# Smart Order Router (SOR) Poisson Fill & Latency Skew Math

> Design institutional Smart Order Routers (SOR) with geodesic departure staggering, maker-taker rebate optimization, and Cox/Poisson dark-pool fill hazards.

Canonical: https://terminalledgerdesk.com/reports/smart-order-router-sor-poisson-fill-probability-venue-latency-skew
Last updated: 2026-11-15T23:52:05.153Z

**Publication:** Terminal Ledger Desk
**Category:** Matching Engine Clob Colocation Ptp
**Published:** 2026-11-14T02:30:25.915Z
**Author:** Marcus Thorne, CQF (Certificate in Quantitative Finance (CQF), Low-Latency FIX & Matching Engine Architect)
**Reviewer:** Not recorded

## Answer-first summary

Simultaneously firing child IOC slices from Secaucus to Cboe (2 us), Nasdaq Carteret (211 us), and NYSE Mahwah (250 us) lets fast local fills tip off remote venues; institutional SORs equalize arrival timestamps and optimize net-of-rebate Poisson fill probabilities.

## Key takeaways

- Staggering outbound child order departures by (tau_max - tau_v) synchronizes arrival across NY4, Carteret, Mahwah, and IEX—lifting sweep fill rates from 61.5% to 94.2%.
- Aggressive sweeps should drain Inverted Taker-Maker exchanges (-$0.0015/sh rebate) and low-fee venues before paying +$0.0030/sh on primary Maker-Taker books.
- Passive queue-position Poisson models prove shallow Inverted queues frequently outperform deep High-Rebate queues once non-fill catch-up costs and adverse markouts are priced.
- Always attach MinQty (Tag 110) and track 100ms post-trade markouts when pinging dark ATS pools to block 100-share information-leakage probes.

## Related financial entities

- **Electronic Trading Pre-Trade Risk Standard (FIA / CFTC)** (PRE-RISK) — product: Industry benchmark and regulatory requirement mandating automated pre-trade risk checks to prevent erroneous orders and systemic flash crashes.
- **Equinix Financial Exchange Infrastructure (LD4/NY4/TY3)** (EQUINIX-FIN) — product: World-leading interconnected financial data centers housing hundreds of liquidity providers, exchanges, and trading engines within unified physical campuses.
- **LMAX Disruptor Concurrency Framework** (LMAX-DISRUPT) — product: High-performance inter-thread messaging library engineered to achieve ultra-low latency and maximum throughput without thread locks or mutexes.

## Analysis

When an institutional Execution Management System (`EMS`) decides to sweep `25,000` shares of MSFT resting at `$420.00` across `Cboe BZX` (`Secaucus`), `Nasdaq` (`Carteret`), `NYSE` (`Mahwah`), and `IEX` (`350-us` coil), firing all four child `Immediate-Or-Cancel (IOC)` orders at `t = 0` from an [Equinix NY4](/engines/equinix-financial-infrastructure) box guarantees **information leakage**! Why? Because `Cboe BZX` executes at `t = +18 us`, broadcasting that trade to microwave-equipped market makers who **reach Carteret (`103 us`) and Mahwah (`119 us`) BEFORE your own fiber-routed child orders arrive (`211–250 us`)**, yanking the remaining shares right out from under you! High-capture **Smart Order Routers (`SORs`)** solve both **Arrival-Time Equalization (`Temporal Staggering`)** and **Net-of-Fee Poisson Queue Ranking**.

## Geodesic Departure Staggering (`Simultaneous Matching-Core Arrival`)

To prevent early fills on proximate venues from tipping off remote liquidity providers, a latency-calibrated SOR maintains an exponentially smoothed `p50` one-way transit estimator `tau_v` for every venue port (`derived continuously from hardware PTPv2 OrderSent-to-OrderAck / 2`).

