Max Drawdown

How max drawdown is measured

Walk through the balance in order. Keep a running high, the peak so far. At each step the drawdown is how far the balance sits below that peak, as a percent of the peak. Max drawdown is the largest of those values.

Take the made-up balances in the picture, thirteen points from $10,000. The running peak reaches $11,200 at point 7. The balance then falls to $9,600 at point 9. The fall is $1,600, which is 14.3 percent of the peak. To get back to the peak from the low, the balance needs to rise $1,600 on $9,600, which is 16.7 percent. A fall is always smaller than the climb back. The guide on drawdown has the full table, and the drawdown calculator does it for your numbers.

The first fall in the example, from $10,900 to $9,800, is 10.1 percent. It is smaller than the one that came later, so it is a drawdown but not the max one.

Why a longer record shows a deeper max drawdown

Each added trade can only add to the list of falls. The measure never improves, so it never goes down. That has a practical meaning. Two bots can have the same real risk and different reported max drawdowns just because one has traded for a year and the other for a month. A shorter record is not safer. It is less tested. When you compare, compare records of the same length and market conditions.

How a bot meets it

• Averaging and ladders. A martingale or grid bot can post a very small max drawdown for a long time and then one very large one. A backtest that has not met a strong trend shows the small figure.
• Open losses. A max drawdown computed from closed trades hides open losses. Measure on equity, including the open trades, or the number is too kind.
• Challenge limits. A loss limit in a prop firm test is a max drawdown the account must not touch. See prop firm drawdown rules.
• Overfitting. A tuned backtest often shows a low max drawdown because the settings were chosen for it. See overfitting.

Percent or money

Quote max drawdown as a percent of the peak, since a fall of $1,600 means one thing on a $10,000 account and another on $100,000. But keep the money figure in view too. A bot that sizes in fixed lots has a drawdown in dollars that stays the same while the account grows, so its percent figure shrinks as the balance rises, and the same bot on a smaller account would show a deeper one. The percent depends on the size you chose as much as on the strategy. Compare bots at the same risk per trade, or scale them to it first, using the position size calculator.

How to use it honestly

Look at max drawdown together with the number of trades and the length of the record. Ask whether you could stay with the bot through a fall of that depth, and through one about half as deep again, because the next one can be worse. Then check how long the longest fall lasted. For the spread of outcomes you could meet, run the Monte Carlo simulator, which reshuffles the trades to show deeper falls than the one history happened to give.

What it does not tell you

It does not tell you the chance of a deeper fall, how fast the balance recovered, or whether the strategy will work from here. It describes the past of one path.