Scaling in and out

Definition

Scaling splits entries and exits into tranches instead of single transactions: scaling IN builds a position across prices or times; scaling OUT harvests it in parts. It trades all-or-nothing timing risk for averaged prices and partial outcomes — and it carries a hard risk-accounting rule: the FULL intended size must pass the risk budget at the first tranche, because adverse paths tend to fill the whole ladder.

How it works / structure

• Scale-in forms (engine-executable): time-based tranches (fixed schedule — dollar-cost averaging is the passive version); price-ladder tranches (add at predefined levels — natural fit for strategy-mean-reversion entries); confirmation adds (add only as the thesis proves — pyramid, the trend-system convention: adds on strength, never weakness).
• Scale-out forms: fixed fractions at R-multiples or targets (mgmt-profit-target partials); trailing remainder (bank part, let the rest run behind a trail — mgmt-stop-loss trailing variant).
• The accounting rule: risk is computed on the LADDER, not the tranche — a 3-tranche averaging-down plan is one position of full size with a worse average entry on exactly the paths where all three fill; the engine sizes the whole ladder against risk-fixed-fractional at entry.
• Parameters: tranche count/sizes, spacing (time, price, or confirmation), direction (with-trend adds vs against-trend adds — declared, because their risk shapes are opposite), and the ladder’s aggregate stop.

When it applies

Uncertain entry timing on sound theses (tranches buy optionality on the entry price); mean-reversion ladders with pre-declared depth; trend pyramids where adds are earned by movement; exits from illiquid size (ms-liquidity — scaling out is execution hygiene at size).

Risk profile & failure modes

• Averaging down as denial: the unplanned add — averaging down beyond the declared ladder to “improve the basis” — is loss aversion buying more of a failing thesis (bias-loss-aversion); the declared-depth rule exists to make this structurally impossible.
• Adverse-selection fills: price ladders fill fully on the paths that keep going against; the ladder’s worst case IS its expected shape on losing trades — hence full-size budgeting.
• Pyramid top-heaviness: adds on strength raise the average basis; a reversal hits the largest position at the worst basis; trend pyramids need trail rules matched to the add schedule.
• Partial-exit drag: scaling out of winners early is the disposition effect in installments — same replay test as mgmt-profit-target.

Evidence & limits

Dollar-cost averaging mechanics are SEC-documented; the literature comparing DCA to lump-sum investing generally finds lump-sum ahead on average (a consequence of positive expected drift — time in market), with DCA reducing regret variance; period-dependent, and the platform treats schedule-vs-lump as a replay question. Ladder and pyramid conventions are practitioner folklore until a specific parameterization survives replay.

Falsifiable-thesis examples

Illustrations only, not signals:

• “A 3-tranche ladder entry (spaced 1 ATR) on strategy S beats single-shot entry on realized average price this quarter in replay” — falsified by the paired comparison.
• “This trend pyramid (add at +1R, +2R, trail at 2 ATR) will outperform the flat-size version on this quarter’s replay” — falsified by the paired result.

Cross-references

• The budget it must pass: risk-fixed-fractional, risk-max-drawdown-budget
• Exit partials: mgmt-profit-target, mgmt-stop-loss (trailing)
• Natural habitats: strategy-mean-reversion (ladders), strategy-futures-trend-following (pyramids), port-rebalancing (the portfolio-level cousin)
• The bias it contains: bias-loss-aversion