Risk and sizing

How to handle drawdowns in a trading system

By Koen van Goch · Updated

Handle drawdowns by planning them before they happen. Decide the deepest drawdown you can live with, size positions so a realistic bad run (worse than the backtest) stays inside it, and write down in advance what would show the system is broken rather than in a normal losing period. Drawdowns are the price of returns. The goal is to make them tolerable, not to avoid them.

Why drawdowns are unavoidable

Any strategy that takes risk will have losing periods. Trend following loses in choppy markets, and a long-only trend system also loses in falling markets. Mean reversion loses when sell-offs keep going, and hedges lose in calm years. A system without drawdowns in its backtest is a warning sign, not a goal.

Set your drawdown before your system

Ask yourself: at what loss would I stop following the rules? Be honest. Most people overestimate their tolerance, so take the number you came up with and, ideally, cut it in half.

Then plan for live drawdowns to be deeper than the backtest's worst one, because the future will contain sequences the past didn't. As a rule of thumb, if a robust portfolio shows a 15% drawdown in the backtest, expect at least 30% live. Single systems on their own will show much deeper drawdowns than that. It's combining systems that do different jobs that generally brings the portfolio's drawdown down.

Size for it

Nobody knows how deep the next drawdown will be. What you can control is your sizing. Lower the position sizing until a realistic worst case, not the backtest's worst case, fits inside your limit. The number of positions and the mix of systems influence it too. See position sizing in trading.

Normal drawdown or broken system?

Decide this before you're in one. Useful questions:

  • Is the drawdown within the range seen in testing, including Monte Carlo tests?
  • Is the system doing what it was designed to do in the current market? For example, a long trend following system losing money in a bear market. If the logic is still sound, it's supposed to lose money there.
  • Has something structural changed, such as costs, liquidity, the market itself or a rule that no longer makes sense?

Only the last one is a reason to stop. Write your stop criteria down and stick to them.

The behavioural part

The biggest risk in a drawdown is you. People change rules at the bottom, switch systems after a losing streak, or turn off the hedge just before it would have paid. Automation helps, and so does having someone to talk it through with who isn't emotionally involved.

All the work on rules, testing and automation is worth nothing if you can't stick to it. Size your systems so you never end up in a position where you feel you have to switch them off, because then everything has been for nothing.

You can see how my own portfolio has gone through drawdowns on my live track record.

FAQ

Common questions

What is an acceptable drawdown for a trading system?

It's personal. It's the deepest loss you would sit through without abandoning the rules, given your capital and goals. Most people overestimate it, so cut your first answer in half. Then plan for a live drawdown of at least twice the portfolio's backtested drawdown, and expect much deeper drawdowns from any single system on its own.

How long can a drawdown last?

For trend-following systems, recovering to a new high can take many months, sometimes more than a year. Look at the longest periods without a new high in your tests, not only the depth.

Should I stop trading a system during a drawdown?

Not because of the drawdown alone. Stop when your pre-written criteria say the system is broken, for example if losses go well beyond the tested range or the logic no longer holds.

Does combining systems reduce drawdowns?

It can, if the systems do different jobs and don't all lose at the same time. Test the combined portfolio, not each system separately. My own portfolio combines trend following, mean reversion and hedge systems for this reason.

Why are live drawdowns often worse than backtested ones?

Backtests show one historical path, often with optimistic costs and some overfitting. The future brings new sequences. Leave a margin: for a robust portfolio, assume at least twice the backtested drawdown.

Want to set your risk before the market sets it for you?

Request a free 30-minute introductory call. We'll look at your systems, your sizing and the drawdown you can live with.