlo simulator, and the article on why the real max drawdown from a Monte Carlo backtest is probably worse walks through reading the output.

The practical rule follows from the spread: size the account for a drawdown well above the backtest figure. A common practice is to plan for one and a half to two times the backtest max DD, or for a figure near the worse end of the Monte Carlo spread, whichever is larger, and to check that the balance at that depth still covers margin. Then write a stop rule for the strategy itself: the drawdown, in money or in time under water, at which you switch it off and go back to research. A strategy that reaches twice its backtest drawdown is telling you that the backtest no longer describes it.

Your next step: take the trade list of one strategy you are considering, run it through the drawdown calculator to confirm the peak-to-trough figure and its duration, then reshuffle it in the Monte Carlo simulator and write down the account size at which the worse end of that spread would still leave margin to spare.

Frequently asked questions

What is a good maximum drawdown percentage?

There is no single good figure, because the right limit depends on the account size and on the margin the position needs. Many traders keep the planned max DD at a modest share of the account, often between a fifth and a third, because deeper holes need disproportionately large gains to recover. Decide the limit in writing before you look at a backtest, and judge the backtest against it rather than the other way round.

What is the difference between drawdown and maximum drawdown?

A drawdown is any fall of equity below its earlier peak, so an equity curve has many drawdowns of different sizes. The maximum drawdown is the largest of them over the whole period measured, from a peak to the lowest point before a new peak. A report usually shows both: the current or average drawdown describes normal conditions, the max DD describes the worst case that happened.

Is max drawdown measured in money or in percent?

Both are used, and the choice should be stated with the number. A fixed-size backtest, for example one futures contract throughout, is normally read in money, because the equity never compounds. A compounding account is read in percent of the peak at the time. The percent figure is the one to compare against the share of the account you are willing to lose.

Can a strategy with a small max drawdown still be risky?

Yes, in three ways. The report may show closed-trade drawdown, so the open losses inside trades were never counted. The strategy may be one of many candidates chosen for its small drawdown, which flatters the figure. And a short backtest simply had fewer chances to hit a bad stretch, since expected max drawdown grows with time. Check the drawdown type, the number of candidates tried and the length of the history before trusting a small number.

References

• Magdon-Ismail, M., & Atiya, A. F. (2004). Maximum Drawdown. Risk, 17(10), 99–102.
• Lo, A. W. (2002). The Statistics of Sharpe Ratios. Financial Analysts Journal, 58(4), 36–52.
• Bailey, D. H., Borwein, J., López de Prado, M., & Zhu, Q. J. (2014). Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance. Notices of the American Mathematical Society, 61(5), 458–471.
• Metropolis, N., & Ulam, S. (1949). The Monte Carlo Method. Journal of the American Statistical Association, 44(247), 335–341.