Practical guide · verified against the real thing
Position sizing and the risk of ruin: the maths most backtests skip
In one line: Two strategies with identical signals can have opposite fates based only on how much they bet. Here is why sizing is a survival question, not an optimisation afterthought.
Backtests usually report a return and ignore the question that actually decides whether you live to collect it: how much was at risk on each bet. Position sizing is the part of a strategy that determines whether a normal run of bad luck knocks you out before the edge has time to show. It is a survival question dressed up as an optimisation detail.
Why a good edge can still ruin you
Bet too large and even a profitable strategy will, sooner or later, hit a streak of losses deep enough to wipe out the account — the risk of ruin. Bet too small and you leave most of the edge unused. The uncomfortable truth is that the size which maximises long-run growth is not the size that feels comfortable, and the size that feels exciting is usually past the point where ruin becomes a real probability rather than a tail story.
Sizing is part of the strategy, not a layer on top
A backtest that omits sizing is testing a signal, not a strategy, because the same signal at different sizes produces entirely different outcomes — including total loss. That means any result reported without a sizing rule is incomplete, and the validation checklist should treat "what was the position size and the worst drawdown" as mandatory, not optional. Drawdown, not just return, is the number that tells you whether the sizing was survivable.
The habits that keep you in the game
Cap the fraction at risk on any single bet so that no realistic losing streak is fatal. Size down when uncertainty is higher rather than when conviction feels stronger — confidence is not the same as edge. And judge a strategy by its drawdown profile as much as its return, because a slightly lower return you can actually hold through beats a higher one that ruins you. This is the same "model the thing you are skipping" instinct as transaction costs: the omitted detail is usually where the real risk lives.
Next