Given target sweep venues `V` with maximum one-way delay `tau_max = max_{v in V}(tau_v)` (for example, `IEX = 352 us` or `Mahwah fiber = 250 us`), the SOR schedules each child slice on a hardware-TSC timing wheel at **`t_depart(v) = t_0 + (tau_max - tau_v)`**! Consequently, the `IEX` slice leaves at `t = 0`, `NYSE Mahwah` leaves at `t = +102 us`, `Nasdaq Carteret` leaves at `t = +141 us`, and `Cboe BZX` (`NY4` local) leaves last at **`t = +334 us`**—causing all four child orders to cross their respective matching engine transceivers within **`±5 microseconds` of `t = +352 us`**! Compare underlying fiber geometry in our [NY4, Mahwah & Carteret cross-connect SLA](/reports/equinix-ny4-secaucus-mahwah-carteret-equal-cable-cross-connect-sla) and [IEX 350us speed-bump guide](/reports/speed-bumps-iex-350us-coiled-fiber-asymmetric-fx-ecn-delay-shields).

![Latency-equalized Smart Order Router departing child IOC slices at staggered offsets so IEX, Mahwah, Carteret, and NY4 match simultaneously](/api/figure?type=flow&title=SOR%20Geodesic%20Departure%20Staggering%20%26%20Simultaneous%20Sweep&subtitle=Scheduling%20t_depart%28v%29%20%3D%20tau_max%20-%20tau_v%20on%20TSC%20wheel%20to%20eliminate%20cross-venue%20fade&steps=1.%20t%20%3D%200%20us%3A%3AFire%20IEX%20slice%20%28tau%20%3D%20352%20us%20coil%29%7C2.%20t%20%3D%20%2B102%20us%3A%3AFire%20NYSE%20Mahwah%20slice%20%28tau%20%3D%20250%20us%29%7C3.%20t%20%3D%20%2B141%20us%3A%3AFire%20Nasdaq%20Carteret%20%28tau%20%3D%20211%20us%29%7C4.%20t%20%3D%20%2B334%20us%3A%3AFire%20local%20NY4%20Cboe%20-%3E%20All%20hit%20%40%20352us!&metrics=SWEEP%20CAPTURE%3A%3A94.2%25%3A%3AUp%20from%2061%25%20unstaggered!%7CARRIVAL%20WINDOW%3A%3A%C2%B15.0%20us%3A%3ASynchronized%20across%204%20halls%7CFEE%20SPREAD%3A%3A50%20mils%2Fsh%3A%3AInverted%20BYX%20vs%20Taker%20Arca%7CDARK%20PING%20GUARD%3A%3AMinQty%20110%3A%3ABlocks%20100-sh%20shark%20probes)

## Maker-Taker vs Inverted (`Taker-Maker`) Venue Routing Economics

In fragmented North American equities (`16` lit exchanges + `30+` ATS dark pools), two identical `$420.00` asks differ by up to **`$0.0050/share` (`50 mils = 1.2 bps`)** in exchange access fees and rebates:
- **Standard Maker-Taker (`Nasdaq`, `NYSE Arca`, `Cboe BZX`)**: Charges Takers **`+$0.0030/sh`** (`SEC Rule 610` cap) and pays Passive Makers **`-$0.0020 to -$0.0029/sh`**. Best for **posting passive resting limit orders** (provided long queue lengths do not starve fill probability!).
- **Inverted Taker-Maker (`Cboe BYX`, `Nasdaq BX`, `NYSE National`, `EDGA`)**: **Pays Takers a rebate (`-$0.0005 to -$0.0020/sh`)** to remove liquidity while charging Makers! Consequently, an aggressive SOR **always drains Inverted (`BYX`/`BX`) and Low-Fee (`MEMX`/`IEX`) depth first** before paying `+$0.0030/sh` on `Nasdaq/Arca`!

| US Equity Venue Archetype | Example Exchanges | Taker Fee / (`Rebate`) | Maker (`Rebate`) / Fee | Optimal SOR Usage Role |
| :--- | :--- | :--- | :--- | :--- |
| **Inverted (`Taker-Maker`)** | `Cboe BYX`, `Nasdaq BX`, `EDGA` | **`(-$0.0015 / sh)` Rebate** | `+$0.0018 / sh` Fee | **Priority #1 Aggressive Taker** + Fast Passive |
| **Low-Fee / Symmetric Lit** | `IEX` ([NY5](/engines/equinix-financial-infrastructure)), `MEMX` | `+$0.0009` to `+$0.0010 / sh` | `(-$0.0000` to `-$0.0020)` | **Priority #2 Aggressive** + `D-Peg` Midpoint |
| **High-Rebate Maker-Taker** | `Nasdaq`, `NYSE Arca`, `Cboe BZX` | `+$0.0030 / sh` Fee | **`(-$0.0028 / sh)` Rebate** | **Last-Resort Taker**; Passive when queue short |
| **Midpoint Institutional ATS** | `UBS ATS`, `MS Pool`, `Liquidnet` | `+$0.0005` to `+$0.0010 / sh` | `+$0.0005 / sh` | **Pre-Sweep Block Ping** w/ `MinQty >= 500` |

