Execution · Risk
Risk Management
Risk is the job. Returns are the by-product of surviving long enough, at the right size, to collect them.
Size before conviction
Position size determines both the gain when you are right and the loss when you are wrong. Most investors set it by how strongly they feel. That is backwards. Feeling is a poor estimator of edge, and edge itself is estimated from noisy data. Size belongs to a risk budget: how much of the portfolio you are willing to lose on this idea if the invalidation hits.
A common working range for active sleeves is 1–2 per cent of capital at risk per idea. That is an example, not a universal law. Full Kelly sizing, which maximises expected log growth in theory, is rarely used by people who survive, because small errors in the estimated edge produce dangerously large recommended bets. Fractional or volatility-targeted sizing sacrifices some theoretical growth for robustness.
From the desk
In 2015, after years of trading my own book with very few losing months, a single position size error took months of gains in a few hours. The direction and the thesis were not the problem. The size was. Experience did not protect against it. A pre-committed size limit would have.
Two drawdown numbers
Financial capacity is what the portfolio can absorb without forcing a change in your life. Psychological capacity is what you will actually sit through. They are frequently different, and the lower one binds. Investors who have never experienced a 25–40 per cent decline in their own money tend to overestimate the second number. Revealed behaviour from 2008, 2020 or 2022 is better evidence than a calm-market questionnaire.
For anyone drawing income, sequence of returns sits on top of both. A 30 per cent decline in year one of retirement, followed by withdrawals, leaves a permanently smaller portfolio than the same decline five years later. The recovery maths is asymmetric.
Correlation is a fair-weather friend
Diversification reduces variance when correlations are moderate. In stress, correlations among risk assets often spike toward one. A portfolio that looked diversified on a historical matrix can behave like a single bet exactly when you need the diversification. That is not an argument against spreading risk. It is an argument for stress-testing the case where the spread fails, and for hard limits on single-name, sector and factor exposure that apply even after a good run.
Bad trade versus bad system
A losing trade, or a short string of them, does not mean the process is broken. Distinguishing noise from signal needs a minimum sample, a defined metric, and a journal that records the thesis and the invalidation at entry — not just the outcome. Most people update in a biased direction: wins are skill, losses are luck. Over time that prevents them from noticing when an edge has actually decayed.
What to do with this
Write your maximum tolerated portfolio drawdown and your maximum risk per idea. Then check whether your current largest position could breach the second number if it hit its invalidation. If it could, the size is already wrong.