Scaling In and Partial Close: What Adding and Taking Profits Really Change

Scaling means not entering or exiting all at once. Scaling in is adding to a trade in the same direction once price moves your way. Scaling out, or a partial close, is closing part of the position before the target. Both sound like ways to make more and lose less, but once you work the numbers step by step, they change the shape of your results, not whether your system has an edge. This article looks at average entry, total risk, win rate and expectancy with real numbers.

We use R for the initial risk of a trade. 1R is the distance from entry to the first Stop loss. If you risk 100 dollars, a 2R profit is 200 dollars.

Scaling in and averaging down go in opposite directions

Both add orders in the same direction. The difference is timing.

• Scaling in (pyramiding) adds once price has moved into profit, and moves the Stop loss up with it. The average entry gets worse, but total risk stays under control
• Averaging down adds when price moves against you. The average entry improves, but total risk grows with every order, in the direction the market is telling you that you are wrong. The mechanics and our test results are in Grid, Martingale and averaging down

A classic pyramid adds smaller orders at each level. The biggest piece sits at the best price, so later entries at worse prices carry less weight.

Example: a 3-level pyramid that stays within plan

Balance 10,000 dollars, risk no more than 1% or 100 dollars. Buy EURUSD, where 1 lot = 10 dollars per pip.

1. Level 1 Buy at 1.1000, SL at 1.0950, 50 pips away. Lot size = 100 ÷ (50 × 10) = 0.20 lot. If stopped out, you lose 50 × 0.20 × 10 = 100 dollars
2. Level 2 Price reaches 1.1060. Move the SL of every order to 1.1010 first, then add 0.10 lot. If the new SL is hit, order 1 makes +10 pips × 0.20 × 10 = +20 and order 2 loses 50 × 0.10 × 10 = −50, a total of −30 dollars, within the 100 budget. Average entry = (0.20 × 1.1000 + 0.10 × 1.1060) ÷ 0.30 = 1.1020
3. Level 3 Price reaches 1.1120. Move every SL to 1.1070, then add 0.05 lot. If stopped out, order 1 makes +140, order 2 +10 and order 3 −25, a total of +125 dollars. The average entry of the 0.35 lot is about 1.1034
4. If price reaches 1.1180 and you close everything: order 1 makes 180 pips × 0.20 × 10 = 360, order 2 120 × 0.10 × 10 = 120, order 3 60 × 0.05 × 10 = 30, a total of +510 dollars

Compare that with holding a single 0.20 lot to 1.1180, which makes +360 dollars. Opening 0.35 lot from the start would make +630, but it risks 175 dollars, or 1.75%, which is over plan. The pyramid gets bigger only on a trade that is already right. The price you pay is a worse average entry, and the upper levels are easily stopped out if price pulls back deeply.

The one rule for adding: move the Stop loss first, then size the new order from the worst-case loss of all orders together, not from the new order alone.

Averaging down breaks the plan immediately

Same example: the first order is 0.20 lot at 1.1000 with SL 1.0950. Price drops to 1.0970 and you add another 0.20 lot.

• New average entry = (1.1000 + 1.0970) ÷ 2 = 1.0985, which looks better
• If the original SL is hit, order 1 loses 100 and order 2 loses 20 × 0.20 × 10 = 40, a total of −140 dollars, or 1.4%, over the 1% plan

The first order used the whole risk budget when it opened, so anything added in the losing direction is pure extra risk. The only way to average in without going over plan is to split the budget from the start, for example plan two orders and size them so the combined loss at the SL is no more than 1%. That is the same position size with split entries, not rescuing a wrong trade.

Partial close: the average entry stays, the shape of results changes

A partial close does not change the entry price of what is left. Close 0.10 of a 0.20 lot trade at 1R and you still hold 0.10 lot at the original entry. What changes is:

• Remaining risk halves at once. If you move the stop on the rest to entry (break-even, BE), the worst outcome for the trade is +0.5R
• Win rate goes up, because every trade that reaches 1R has already booked a profit
• Profit on trades that run far halves, because only half the lot reaches the target

The real question is what you gain and lose overall. To answer it, every plan has to face the same price paths.

Expectancy of each exit plan

Expectancy is the average profit per trade in R: each outcome times its probability, added up. We use a simple model where price moves in steps of 0.1R, with a 50.5% chance of up and 49.5% of down, a small edge in our direction. The chance of reaching each level comes from the random-walk formula, computed in node:

• Reach +1R before the SL: 55.0% · reach +2R before the SL: 40.2% · reach +3R before the SL: 32.9%
• Once at +1R: reach +3R before returning to entry 40.2% · reach +3R before returning to the original SL 59.9%

Worked example for the plan close 50% at 1R, move the rest to BE, target 3R, per 100 trades:

1. 45.0 trades never reach 1R and hit the SL: 45.0 × (−1R) = −45.0R
2. 55.0 × 40.2% = 22.1 trades close half at 1R and the other half reaches 3R: 0.5 + 1.5 = +2R each, +44.2R in total
3. 55.0 − 22.1 = 32.9 trades close half at 1R and the other half is stopped at BE: +0.5R each, +16.4R in total
4. Total −45.0 + 44.2 + 16.4 = +15.6R per 100 trades, or +0.156R per trade, with a 55.0% win rate

Exit plan	Win rate	Expectancy per trade
Close all at 1R	55.0%	+0.100R
Close all at 2R	40.2%	+0.205R
Close all at 3R	32.9%	+0.317R
Close 50% at 1R, rest to 3R (SL stays)	32.9%	+0.208R
Close 50% at 1R, rest to BE then 3R	55.0%	+0.156R
Target 3R, SL to BE at 1R	22.1%	+0.213R

Row 4 wins only 32.9% because a trade that reaches 1R and then comes back to the original SL makes +0.5R and loses −0.5R, exactly break-even (22.1% of all trades). The table shows two things.

