Scheduling algorithms decide how much to trade and when; the router decides where. Between them sits a tactical layer that quietly determines a surprising share of realized performance: for each child order, at what price should it be placed, at what size, and how much of it should the market see? This part covers that placement logic.

Pricing: The Aggression Ladder

Every child order sits somewhere on a ladder of aggression, and each rung buys fill probability with cost:

• Behind the touch (below best bid for a buyer): cheapest if filled - you earn more than the half-spread - but fills arrive mostly when the market trades through you, which is adverse-selection-rich territory.

• At the touch (joining the best bid): the standard passive posture; you earn the half-spread but wait in a queue.

• Improving the touch (a new best bid): jumps every existing queue at the cost of one tick - often the best trade in wide-spread instruments, where one tick is much less than the spread you save.

• Midpoint: split the spread, typically via hidden or dark orders.

• Crossing (marketable): certainty, at the price of the half-spread plus taker fees plus possible walk up the book.

Good pricing logic is spread-aware and state-aware: in a one-tick-wide book, joining the queue and crossing are the only real choices, and queue position is everything; in a wide book, price improvement and midpoint work harvest most of the value. The scheduler’s urgency (its band position - see Part 4) sets which rungs are permitted at any moment.

Queue Position: The Invisible Asset

On price-time venues, time priority makes a resting order’s queue position a genuine asset: an order near the front collects the “good” fills (from routine two-way flow), while an order at the back gets filled disproportionately when the level is about to break - the classic adverse-selection fill. Two practical consequences:

• Don’t reprice casually. Cancelling and replacing resets you to the back of the queue. Placement engines weigh the value of a better price against the queue position being surrendered.

• Age orders deliberately. Posting early at levels the market is likely to reach builds priority before it is needed - one of the quiet edges of well-built passive engines.

Sizing and Display

How Big Should a Child Order Be?

Clip size balances three pressures: too small leaks a steady metronome of prints (and burns message capacity); too large distorts the book and advertises intent. Reference points are the venue’s typical trade size and displayed depth at the touch - a child order that looks like everyone else’s is the goal. Sizes should be randomized around that reference; uniform clips are the single most recognizable algorithmic fingerprint on a tape.

Icebergs and Reserve

Display logic decides how much of the working quantity is visible:

• Iceberg / reserve orders show a small tip and hold the remainder hidden at the venue, automatically replenishing the tip as it fills. The replenishment pattern itself can be detected, so better engines randomize tip sizes and refresh timing.

• Fully hidden orders rest invisible at or inside the touch on venues that allow them, sacrificing queue priority (displayed size fills first - Part 2) in exchange for zero footprint.

• Dark placement hands the quantity to the dark aggregator (Part 9) to rest at midpoints with minimum-fill protection.

A useful discipline for thin markets: cap displayed size as a fraction of the visible book, not as an absolute number. Showing 5,000 units in a book that usually displays 800 is a broadcast, not an order.

Repricing: Chasing vs. Sitting

When the market drifts away from a resting order, the engine faces the chase decision. Repricing after every tick means perpetually paying fresh spread at the worst moments and abandoning accumulated queue position; never repricing means fill rates collapse in a trend. Production engines use hysteresis - reprice only after the market moves a threshold distance or dwell time - plus the scheduler’s band state: sitting is fine while comfortably inside the bands, chasing becomes mandatory as the lower band approaches.

Reading the Book: Micro-Signals

Placement engines increasingly condition on short-horizon microstructure signals:

• Book imbalance: when displayed bid size dwarfs ask size, the next move is more likely up - a buyer should cross now rather than wait to be run over; symmetric logic says be patient when imbalance favors you.

• Micro-price: the depth-weighted fair value between bid and ask tells you whether the midpoint is currently a good or bad deal - worth checking before every midpoint fill.

• Trade-flow momentum: a run of aggressive prints in one direction warns the passive side that its level is about to trade through.

These signals are the same raw material the opportunistic strike logic of Part 7 uses at a larger scale - placement is where they operate order by order.

The Anti-Gaming Checklist

Pulling the tactics into one list - the properties a placement layer needs before it can be trusted with a sensitive order:

1. Randomized clip sizes and action timing (no round numbers, no clock alignment).

2. Queue-aware repricing with hysteresis.

3. Display caps scaled to the live book.

4. Minimum-fill protection on hidden and dark placements.

5. Imbalance-aware crossing to avoid the worst adverse-selection fills.

With scheduling, routing, and placement covered, the execution stack is complete. What remains is accountability: proving what all this machinery earned. That is Part 12: Benchmarks & Transaction Cost Analysis.