Scaling in and out of trades: average price, realized R and expectancy
Scaling changes your average price, your realized R and the shape of your results. Selling a partial can raise win rate while lowering expectancy, so compare both on the same trades.
Start in plain English
Scaling means entering or exiting a position in parts instead of all at once. Scaling out sells some shares at a first target and holds the rest; scaling in buys a starter position and adds to it. Small-cap scalpers use both, often with a hotkey that sells half. This lesson shows what scaling does to the numbers that decide whether a process makes money: your average price, your realized R and your expectancy after costs.
The core idea: scaling does not change what the stock does; it changes which part of each outcome you keep. Selling half at the first target turns some losers into small winners and some large winners into medium ones. Whether that helps depends on how often each of those outcomes happens in your own trades, and that can only be answered by counting.
Two conventions keep the math honest. Realized R for a scaled trade is the total net P&L of all its parts divided by the trade's planned 1R (see R multiple). For a planned scale-in, 1R is the risk of the full planned position at the initial stop.
Two paths from the same start
All prices, sizes and costs in this example are Illustrative. Tamsin Robotics is fictional.
The plan, written before 10:12 ET. Long 1,000 shares at $4.20, stop $4.12. Planned risk $0.08 per share, so 1R = $80. Fees $0.005 per share round trip, $5.00 for the full position. Two exit plans on the same entry:
- All-out: sell 1,000 at the target $4.36 (2R), or at the stop.
- Scale-out: sell 500 at $4.28 (1R), move the stop on the other 500 to break-even ($4.20), sell those at $4.36.
Path AThe partial locks in a gain before a reversal
Price trades $4.29 at 10:15, so the 500-share partial fills at $4.28. At 10:18 the stock reverses and trades back through $4.20, then $4.12.
Scale-out: +$40 on the partial, $0 on the rest at break-even, minus $5 in fees: +$35, +0.44R. All-out on the same path: stopped at $4.12, −$80 − $5 = −$85, −1.06R.
Path BThe partial caps a large winner
Same start. The partial fills at $4.28, the stock keeps going, and the rest sells at $4.36 at 10:21.
Scale-out: +$40 + $80 − $5 = +$115, +1.44R. All-out: 1,000 × $0.16 − $5 = +$155, +1.94R. The partial gave up half a risk unit on the trade that went furthest.
Path CNothing much happens
Price chops between $4.15 and $4.26 and does not reach $4.28. The time stop exits all 1,000 shares at $4.20 at 10:22.
Both plans: $0 gross, −$5 in fees: −0.06R (−$5.00).
Review note for all three: "Plan followed; partial at 1R as written; outcome A, B or C." Neither exit plan was wrong on any single trade. Which plan does better over many trades depends on how often Path A and Path B happen, which the next sections turn into numbers.
Odds box
In this split, scaling out raised the win rate by ten points and lowered expectancy by about 0.08R per trade (computed from the assumed split, Illustrative). The partial helps on the 10 Path A trades (+1.5R each compared with all-out) and costs on the 45 Path B trades (−0.5R each). With these payoffs, scaling out comes out ahead only when Path A trades are more than about one third as common as Path B trades. That ratio is a property of your setups and your exits, not of scaling in general.
A higher win rate feels better, which is part of why scaling out is popular. Shefrin and Statman (1985) called the wider tendency the disposition effect: selling winners early and holding losers. A partial at 1R is a planned version of the same instinct, so judge it by expectancy, not by how often it lets you book a win.
Average price and realized R
Average price is total cost divided by total shares. Buy 700 at $2.45 and add 300 at $2.55: (700 × $2.45 + 300 × $2.55) ÷ 1,000 = $2.48 (Illustrative). The simple average of the two prices, $2.50, is wrong whenever the parts differ in size.
Realized R adds the net P&L of every part and divides once. In Path A that was ($40 + $0 − $5) ÷ $80. Two common errors:
- Counting the partial as its own "+1R win". It was +1R on half the size, which is +0.5R of the trade.
- Re-basing 1R on the break-even stop after the partial. The trade's 1R stays $80.
Scaling in: planned adds vs. averaging down
A planned scale-in writes the add into the plan: where it happens, how many shares, and where the stop for the whole position sits. Example (Illustrative): buy 500 at $4.20, add 500 at $4.26 on a break of $4.25, stop $4.12 for both. If both fill, the average price is $4.23 and the risk at the stop is ($4.23 − $4.12) × 1,000 = $110. That $110 is the trade's 1R, even on days when only the first 500 fill. A starter that is stopped before the add loses $40 plus $2.50 in fees, −$42.50 ÷ $110 = about −0.39R on this plan.
Averaging down adds as price moves against you, lowering the average price. It is a different decision: unless the full size and a stop for the full size were written in advance, each add raises the risk above the plan. Odean (1998) found that individual investors were more reluctant to realize losses than gains; an unplanned add to a loser is one way that reluctance shows up in a scalper's log, and it often appears as a loss well beyond −1R.
