Learn / Risk & Position Sizing

What partial exits do to your expectancy

Scaling out of a winner raises your win rate and lowers your average win. Whether that trade is worth making is an arithmetic question, and the answer is frequently no — but it depends on the shape of your distribution, not on how it feels.

What actually changes

Take a strategy risking 1R with a 3R target, won 35% of the time:

expectancy = (0.35 × 3) − (0.65 × 1) = 1.05 − 0.65 = +0.40R

Now take half off at 1R and move the stop to break-even on the rest. Three outcomes replace two:

  • Never reaches 1R — full −1R. Unchanged.
  • Reaches 1R, then reverses — +0.5R on the half, 0R on the rest. A small win where before there was a full loss.
  • Reaches 3R — +0.5R on the half, +1.5R on the rest = +2R, not +3R.

So the modification converts some losers into small winners and converts every big winner into a medium one.

Whether that is a good trade depends entirely on how often price reaches 1R and then reverses. If that happens often, you are rescuing many trades. If it rarely happens, you are paying a third of your best outcomes for almost nothing.

Working it through

Suppose of 100 trades: 35 reach 3R, 25 reach 1R then reverse, 40 never reach 1R.

Without partial exits — the 25 that reversed are full losses:

(35 × 3) + (25 × −1) + (40 × −1) = 105 − 25 − 40 = +40R  →  +0.40R per trade

With partial exits:

(35 × 2) + (25 × 0.5) + (40 × −1) = 70 + 12.5 − 40 = +42.5R  →  +0.425R per trade

Marginally better, and it required 25% of trades to hit the rescue case.

Now suppose only 10 trades reverse after 1R, and 55 never reach it:

Without: (35 × 3) + (10 × −1) + (55 × −1) = 105 − 65 = +40R   →  +0.40R
With:    (35 × 2) + (10 × 0.5) + (55 × −1) = 70 + 5 − 55 = +20R →  +0.20R

Half the expectancy. Same modification, opposite conclusion, driven entirely by a frequency you have to measure rather than assume.

Why it feels better than it is

Two effects, both psychological rather than arithmetic.

Win rate rises visibly. Trades that would have been losses become small wins. That number is emotionally salient, and improving it feels like improvement — see expectancy vs win rate.

The remaining position feels free. Once the initial risk is “paid for”, the rest is experienced as house money and held with less discipline. Which is itself a cost, because the stop tends to drift.

Regret asymmetry. Giving back an open profit is more painful than never capturing it, so scaling out relieves an anticipated regret. That is a real benefit to your experience of trading; it is not a benefit to the account.

Where it genuinely helps

When your distribution has a fat middle. If price frequently reaches 1R and stalls, partial exits capture value that a binary approach loses.

When execution is unreliable. If large exits move the market, scaling out is slippage management rather than risk management, and it is justified on those grounds.

When it keeps you in trades. If the alternative is closing the whole position from discomfort, a partial exit that lets you hold the rest to target is better than the trade you would otherwise have taken. That is a real argument and worth being honest about.

When the position is oversized. Scaling out of a position that was too big is correcting an error, not managing a trade.

The cost nobody prices

Partial exits corrupt your R statistics. Half at 1R and half at 3R is not a 2R trade in the sense your other records use, because the risk was not constant across the position’s life.

If you scale out, record the components — the fraction exited, at what R, and the final result — rather than collapsing it to a single number. Otherwise your expectancy calculation is averaging things that are not the same kind of thing, and the conclusions you draw from it are unreliable.

How to decide

Do not decide from principle. Decide from your own data:

  1. Over your last sample, how many trades reached 1R and then failed to reach target?
  2. How many reached target?
  3. Compute expectancy both ways, as above.

If you cannot answer 1 and 2, you cannot evaluate the change — which is one more reason to keep a record of what happened after each entry, not just the final result. See logging tool calls for a trading audit trail.

The general shape: partial exits help distributions with many near-misses and hurt distributions where the winners are the whole edge. High-R, low-win-rate approaches — where a minority of trades carries everything — are the ones most damaged by capping the tail.

FAQ

Should I take partial profits?

It depends on how often your trades reach a partial-exit level and then reverse. If that happens frequently, scaling out rescues trades that would have been losses and expectancy improves slightly. If your winners run and your losers fail early, capping the winners while leaving losers at full size reduces expectancy, sometimes by half.

Does taking half off reduce my risk?

Only after the position has already moved in your favour. It does not reduce the risk you took on entry, which was determined by your stop and position size. What it does is convert some potential losers into small winners and all large winners into medium ones.

Why does my win rate improve but my results get worse?

Because those are different measurements. Partial exits reliably raise win rate by turning would-be losses into small gains, while lowering average win by capping the trades that carried your edge. Expectancy — win rate times average win, minus loss rate times average loss — is the number that combines them, and it can fall while win rate rises.

How should I record a scaled-out trade?

Record the components rather than a single R figure: what fraction exited at which R, and the final outcome of the remainder. Collapsing it into one number makes it incomparable to your full-size trades, which quietly corrupts every statistic computed from the set.