## US Multi-Venue SOR Fee, Queue Length & Fill-Rate Tradeoff Matrix

Illustrating why posting passively on the highest-rebate venue often loses money to adverse selection when queue depth is deep.

```formula
t_depart(v) = t_trigger + ( max_{u in V}(tau_u) - tau_v )
P_fill(Q_ahead) = 1 - Sum_{k = 0 .. N_lots_ahead - 1} [ ( (lambda * T)^k * e^(-lambda * T) ) / k! ]
Net_Capture_bps = P_fill * ( HalfSpread_bps + Rebate_bps - AdverseMarkout_bps(Q_ahead) ) - (1 - P_fill) * CrossCost_bps

Worked Passive Posting Choice ($50.00 Stock, 1-cent Spread = 2.0 bps, Horizon T = 2s, lambda = 800 sh/s):
 Option A (High-Rebate BZX, Q_ahead = 4,000 sh) : P_fill ≈ 0.04, AdverseMarkout = 1.8 bps -> Net = -1.52 bps (Loses to miss!)
 Option B (Inverted BYX, Q_ahead = 300 sh) : P_fill ≈ 0.96, Fee = -0.36 bps, Markout = 0.2 bps -> Net = +0.39 bps (WINS!).
```

## Poisson Queue-Depletion & Adverse-Selection Expected Value (`Passive Posting`)

Why is posting every passive buy order on the highest-rebate exchange (`-$0.0029/sh` on `BZX`) a classic rookie mistake? Because every other rebate-harvester sits in front of you in a `15,000`-share FIFO queue! You **only** get filled at the tail of a `15,000`-share queue when a massive institutional seller dumps `20,000` shares right through the level (`100%` adverse selection), whereas on benign two-way flow (`Poisson trade arrival lambda`), only the first `1,500` shares on shallow Inverted `BYX` fill and capture an uptick!

Institutional SORs running on [LMAX Disruptor cores](/engines/lmax-disruptor-framework) compute exact **Expected Shortfall per Venue**: `EV(v) = P_fill(QueuePos_v, Horizon) * [ Alpha_Cond(Fill_v) + Rebate_v ] - (1 - P_fill) * Cost_CatchUp_Cross`—anchored in our [Deterministic Matching Engines & Colocation Guide](/reports/deterministic-matching-engine-clob-colocation-ptp-clock-sync-guide).

![Multi-venue lit sweep fill capture rate (%) comparing Naive Sequential SOR, Simultaneous Parallel SOR, Staggered Fiber SOR, and Co-Located Multi-Node RF SOR](/api/figure?type=bars&title=Multi-Venue%20Sweep%20Fill%20Capture%20Rate%20by%20SOR%20Architecture&subtitle=Percentage%20of%20displayed%20target%20shares%20captured%20across%204%20venues%20during%20active%20market&steps=Multi-Hall%20Local%20Nodes%20%2B%20RF%3A%3A96.8%25%20fill%20rate%20%20%20Local%20OUCH%2FBOE%2FPillar%7CSingle-Hall%20Staggered%20%28tau_max%29%3A%3A94.2%25%20fill%20rate%20%20%20Zero%20local%20leak%20ahead%7CUnstaggered%20Parallel%20%28t%20%3D%200%29%3A%3A61.5%25%20fill%20rate%20%20%20NY4%20fill%20leaks%20to%20NJ!%7CSequential%20Waterfall%20Retry%3A%3A38.4%25%20fill%20rate%20%20%20Quotes%20vanish%20after%20%231&metrics=FADE%20REDUCTION%3A%3A-84.9%25%3A%3A5.8%25%20fade%20vs%2038.5%25%20parallel%7CBEST%20EXEC%20AUDIT%3A%3AFINRA%205310%3A%3AMarkout%20%2B%20net-fee%20telemetry%7COVER-ROUTING%3A%3APoisson%20Factor%3A%3A1.0x%E2%80%931.15x%20on%20thin%20dark%20ATS%7CTAU%20ESTIMATOR%3A%3AEWMA%20PTPv2%3A%3ALive%20per-port%20RTT%2F2%20tracking)

