Position Management and Scaling

The short answer

Adding, reducing, trailing and taking partials are not neutral housekeeping. Each one moves your expectancy and your risk in a specific, predictable direction, and none of them can turn a negative-expectancy method into a positive one.

"Risk-free trade" is the phrase to be most suspicious of. Moving a stop to break-even does not remove risk; it exchanges one set of risks for another, and it raises the chance of being taken out of a trade that was going to work.

The only question that matters

For any management action, ask: what does this do to the distribution of outcomes, and did I decide it before or after the position existed?

The second half of that question does most of the work. A scaling schedule written in the plan and tested is a rule. The same action taken at 15:40 because the position is uncomfortable is a discretionary override wearing a rule's clothes, and it will not appear as one in your journal.

Adding to a position

Adding to a winner (pyramiding)

You are long, price moves in your favour, you add.

• What it does to exposure: increases it, at a worse average price than your first entry.
• What it does to expectancy: increases the size of the largest winners and increases the size of the occasional large loss, because you now hold more units when a reversal arrives.
• What it costs: an additional round trip in fees, and a worse blended entry.

Pyramiding is defensible when a method's edge comes from long continuation moves, which are rare and large. It is destructive when the method's edge comes from a high hit rate on small moves, because you have added size precisely where the remaining move is shortest.

The non-negotiable part: the combined position's risk against the current invalidation level must still sit inside your per-trade limit. In practice that means the added units come with a raised stop for the whole position, or the add is smaller than the original. If adding units leaves you risking more than your stated limit, you have not scaled in — you have abandoned the limit. The risk/reward calculator will show you the blended figures before you commit to them.

Adding to a loser (averaging down)

You are long, price moves against you, you add to improve the average entry.

• What it does to exposure: increases it while the thesis is, by your own definition, performing worse than expected.
• What it does to planned risk: increases it. Same invalidation level, more units, bigger loss.
• What it does to your record: raises your win rate slightly and lengthens the left tail. Most accounts that end suddenly end this way.

The improved average entry is cosmetic. It makes the position look better in the platform's P&L column while making the outcome of being wrong strictly worse.

There is a narrow legitimate version: a planned, pre-specified entry ladder where the full ladder was sized as one position from the start and the invalidation sits below the entire ladder. That is not averaging down. That is one position entered in pieces, and the distinction is whether the total was sized before the first unit was bought.

Reducing and partial exits

Taking part of the position off at a defined level does two things at once.

The numbers below are invented for teaching. They describe no real instrument and are not a trade idea.

Take a method with a 40% hit rate that either loses 1R or runs to +3R. Expectancy is (0.4 × 3) − (0.6 × 1) = +0.6R per trade.

Now add a rule: exit half at +1R, move the rest to break-even, and let the remainder run to +3R. In the same idealised world, a winner returns (0.5 × 1) + (0.5 × 3) = 2R instead of 3R. Losers still cost 1R, but some trades that would have been full losers now end at roughly zero because the first half was banked.

Suppose that converts one in six losers into a scratch. Expectancy becomes (0.4 × 2) − (0.5 × 1) = +0.3R. Lower. The win rate went up and the equity curve got smoother, and the per-trade average fell by half.

That arithmetic is not an argument against partials. It is an argument for computing them against your own numbers, because the direction of the effect depends entirely on where your expectancy comes from:

• If the method's return is concentrated in a few large winners, partials cut into exactly the trades that pay for everything else.
• If the method's return comes from frequent modest moves that often give back their gains, partials capture something real.

You cannot tell which case you are in by introspection. You can tell by running both versions over the same sample, which is the point of holding a backtest to the standard described in backtesting and forward testing.

Moving the stop to break-even

This is the most common management action and the most misunderstood.

What it does:

• Removes the planned loss if the stop fills at its trigger price.
• Leaves gap risk, slippage risk and venue risk fully intact. A stop is an instruction to transact at market once touched, not a guarantee of price.
• Leaves your capital, margin and attention committed.
• Raises the probability of exiting before the thesis resolves, because price frequently retraces through the entry area before continuing. The tighter the stop, the more of ordinary noise sits inside it.

That last point is the cost nobody prices. Every break-even stop moved too early converts a trade that would have worked into a scratch, and those scratches come out of the same right tail that pays for your losers.

The defensible version ties the move to a market event rather than to a P&L threshold: move the stop when a new structural level has formed beneath the position — a higher low, a completed range, a level defined in advance — not when the position happens to be up by some multiple of the risk. One is a statement about the market; the other is a statement about your comfort.

Why "risk-free trade" is a bad phrase

Strictly, it is wrong: gap and slippage risk remain, and the fill on a stop can be far from its trigger.

The bigger problem is behavioural. Once a trade is labelled risk-free, three things tend to follow: the position is held past its exit rule because "it costs nothing to see," the next position is sized larger because the last one "didn't use any risk," and the scratch outcomes stop being counted as costs at all. None of those are conclusions the arithmetic supports.

A more accurate description: the position now has a planned loss of approximately zero, at the price of a higher probability of early exit. It is longer and less satisfying, and it does not lead anywhere unhelpful.

Writing a scaling schedule you can test

If you scale, it belongs in the plan as a schedule specific enough to backtest:

total position risk : 1.0% of equity, fixed
entry 1             : 60% of units at trigger condition A
entry 2             : 40% of units only if condition B occurs before invalidation
invalidation        : single level, below both entries, set before entry 1
partial exit        : 50% of units at +1.5R, limit order
stop after partial  : moved only when a higher low forms above entry
final exit          : trail below each new higher low, or time stop at session close

Every line is a condition, an order type or a number. Nothing in it can be satisfied by how the chart feels.

What to do with this

Pick one management action you currently take by instinct — most people should start with the break-even stop — and do three things.

Write it as a condition rather than a P&L threshold. Run your existing journal twice: once with the action, once without, using the trades you already have. Then keep the version that survives the comparison, and specify it in the plan so the next hundred trades test the same thing.

Management is the part of the plan that gets improvised most and reviewed least. It is also, because it is applied to every open position, the part where an unexamined habit compounds fastest.