• A partial close is a weighted average. Row 4 is exactly half of the 1R plan plus half of the 3R plan: (0.100 + 0.317) ÷ 2 ≈ 0.208. Row 5 is half of the 1R plan plus half of row 6: (0.100 + 0.213) ÷ 2 ≈ 0.156. Two half lots are two independent orders on the same price path, so a partial close can never beat the best plan in its own mix
• With no edge at all (a 50/50 chance up or down), every plan has zero expectancy before costs. Closing half and moving to BE wins 50%, the full 3R target wins only 25%, but the expectancy is the same. The higher win rate comes from splitting the outcome, not from an edge. This ties into A high win rate doesn’t mean profit

This model has price drifting slightly one way, so far targets win. In a market where price tends to come back, the result can flip. So the numbers don’t tell you which plan is best. What holds in every market is the structure: a partial close is always a mix of full-exit plans. To know which plan suits your system, measure how often its trades reach 1R, 2R and 3R. See also Near or far Take profit.

From our backtests

We tested adding positions, profit guards and trailing stops on several systems on gold (XAUUSD) in the MT5 Strategy Tester. The test was always against simply raising the lot at the same drawdown, because any way of adding size makes profit look bigger. The question is whether it makes more per unit of risk.

Test	Without	With
Add an order each time price passes the next Fibonacci level (M1, Jan–Jul 2026, every-tick test, 0.02 lot per order)	+5,608, DD 10.6%	+7,687, DD 27.3%
Add to winners on a far-target system (real ticks 2026, 3,000 balance)	+2,532, equity DD 265	+3,644, equity DD 495
SL to break-even at 30% of the target distance (multi-strategy set, real ticks 2026, 6,000 balance)	+12,757	+11,621
Lock in part of the profit at 30% of the target distance (same set)	+12,757	+10,196
Four stepped trailing-stop ladders (on the near-target leg of the same system, real ticks 2026, 3,000 balance)	+2,532, equity DD 265	+2,096 to +2,319, equity DD 292–331

• Most adding is just a roundabout way to raise the lot. In the first row profit rose 37% but drawdown nearly tripled, while running the original setup at double lot was estimated at about +11,216 at roughly 21% drawdown, better on both counts. In the second row profit rose 44% but equity drawdown rose 87%
• There was one exception. A pullback system on gold M15 that added a single order only after price had moved half the frame in its favour, and moved the first order’s SL to entry first. In 2026 (real ticks) profit went from +2,843 to +5,868 with drawdown 24.75% → 26.02%, and 2022–2025 (simulated ticks) went from +1,300 to +2,261 in total with the worst drawdown 37.7% → 25.2%. But 2022 and 2023 still lost money either way. It helped only on a system whose trades tend to run far once they are right, and it still had to pass the same comparison
• The profit guard added trades, from 2,444 to 3,531, because trades cut at break-even were re-entered and some lost again. In 2025 and 2024 (simulated ticks) the differences were under 100 dollars either way. This matches Moving the Stop loss to break-even
• Trailing lost to a fixed target in order. The faster the stop moved in, the less it made, and drawdown went up, not down

All of these are backtest results. Past results do not predict future results, and they are not returns you should expect. To read profit per drawdown in a report, see The MT5 backtest report explained.

Using scaling sensibly

• Write the plan before entering: where you add, where you take partial profits and how big each piece is. Don’t decide while the trade is running
• Total risk across all orders ≤ the original plan, every time you add. Use the lot size calculator to size the new order from the budget left after moving the SL, not from a fresh 1%
• Keep upper levels smaller than lower ones, so the average entry doesn’t climb toward the SL too fast
• Count correlated pairs as the same trade. Buying EURUSD and then GBPUSD is close to adding in the same direction. Read Correlation and portfolio risk
• Compare with simply raising the lot before using it for real. If you want size to grow with the account, compounding by balance is usually easier to control. Read Compounding with position size
• Check the minimum lot. 0.01 lot can’t be split in half. To close half you need at least 0.02 lot

Common mistakes

• Adding before moving the SL. In between, total risk is over plan, and if price turns at once you lose more than intended. In the example above, adding 0.10 lot with the SL still at 1.0950 means a total loss of 210 dollars if it’s hit
• Treating a partial close as free profit. The win rate goes up, but expectancy is only the average of the plans in the mix. If the system has no edge, no way of closing makes money after costs
• Calling averaging down “scaling in”. If you add while in a loss, the average entry improves but risk goes over plan. That is a rescue trade, not a pyramid
• Sizing each add from a full 1%. Every level risking a new 1% means three levels risk close to 3% in one direction
• Judging from the trades you remember. The trade you closed half of and then watched run, or the one you held that came back to the SL, is just one trade. Count them all in your journal, or work out your actual RR with the RR calculator