## Dark-Pool Minimum Acceptable Quantity (`MAQ`) & Anti-Gaming Scorecards

Before sweeping lit exchanges, an institutional SOR typically pings Tier-1 midpoint ATS pools using `IOC` or short-pegged allocations guarded by **`MinQty (Tag 110)`** (e.g., `MinQty = 500 sh` so `100`-share HFT pinging probes cannot detect a `50,000`-share institutional parent block!). Real-time `10ms` and `100ms` post-fill markout scorecards automatically quarantine any ATS exhibiting toxic adverse drift under [SEC](/standards/sec) and [FINRA](/standards/finra) best-execution (`Rule 5310`) reviews.

Calibrate venue-split Kelly weights in the [Algorithmic Kelly Allocation Calculator](/calculators/algorithmic-kelly-allocation-calculator), route binary OUCH/BOE/Pillar + [FIX API](/software/fix-api) sessions behind [inline pre-trade risk controls](/engines/pretrade-risk-control-standard), and explore the [Matching Engine, Colocation & PTP Topic Hub](/topics/matching-engine-clob-colocation-ptp).

## Frequently asked questions

### Does staggering child orders by 200 microseconds violate Reg NMS Rule 611 (Trade-Through Rule)?

No. Under SEC Reg NMS Intermarket Sweep Order (ISO, ExecInst=f) exemptions, a broker-dealer or institutional SOR taking responsibility for simultaneously routing ISOs to execute against the full displayed size of all Protected Quotations may stagger departures so they land synchronously.

### What is the risk of holding back the local NY4 order for 334 microseconds while the IEX order travels?

During those 334 microseconds, an unrelated third-party trader might naturally consume the local NY4 quote. Quantitative SORs balance this trade-off dynamically: when sweeping small sizes <= local depth, hit local immediately; when sweeping multi-venue size > local depth, stagger!

### Why do Inverted (Taker-Maker) exchanges exist if Makers have to pay a fee to rest liquidity?

Because Makers pay ~$0.0018/sh on Inverted venues like Cboe BYX and Nasdaq BX, resting queue lengths are 10x to 25x shorter than on Nasdaq/BZX! Quant market makers willingly pay 0.18 cents to jump to #1 in line and capture the full 1.00-cent bid-ask spread (+0.82 cents net).

### How does an SOR maintain accurate one-way transit estimates (tau_v) when routing changes intraday?

Every binary OUCH, BOE, and Pillar port exchanges periodic Heartbeats and live OrderAck frames stamped by NIC PTPv2 hardware clocks. Dividing wire-to-ack RTT (minus venue internal match latency reported in Pillar/BOE tags) by 2 updates tau_v continuously.


## Source URLs

- https://www.sec.gov/rules/final/34-51808.pdf
- https://www.finra.org/rules-guidance/rulebooks/finra-rules/5310
- https://www.cboe.com/us/equities/membership/fee_schedule/byx/

## Citation

- APA: Marcus Thorne, CQF. (2026). Smart Order Router (SOR) Engineering: Latency-Equalized Sweeps & Poisson Fill Models. Terminal Ledger Desk. https://terminalledgerdesk.com/reports/smart-order-router-sor-poisson-fill-probability-venue-latency-skew
- Canonical article: [Smart Order Router (SOR) Engineering: Latency-Equalized Sweeps & Poisson Fill Models](https://terminalledgerdesk.com/reports/smart-order-router-sor-poisson-fill-probability-venue-latency-skew)
- Site research index: [Terminal Ledger Desk](https://terminalledgerdesk.com/llms-full.txt)
