Why position sizes drift upward

2026-12-20

Almost nobody sits down and decides to risk too much. Position sizes drift upward through a set of individually reasonable steps, and by the time the size is wrong, each step that got you there still looks defensible.

Recognising the mechanisms is more useful than resolving to be disciplined, because five of the six are not about discipline at all.

1. The percentage moved without you

Fixed fractional sizing recomputes from the current balance. After a good run, 1% is a larger number — correctly, and that is the method working.

The drift is when the percentage creeps up alongside it. A good stretch makes a larger percentage feel earned, and the change is never made explicitly. It happens as a series of small roundings.

The check is mechanical: log the realised risk percentage per trade and look at the series. It is either flat or it is not.

2. Averaging down, one reasonable step at a time

The most common single path. Adding to a losing position lowers the break-even, which is genuine arithmetic — and it increases the size of a position whose thesis is unconfirmed.

Each addition feels like managing. The aggregate is a position larger than anything you would have approved at the outset, concentrated in the trades that kept going against you. See averaging down vs dollar-cost averaging.

3. Correlation you did not count

Three positions at 1% each in instruments that move together is closer to one 3% position. Per-trade sizing cannot see this, because each calculation considers one trade.

You did not oversize any position. You oversized the book, and the sizing formula had no term for it — correlation and position limits.

4. The trade count, which nothing constrains

A 1% per-trade rule bounds one decision, not a day. Ten trades at 1% is nearly 10%; twenty is over 18%.

The count rises exactly when it should fall — after losses, in volatile conditions, and when a model is generating candidates faster than you can evaluate them. See risk per trade vs risk per day.

5. The tight stop that bought a bigger position

Position size is inversely proportional to stop distance. A tighter stop means a larger position for the same stated risk.

So a stop moved in to “reduce risk” doubles the notional, doubling the fees and slippage — and if leverage crosses the point where liquidation arrives before the stop, the real loss is the margin rather than the 1% you sized for. See why tight stops can cost more and the leverage and stop-loss table.

6. Recovering

The one that is about psychology, and it is the most destructive because it increases f exactly when the balance is smallest.

The arithmetic is unforgiving: fixed-fractional sizing protects you by shrinking positions during a drawdown, and raising the risk percentage inverts that protection at the worst possible time. It is the most common path to ruin for accounts that had a genuine edge — risk of ruin.

What actually works

Resolving to be disciplined addresses one of six mechanisms. Structure addresses all of them.

A per-account position cap evaluated before submission. Catches drift regardless of which mechanism produced it — including the ones you did not notice.

A daily loss limit enforced in code, with closing orders exempt. Bounds the sequence that per-trade limits leave open.

A trade count limit. Crude, and it binds before the losses rather than after.

A correlation group cap. The only thing that sees mechanism 3.

Logging realised risk percentage per trade. Makes mechanism 1 visible, which is otherwise invisible.

A cooldown after consecutive losses. Removes the window in which mechanism 6 operates, without requiring you to notice it is operating.

All of them have to be code, not intention. The moment each binds is the moment you have a good-sounding reason to exceed it — which is the same argument as why MCP needs an approval layer, applied to yourself rather than to a model.

The AI-assisted version

Worth naming because it is new and it accelerates mechanism 4.

A model produces as many candidates as you ask for. Per-trade limits do not scale down as candidate volume scales up, and an approval step under high volume stops carrying information.

The specific control: cap proposals per session, not only executed trades.

FAQ

Why do I keep taking positions bigger than I planned?

Usually not one decision but several small ones — a risk percentage that crept up, adding to a loser, correlated positions counted separately, more trades than planned, or a tightened stop that doubled the notional. Each looks reasonable in isolation, which is why a cap evaluated in code catches them and vigilance does not.

How do I stop myself from oversizing?

Structural limits rather than resolve: a per-account position cap, a daily loss limit, a trade count limit, and a correlation group cap — all evaluated before an order is submitted. Log your realised risk percentage per trade so drift becomes visible rather than gradual.

Is it ever right to increase position size?

Deliberately and in advance, yes — a risk percentage set from your drawdown tolerance and expected losing streak is a legitimate parameter to revisit. What does not work is increasing it during or after a drawdown, which is when the protective effect of fixed-fractional sizing is most needed.

Why is averaging down so common?

Because the arithmetic that motivates it is real: adding at a lower price does lower the break-even. What the arithmetic omits is that it also enlarges a position whose thesis is currently unconfirmed, and applied consistently it puts your largest size in the trades that went furthest against you.