How I Size and Stress-Test a Position

The first number I set has nothing to do with the trade itself: it is the fixed amount of the portfolio I am willing to lose if the position is wrong. Once that is fixed, the size of the position falls out of the maths. This inversion — risk first, size second — is the foundation everything else rests on.

Keeping risk per trade constant is what allows a string of losses to be survivable. A book that risks a steady fraction per idea can absorb a losing streak; a book that sizes by conviction or gut feel eventually meets the trade that takes too much.

A stop placed at a round number or a fixed percentage ignores how the instrument actually moves. I anchor stops to volatility — typically a multiple of Average True Range — so the level sits beyond the market's normal noise rather than inside it.

The point of a volatility-based stop is to be wrong only when the thesis is genuinely wrong, not when ordinary fluctuation happens to clip an arbitrary line. Once the stop distance is set this way, it feeds directly back into position size: wider volatility means a smaller position for the same dollar risk.

I measure every trade in R — multiples of the initial risk. A trade that makes three times what it risked is a +3R trade whether the position was large or small. Expressing results this way strips out size and currency and exposes the only thing that matters over a long run: the distribution of R-multiples the strategy actually produces.

Thinking in R also reframes the planning stage. Before entry I want to see the trade's reward-to-risk and the win rate it would need to break even. If a setup needs to be right two times in three just to tread water, that is worth knowing before any capital is committed, not after.

An individual trade's return is only half the story. What matters to the book is its contribution — the trade's return scaled by how much of the portfolio it represents. A large move on a small allocation and a small move on a large allocation can mean the same thing to the bottom line.

This is exactly how the live Trade Log on this site reports — percentage return on the position, and then the portfolio-weighted contribution. It keeps the focus on what each position is actually doing to total capital rather than on headline percentages in isolation.

A single historical path tells you what did happen, not what could have. Before I lean on a strategy I run it through randomised simulation — reordering and resampling outcomes to see the range of equity curves and, crucially, the drawdowns it can produce. The worst path in a Monte Carlo run is almost always deeper than the one backtest you happened to observe.

The goal is not a precise forecast — it is to size and prepare for a drawdown you can actually sit through. Most strategies do not fail because the edge disappears; they fail because the operator abandons them inside a drawdown they never imagined was normal.