Sized positions
The fastest way to ruin a good edge is to bet too much on a single trade. Sizing — not stock-picking — is what keeps an account alive long enough for the edge to show.
Every approach will be wrong often, sometimes for an uncomfortable stretch. The job of position sizing is to make sure that no single loss, and no short cluster of them, can take you out of the game before that edge has time to work. That means deciding in advance the small, fixed share of the account any one trade may risk, and sizing each position to honour that cap rather than to chase the trades that feel most exciting.
The trap is treating the trade that feels most certain as the one to oversize. Conviction without measurement is just confidence, and confidence is what gets accounts blown up. The discipline is to keep the per-trade risk small enough that being wrong — even repeatedly — is an inconvenience, not a catastrophe.
A worked example: the same call, sized three ways
This is an illustrative example, not a specific recommendation; the numbers are invented to make the arithmetic visible. Take a £25,000 account and a per-trade cap of 0.4%, which is £100 of risk. A setup gives an entry and a stop £0.50 apart per share. The size that risks exactly the cap is the budget divided by the per-share risk:
Two hundred shares is the ceiling for this setup, not a target to beat. Now let conviction place the position inside it. A D-grade call might take a quarter of that — 50 shares, £25 at risk. A B takes the full baseline. An A sits near the top of what the cap allows. What never happens is the cap moving: a strong feeling does not buy 400 shares, because 400 shares risks £200, and £200 was never on the table. The grade redistributes weight under a fixed line; it does not raise the line.
How a conviction grade turns sizing into a measured scale
A flat approach risks the same amount on every trade. A graded one can do better, because it has measured which calls are stronger. On the measured system here, every call carries an A-to-D conviction grade set by where it sits in that model's own return distribution — and the grade-A bar is calibrated per model, around 0.70% average per trade on the fastest model and roughly 6.00% on the one held about one to four weeks. That gives you a built-in sizing scale: lean harder on an A, lighter on a D, all while staying under your per-trade cap. The grade does not replace the cap; it tells you where, inside it, the system itself would weight the position — and because the grade was fixed before the outcome, that weighting cannot be rewritten after a win.
What an A means on each model
For the grade to read the same way across very different holding times, the bar that earns an A is set per model rather than as one absolute number:
| Model | How long it carries | Grade-A bar (per trade) |
|---|---|---|
| Day Trade | opened and closed inside one session | 0.70% average per trade |
| Multi Hour | carried from part of a session to a couple of sessions | 4.50% average per trade |
| Swing Trade | held about one to four weeks | 6.00% average per trade |
| Investing | carried on a long horizon | graded on a long horizon |
An A is the strongest band; D is the lowest still published. Because the bar is calibrated per model, an A on the fastest model (around 0.70% a trade) and an A on the one held about one to four weeks (around 6.00%) both mean “top band for this horizon.” That is what lets the grade act as a sizing scale you can read the same way across every clock. There is no E grade — it was retired so the four-step ladder keeps its meaning.
What a bad version of sizing looks like
The fragile version is built backwards. It sizes from the hoped-for gain rather than the tolerable loss: “this could run, so I will buy a lot.” It lets conviction lift the cap instead of allocating within it, so the “sure thing” quietly carries three times the normal risk — which is precisely the trade that, when it fails, does outsized damage. It sizes to a tight stop just to justify a bigger position, putting the stop where the size wants it rather than where the chart says the idea is wrong. And it varies the per-trade cap by mood, large after a win, larger still after a loss to “make it back.” Each of these unties sizing from the loss and re-ties it to a feeling, which is the one thing position sizing exists to keep out of the decision.
Sizing from the loss almost never blows up an account; sizing from the hoped-for gain eventually does. The grade lives inside the cap, never above it.