Guide · Drawdown
The short answer
A 20% loss needs 25% to recover, a 50% loss needs 100%. The asymmetry is the whole argument for small size.
A loss and the gain that undoes it are not the same size, and the gap widens fast. Losing 50% requires 100% to recover; losing 90% requires 900%.
The reason is that the gain is computed on a smaller base. Down 50%, every subsequent percentage point is worth half of what it was.
| Drawdown | Remaining | Gain to recover |
|---|---|---|
| 5% | 95% | 5.3% |
| 10% | 90% | 11.1% |
| 20% | 80% | 25.0% |
| 30% | 70% | 42.9% |
| 40% | 60% | 66.7% |
| 50% | 50% | 100.0% |
| 70% | 30% | 233.3% |
| 90% | 10% | 900.0% |
Up to about 20% the two columns are close enough that recovery is a matter of a normal good run. Past 30% they separate, and past 50% the required gain is larger than most strategies produce in years.
Every risk-management rule reduces to this table. Small size is not caution for its own sake; it is keeping the account in the region where the two columns are still nearly equal, because that is the region a normal winning run can undo.
It also explains why maximum drawdown, not total return, is the number to compare strategies on first. Two strategies returning 40% a year are not equivalent if one did it through a 15% drawdown and the other through 55%.
Worth knowing
The percentages are only half of it. A 40% drawdown also takes time to recover, and time spent recovering is time the capital is not compounding. Two years back to break-even is two years of returns that never existed.
The arithmetic says a 30% drawdown is recoverable. Whether you recover from it is a separate question, because most people change the strategy partway down — usually by increasing size to make it back faster, which is the row-by-row worst response the table permits.
Decide the maximum tolerable drawdown before trading, then size for it. See how much to risk per trade for the connection.
The recovery table says what gain is needed. It does not say how long that gain takes, and for a real account the time is often the larger cost.
A strategy earning 15% a year needs roughly two years to recover a 25% drawdown and just over five to recover a 50% one — assuming it performs at its average the entire way, which is exactly what a strategy in drawdown is demonstrably not doing.
Five years of returns that produce no new capital is the real meaning of a 50% drawdown, and it is why the number to compare strategies on is not return but return relative to the worst drawdown they took to get it.
Two strategies returning 40% a year are not equivalent if one took a 15% drawdown to get there and the other 55%. The standard way to say that in one number is the MAR ratio.
Above 1.0 is respectable for a trading strategy and above 2.0 is rare. The ratio is crude — it uses a single worst-case number that depends heavily on the sample — but it forces the comparison onto the axis that decides whether a strategy is holdable.
Maximum drawdown is the deepest hole in the sample you happen to have. It is an extreme-value statistic, and extreme-value statistics from short samples systematically understate the tail.
A strategy with two years of history and a 20% maximum drawdown has not demonstrated that 20% is its limit. It has demonstrated that 20% was the worst of the particular sequence it encountered — and a different ordering of the same trades would have produced a different figure, often a considerably worse one.
Worth knowing
Plan for a drawdown roughly one and a half times the worst you have seen. That is not pessimism, it is what re-ordering the same trades tends to produce.
Planned drawdown is a number chosen in advance from the arithmetic. Experienced drawdown is that number arriving slowly, in a sequence that offers several opportunities to abandon the plan.
The gap between them is where most position-sizing discipline is lost, and it is not an arithmetic problem — the maths was settled before the trade. Deciding the stop-trading level in advance, and writing down what would justify changing it, is the only part of this that can be done while thinking clearly.
gain required = loss ÷ (1 − loss). A 10% drawdown needs 11.1%, 20% needs 25%, 50% needs 100% and 90% needs 900%. The asymmetry is why small position sizes matter more than entry signals.
At a 15% annual return, a 25% drawdown takes about two years and a 50% drawdown about five — and that assumes the strategy performs at its average throughout, which a strategy in drawdown is demonstrably not doing.