Two Take-Profits and a Break-Even: The Math Behind Partial Exits

Two Take-Profits and a Break-Even: The Math Behind Partial Exits

21 十一月 2026, 01:12
Xiang Jun Jiang
0
4
Most EAs I audit have one take-profit and one stop-loss. That is the simplest design, and simplicity is good — but it quietly forces you into a single trade-off: either you take profit early and leave money on the table on the days the move runs, or you let it run and give back a large part of the gain on the days it reverses.

A two-stage exit with a break-even move solves part of that problem. This article walks through the arithmetic, the failure modes, and a few implementation details that decide whether the idea actually helps or just looks good on a chart.

WHAT PARTIAL EXITS ACTUALLY CHANGE

Your expectancy per trade is the same whether you close it in one piece or in four pieces, as long as the total size and the total exit average are the same. Splitting a position does not create edge. What it changes is the distribution of outcomes:

- it lowers variance
- it converts some "gave it all back" trades into "small win" trades
- it raises the cost base slightly, because each partial close pays spread and commission again

So the honest way to judge a two-stage exit is not "does it make more money" but "does it improve the ratio of drawdown to return, at an acceptable cost per trade".

A SIMPLE MODEL

Take a EURUSD trade with a 20 pip stop and a 100 pip target, full size. Over a sample of 200 trades the hit rate on the target might be 30%. That gives roughly:

- 60 winners at +100 pips
- 140 losers at -20 pips
- net = 6000 - 2800 = +3200 pips per 200 trades

Now split the same trade: 50% closes at 40 pips, the remaining 50% runs to 100 pips, and once the first stage fills we move the stop on the remainder to break-even. In the same sample:

- of the 60 trades that eventually reached 100, all 60 filled the first stage: 60 x 40 x 0.5 = 1200 pips
- those 60 second halves added 60 x 60 x 0.5 = 1800 pips
- the break-even stop on the remainder converts some of the 140 losers into scratches; say 40 of them reached 40 pips before reversing and then closed at break-even instead of -20. Those 40 now return 40 x 0.5 = 20 pips each instead of -20, which is a swing of 1600 pips across the sample
- the remaining 100 losers are untouched: -2000 pips

Net: 1200 + 1800 + 800 - 2000 = +1800 pips. That is less than the single-target version in this example, and it is supposed to be. What you bought for those 1400 pips is a much flatter equity curve: the worst clusters of losses are gone.

Run the same numbers with a lower hit rate (target reached only 20% of the time) and the two-stage version pulls ahead, because the "reached first target, then reversed" bucket grows and the break-even conversion pays for itself repeatedly.

THE COST NOBODY MODELS

Every partial close is a separate market order. On a 0.10 lot EURUSD position, closing half costs roughly the same commission as the original entry. If your round-trip cost is 1.2 pips equivalent per lot side, a two-stage exit adds about 1.2 pips per trade on the second half.

That means the first target must be far enough from entry to cover: spread + slippage + the extra commission, plus enough to make the risk you removed worth it. On 5-digit FX, a first target of 10-15 points is almost always wasted — the cost eats most of it.

Rule of thumb I use: first target at least 8x the spread, and at least 1.5x the stop distance divided by the number of stages.

WHY BREAK-EVEN IS NOT FREE

Moving the stop to entry feels protective. Two things go wrong with it:

1. Stop hunting near round numbers. If the price is at your entry, so is everyone else's, and a wick through entry takes you out before the move continues.
2. You cut off the trades that would have paid for the losers.

The fix that usually survives backtesting: place "break-even" slightly in profit, not at exactly entry. On gold, 20-30 points past entry; on EURUSD, 2-4 pips. It converts "zero" into "small positive", which is what actually keeps the curve smooth.

The other fix is timing: move the stop only after the first stage has filled, not when price merely touches the first target. If you move it on touch and the first stage has not filled, you have changed the trade without taking any money off the table.

IMPLEMENTATION DETAILS THAT DECIDE THE RESULT

If you code this yourself in MQL4 or MQL5, these are the parts that break:

- Track stages per ticket. A position keeps its ticket when you partially close it, so a simple "closed half" flag keyed by ticket is enough. Without that flag, a per-tick check will close a second slice on the next tick.
- Handle gaps. If price jumps through the second target, you must mark the first stage as done at the same time, otherwise the position is left with a stale "stage 1 pending" state.
- Respect the minimum lot. If the remaining size after a partial close is below the broker minimum, close the whole position instead; otherwise you get "invalid lots".
- Respect the stop level. Setting a stop closer to price than the broker's stop level fails silently in some wrappers. Check the distance before modifying, and retry on later ticks rather than dropping the idea.
- Do not re-issue the same modify every tick. If the stop is already at or better than your break-even level, skip.

WHAT TO MEASURE AFTER A MONTH

Do not judge the change by profit alone. Compare these five numbers before and after:

1. maximum drawdown, in currency and in R
2. average loss per losing trade
3. percentage of trades that end within +/- 0.1R (the "scratches")
4. profit factor on the second stage only
5. total commissions

If drawdown fell and the average loss fell, while profit factor on the second stage stayed above 1.3, the change is doing its job. If commissions grew faster than the drawdown improvement, tighten the first target or drop the second stage entirely.

CONCLUSION

Partial exits do not create an edge, and break-even stops do not remove risk — they move it. What they do reliably is reshape the outcome distribution, and that is worth something when you trade someone else's money or need to survive a bad month without changing your lot size.

The version that survives contact with a live account is usually the boring one: one meaningful first target, a stop moved slightly past entry rather than exactly to it, stage tracking per ticket, and a decision made on drawdown, not on the best backtest.