Partial exits on M5 gold scalps without turning winners into scratches
Posted: Sat Sep 05, 2026 4:28 pm
Partial exits on M5 gold are where many scalpers quietly turn a clean winner into a scratch — or worse, a late loser. I still scale out. I just refuse to do it by feel mid-candle.
The problem is not taking profit. The problem is random first targets that leave the remainder oversized into noise, then moving the stop to entry before the structure that justified the trade has done anything. That is how +1.5R ideas become +0.2R with stress.
My fixed partial framework for XAUUSD M5:
1. Entry only with a pre-written invalidation. Stop is beyond that micro structure, sized so full risk is known before the first click.
2. First partial is mechanical: 40–50% at +1R measured from entry to stop. No “it looks extended.” R is the unit.
3. After partial one, stop moves to breakeven only if M5 has printed a clear higher low (long) or lower high (short) that protects the idea — not because I am nervous.
4. Second partial (optional): another 25–30% at a pre-marked M15 level or liquidity pool. Remainder trails behind structure, not behind an arbitrary tick trail.
5. If price never reaches +1R and invalidates, full stop. I do not save” the trade by cutting early into noise unless a hard session rule forces flat (news, daily stop).
What I refuse:
- Scaling out 80% at +0.4R because the candle paused. That is fear dressed as management.
- Leaving full size to “let it run with no partial plan when my edge is mean-reversion-ish reclaim, not trend riding.
- Moving stop to entry immediately after partial one on a still-developing impulse. You convert a valid stop into a free scratch lottery.
A practical example: long gold after a London sweep-and-reclaim. Stop under the sweep low. At +1R I take half. I do not move stop to entry until M5 holds a higher low above the reclaim. If gold stalls and revisits the reclaim zone, the remainder can still be wrong — that is allowed. Partial one paid for process; remainder is still risked on structure.
Journal tag I use: partial_plan_followed (Y/N). Wins that ignore the plan score lower on process than losses that followed it. Over a month, expectancy improves when partials are boring and repeatable.
If your gold winners keep dying as scratches, write the partial percentages and the breakeven rule before the session. Execute the plan. Do not renegotiate R while the candle is open.
The problem is not taking profit. The problem is random first targets that leave the remainder oversized into noise, then moving the stop to entry before the structure that justified the trade has done anything. That is how +1.5R ideas become +0.2R with stress.
My fixed partial framework for XAUUSD M5:
1. Entry only with a pre-written invalidation. Stop is beyond that micro structure, sized so full risk is known before the first click.
2. First partial is mechanical: 40–50% at +1R measured from entry to stop. No “it looks extended.” R is the unit.
3. After partial one, stop moves to breakeven only if M5 has printed a clear higher low (long) or lower high (short) that protects the idea — not because I am nervous.
4. Second partial (optional): another 25–30% at a pre-marked M15 level or liquidity pool. Remainder trails behind structure, not behind an arbitrary tick trail.
5. If price never reaches +1R and invalidates, full stop. I do not save” the trade by cutting early into noise unless a hard session rule forces flat (news, daily stop).
What I refuse:
- Scaling out 80% at +0.4R because the candle paused. That is fear dressed as management.
- Leaving full size to “let it run with no partial plan when my edge is mean-reversion-ish reclaim, not trend riding.
- Moving stop to entry immediately after partial one on a still-developing impulse. You convert a valid stop into a free scratch lottery.
A practical example: long gold after a London sweep-and-reclaim. Stop under the sweep low. At +1R I take half. I do not move stop to entry until M5 holds a higher low above the reclaim. If gold stalls and revisits the reclaim zone, the remainder can still be wrong — that is allowed. Partial one paid for process; remainder is still risked on structure.
Journal tag I use: partial_plan_followed (Y/N). Wins that ignore the plan score lower on process than losses that followed it. Over a month, expectancy improves when partials are boring and repeatable.
If your gold winners keep dying as scratches, write the partial percentages and the breakeven rule before the session. Execute the plan. Do not renegotiate R while the candle is open.