An exit strategy specifies when and how to reduce open exposure. Entry price, position size, closing fills and costs jointly determine the result. A useful plan lets you reconstruct which condition occurred, which volume should have been closed and what the broker actually executed.

Write the exit hierarchy before opening

Separate the initial protective stop from profit-taking and a time limit. Record whether each instruction closes the whole position or a specified part, what remains active after a fill and who manages the remainder. If conditions occur close together, the actual sequence matters. Each subsequent instruction must reflect already executed closes so it does not inadvertently open opposite exposure.

For example, keep the initial SL active; after a half-position close, verify the filled and remaining units before updating protection. After a full close, check and cancel orphaned pending orders. This is an operating procedure, not evidence that closing half is superior. The wider process appears in managing an open position step by step.

A better first exit need not give a better whole-policy total

Looking only at the first position, B nets 3.9 R and A 1.9 R. With a one-position limit, A’s earlier exit frees capacity for the second signal. A totals 1.9 + 2.9 = 4.8 R in this fictional scenario; B totals 3.9 R. The 0.9 R difference comes from different possible entries, not the first exit.

One occupied slot changes signal availabilityOne timeline spans zero to three hours. Policy A holds its first position from zero to one hour, then its second from two to three. B occupies the only slot for all three hours and skips the second signal. Fictional net totals: A 4.8 R, B 3.9 R.Limit: 1 position; signal 2 at 2 hAB1.9 R2.9 R3.9 R0 h1 h2 h3 hNet total: A 4.8 R; B 3.9 R
Bars show capacity occupancy. Results are assigned for this invented sequence, not estimates of early-exit performance.

Compare exit quality for a common entry separately from the whole state machine with its position limit. Retain signals skipped through occupied capacity, their validity rules and all accepted-cycle costs in a complete test. This does not prove A better: a different second-position result could reverse the comparison. Do not add its win to an actual report if it was never executed.

Assumptions: Entirely invented scenario: max 1 position; the new signal is valid only at 2 h. Fixed R = 100 USD. A exits its first at 1 h for +2 R gross and accepts the second for +3 R gross; B holds its first to 3 h for +4 R and skips the second. Each complete cycle costs 0.1 R = 10 USD; no other costs.

Five ways to close a position

An exit rule must specify an action rather than just a method name. The five approaches below need explicit parameters and execution assumptions. Their names do not establish a ranking of performance.

Five methods: the missing implementation details
Fixed take profitSpecify the level, units, order type and fill conditions. Reaching a chart level need not fill an order.
Trailing stopSpecify activation, update rules and behavior when the terminal stops. A moved stop does not guarantee its fill price.
Scaling out partiallySpecify units at each target and rules for the remainder. Profit on one part need not make the whole trade profitable.
Time-based exitSpecify completed bars or a timestamp/time zone and how to handle closed markets. Time passing does not itself submit an order.
Signal-based exitSpecify the condition on available data, decision time and executable instruction. An opposite signal has no automatic performance advantage.

In MT5, position SL and TP operate on the server; its built-in trailing stop updates SL in the terminal. If the terminal stops, the last SL remains but is no longer moved. Other tools can behave differently: see how a trailing stop works. Do not infer a smaller drawdown you are willing to tolerate from one partial profit. For order distinctions, see the take-profit glossary entry.

The tension between letting winners run and protecting them

A fixed target and a trailing rule produce different exposure after the first target. Either can give the higher result on a particular path. Compare specified rules including missed fills, reversals and costs; a few longest-held winners do not establish what determines every trend strategy’s annual result.

A possible rule closes half at one initial price-risk unit and trails the rest. Define that initial unit once and retain the denominator after the partial close. A realized first portion can be offset by the remainder and fees. In MT5 netting, a smaller opposite deal reduces the position; in hedging, use the appropriate position-closing instruction, since a new Sell can open a separate short.

Entry-price protection is not necessarily net break-even

A stop at the entry price is not risk-free: an ordinary stop may fill worse, and costs can make even an exact entry-price close negative. Whether an earlier move helps or harms a defined strategy is a testable question, not a universal rule about market noise.