Observe vs. infer
At 10:16 ET in Path A, the partial has just filled at $4.28. Here is what the trader could see.
| What the screen shows (observed) | What a trader might infer | Why it is not established |
|---|---|---|
| 500 shares sold at $4.28; 500 still held with a stop at $4.20 | "Now it's a free trade" | The rest can still lose fees, slippage on the stop, or much more on a halt |
| Five higher closes in a row | "Momentum is strong; the rest is on its way to $4.36" | The same bars start Path A, B and C |
| The 10:15 bar had the highest volume of the six | "Buyers are stepping in" | Volume at a new short-term high fits buyers lifting offers and sellers meeting them |
What shifts the odds
No FennEdge study measures partial-exit outcomes. The items below are hypotheses to record, not findings. Log them with each trade so your own sample can test them:
- Where the partial sits. A partial at 0.5R changes more trades than one at 1R; one at 1.5R changes fewer.
- What happens to the stop after the partial. Break-even, unchanged or trailed; each gives a different Path A result.
- Spread and liquidity at the partial price. A partial that needs to sell into a thin bid can slip as much as a stop.
- Time of day. If your late-morning trades reach 2R less often than your early ones, the best partial rule may differ by time bucket.
- Halt frequency in the names you trade. Halts affect the share still held after the partial.
Cost of being wrong
Using the Tamsin Robotics scale-out plan (1,000 shares, 1R = $80; Illustrative):
| Outcome | Exits | R after costs | Dollars after costs |
|---|---|---|---|
| Stopped before the partial | 1,000 at $4.12 | −1.06R | −$85.00 |
| Stopped before the partial, 2 cents slippage | 1,000 at $4.10 | −1.31R | −$105.00 |
| Partial, then halt-down; rest reopens at $3.95 | 500 at $4.28, 500 at $3.95 | −1.13R | −$90.00 |
| All-out, halt-down to $3.95 | 1,000 at $3.95 | −3.19R | −$255.00 |
| Partial not filled (bid pulled at $4.27), then stop | 1,000 at $4.12 | −1.06R | −$85.00 |
Splitting an exit also adds orders. On a per-share fee plan the total fee is the same; on a per-order plan or with minimum charges, each partial costs extra (ECN fees). Each partial is a separate order with its own fill uncertainty; Marketable limit orders covers partial fills.
How many trades before this means anything
The question is not "does scaling work?" but "on my trades, does scaling at this level beat all-out after costs?" Compare both exits on the same trades: record where each trade went (did it touch 1R, did it reach 2R) so you can compute both results for every row. A paired comparison like this is less noisy than comparing two separate samples, but the gap you are looking for is small. In the odds box, the two plans differ by about 0.08R per trade; separating a difference that size from noise typically takes well over 100 trades (Illustrative). How many trades before your numbers mean anything explains why.
What this lesson does not claim
- It does not claim scaling out is better or worse than all-out exits. The split in the odds box is Illustrative.
- A higher win rate is not a higher expectancy; the example shows they can move in opposite directions.
- Planned scale-ins are a sizing choice, not a way to make a losing trade right.
- Tamsin Robotics, its bars and all costs are fictional and Illustrative.
Go deeper
- FennEdge first.
- DAS course: Order objects builds and inspects orders before sending, which matters when one keystroke places a partial; Order filters, identity and cancellation scope shows how to cancel only the remaining exit order, not every order; Orders, fills and protection are separate states covers why a filled partial does not move your stop by itself.
- The planned DAS Hotkey Script Builder is to generate risk-sized entry keys with an optional stop key; as of 2026-10-05 it is not live and has no partial-exit templates, so build those with the DAS course lessons above and test every script in a simulator first.
- If you trade live, the Trading Journal syncs real fills from supported brokers; on Plus and above, a scaled trade appears on its 1-minute chart with each partial fill marked, and per-fill fee detail is on Pro and Max (as of 2026-10-05). It holds broker-synced trades only, not practice entries.
- Exits compared. Exits: fixed targets, trailing exits and time stops compares exit rules on the same trades using MFE (MAE and MFE).
- Partial fills when scaling in. If the add fills only partly, the position is smaller than planned and 1R stays at the full plan; note the shortfall.
- Routing. Where a partial order goes changes fees and fill odds; see Order routing.
- Sizing. Position sizing for scalps covers the liquidity cap that limits how much you can sell at the partial price.
- Related: Risk/reward and breakeven win rate on planned vs. realized R:R.
Exercises
Six exercises follow this lesson: three concept checks (what a partial changes; planned adds vs. averaging down; why win rate and expectancy can move apart) and three calculations (average price after an add; realized R for a scaled trade after fees; expectancy of a scale-out plan from an assumed split). Exercises, answers and progress are available after signing in.
Primary sources
- Shefrin, H. and Statman, M. (1985). The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence. The Journal of Finance 40(3).
- Odean, T. (1998). Are Investors Reluctant to Realize Their Losses? The Journal of Finance 53(5).
- FINRA. Types of orders.
Sources checked 2026-10-05. Tamsin Robotics and all prices, volumes, outcome splits and costs in this lesson are fictional and Illustrative. This lesson is education, not a recommendation to buy or sell any security. Past patterns do not indicate future results.