Drawdowns and Position Sizing

Returns get the headlines, but drawdowns determine whether an investor survives to collect them. This guide defines drawdown precisely, explains why the typical and worst cases both matter, and lays out the basic arithmetic of position sizing — how much to commit to a single idea — including the plain-terms logic of risk of ruin.

What a drawdown is

A drawdown is a decline from a peak. In the context of a single signal, it's usually measured from the entry price to the lowest price reached over a defined period afterward: if a signal fired at $100 and the stock touched $82 within the following weeks, the drawdown was 18% — regardless of where the stock eventually ended up. In the context of a whole portfolio, drawdown is measured from the account's high-water mark to its subsequent low.

The signal-level version matters because it describes the experience of following the signal. Two signals can both finish +20% while one never dipped more than 4% and the other spent a month 25% underwater. Reported returns call those identical. Anyone who lived through them knows they aren't. This is why drawdown belongs next to every return statistic in any track record you evaluate: the median drawdown tells you what the typical ride felt like; the worst drawdown tells you what the hardest ride demanded of you.

Median versus worst

Averages hide tails; drawdowns live in the tails. A method whose average worst dip is 10% may still produce individual signals that fall 40%. The honest summary uses at least two numbers: the median (half of signals dipped less, half dipped more — the typical experience) and a bad-case figure (the worst, or a high percentile — the experience you must be built to withstand). When a record quotes only an average drawdown, assume the worst cases are considerably deeper, because they almost always are.

There's also a time dimension: how long a position stays underwater. A 15% drawdown recovered in three weeks and one that takes eighteen months differ in opportunity cost and in the psychological tax they levy. "Time under water" is worth asking about alongside depth, even though depth is the number usually published.

The arithmetic that makes drawdowns dangerous

Drawdowns are asymmetric. A 10% loss needs an 11% gain to recover. A 25% loss needs 33%. A 50% loss needs 100% — the position must double just to break even. This convexity is the mathematical root of risk management: deep losses are not proportionally worse than shallow ones, they are disproportionately worse, because the recovery required grows faster than the loss itself. Avoiding the deepest drawdowns matters more than squeezing out extra return, for arithmetic reasons alone.

Position sizing in plain terms

Position sizing is deciding what fraction of the account a single idea gets. Its purpose is to make sure no one signal — and no plausible cluster of bad signals — can inflict a drawdown the account can't recover from.

The common approach works backward from a risk budget. Decide the most you're willing to lose on one position if it goes wrong (a small fraction of the account is the usual scale — a fraction of a percent to a couple of percent for most disciplined frameworks). Estimate how far the position could fall before you'd exit — for a stock, guides like the ATR in Popular Technical Indicators help calibrate what "normal" movement looks like. Divide the risk budget by that distance and you get a position size. Small positions relative to the account feel unexciting; that's the point. They convert any single mistake from an event into a statistic.

Sizing also interacts with volatility. A stock that swings 6% in a normal day and one that swings 1% should not get the same allocation if the goal is equal risk, because the first can do three weeks of the second's damage in an afternoon.

Risk of ruin, in plain terms

Risk of ruin is the probability that a run of losses shrinks an account to a point where recovery is impractical. It depends on three things: how much is risked per position, how often losses occur, and how deep they run. The uncomfortable truth is that even a genuinely good method — one that wins more often than it loses — carries a meaningful ruin probability if positions are sized too large, because losing streaks are guaranteed eventually. A method with, hypothetically, a 60% win rate will still produce five consecutive losses about one time in forty sequences of five — and across a long record, that sequence will arrive. Sized so each loss costs 1% of the account, that streak costs about 5%. Sized so each costs 10%, it costs about 41% — and the recovery math above takes over from there.

The defense is boring by design: risk a small, fixed fraction per idea, and the worst realistic streak becomes survivable. Investors rarely blow up because their analysis was wrong once. They blow up because one wrong idea — or one bad month — was allowed to matter too much.

Putting the two together

Drawdown statistics tell you what a method's signals have historically demanded; position sizing decides how much of that demand lands on your account. Read together, they answer the only risk question that matters: if this method has its normal bad stretch while I'm following it, what happens to me? A method with deep typical drawdowns can still be followed — with smaller positions. A method with shallow drawdowns can still ruin an account — with oversized ones. The Glossary collects the terms used here; the companion piece on judging published numbers is How to Read a Track Record.

Educational material only — not investment advice.