If using a confirmed higher low as the update condition, define confirmation and the precise offset below it in advance. Keep that rule distinct from time-based or fixed-distance updates. A chart point can invalidate a chosen premise without proving one update method superior.

Cutting losers at a pre-defined invalidation

A stop can express a pre-defined invalidation or a separate risk rule; its name does not determine the level. Record the trigger, units and order mechanics before opening. Neither a support break nor the trigger guarantees the execution price or a maximum loss. The difference between a stop loss and a take profit explains their distinct roles.

Win rate is one component of expectancy. In a fictional distribution with a 30% chance of +3 R and 70% of −1 R, the gross mean is 0.3 × 3 − 0.7 × 1 = 0.2 R. An assumed 0.3 R cost on every trade makes the mean −0.1 R. These assigned probabilities are not research results; observed estimates are uncertain and do not promise future profitability.

A step-by-step worked example, entirely hypothetical

For a hypothetical calculation, assume long entry 1.0850, stop 1.0820, and half closed at 1.0880: the first half gains 30 pips. Assume a later confirmed higher low at 1.0866, a stop just below it at 1.0865, and an actual second-half fill exactly there: 15 pips. The volume-weighted average is 22.5 pips, or 0.75 of the initial 30-pip distance, before costs. A stop below a low of 1.0865 would instead give less than 15 pips; actual fills are not guaranteed.

For a separate fictional USD account example, buy 20,000 EUR at Ask 1.1000 with initial Bid 1.0998 and SL Bid 1.0970. Define R₀ = 20,000 × (1.1000 − 1.0970) = 60 USD before commission. This initial price-distance unit is not a maximum loss. One pip is 0.0001: 2 USD for the full volume, 1 USD for half.

Assume commission of 0.0001 USD per EUR at opening and closing, no minimum charge per order, and all exits before funding: 2 USD opening plus 2 USD total closing, or 4 USD regardless of splitting. Each table row is a separate path with the specified actual fills, not six exits available on one path.

Executed closesGross USDNet USDNet / R₀
20,000 at 1.103060560.9333
10,000 at 1.1030; 10,000 at 1.101545410.6833
10,000 at 1.1030; 10,000 at 1.10901201161.9333
10,000 at 1.1030; 10,000 at 1.09700−4−0.0667
20,000 at SL 1.0970−60−64−1.0667
20,000 at worse stop fill 1.0960−80−84−1.4000

The half exits at 1.1030 and 1.1015 give 30 + 15 = 45 USD gross, not 45 pips on the full volume. Their weighted move is 22.5 pips, or 0.75 R₀ gross. Half at the target and half at the initial stop instead give zero gross and −4 USD net. Exact full closure at 1.1000 also nets −4 USD; the model’s net-zero sell price would be 1.1000 + 4/20,000 = 1.1002. Actual funding, minimum fees and fills can change this. Bid/Ask sides are already in executed-price differences: do not deduct spread again.

The same post-entry fictional Bid OHLC, O 1.1000 / H 1.1040 / L 1.0960 / C 1.1010, can hide opposite outcomes. Assume continuous paths, no gaps and exact fills at the first of SL 1.0970 or TP 1.1030. O → H → L → C reaches TP first and nets +56 USD; O → L → H → C reaches SL first and nets −64 USD. The 120 USD difference comes from event order, not the final Close. Without intrabar data, mark the case unresolved or apply a convention fixed before testing rather than selecting the winning path afterwards.

What to do tomorrow

  1. Record the initial stop, every partial-close unit count, update conditions, time limits and precedence. Verify fills and cancel orphaned pending orders after full closure.
  2. Fix the definition of R and cost assumptions before comparing variants. Count a partially closed position as one entry opportunity, not several independent trades.
  3. To isolate exits, retain the same entries, volumes and initial protections. Separately test the whole policy: earlier closing may free capacity for a different set of later entries, as the unchanged diagram shows.
  4. Report net results, holding time, drawdown and deviations. Separate method selection from evaluation on untouched data; twenty trades or mechanical compliance alone do not establish an edge.