2.1.3 Exchanges, Dark Pools and OTC

Basic Prerequisites: 02 - Order Types.

---

1. Intuition & Practical Objective

An order type is only half the story; the other half is where it is sent. The same limit order behaves very differently on a lit exchange (its price is displayed and joins a queue), in a dark pool (it is invisible and matches at the midpoint), and at an OTC dealer (there is no book at all - you trade against a quote). This page maps the venue landscape, the matching rules that decide who is first, the fee/rebate economics that make some orders pay you to post, and the regulation that stitches fragmented venues into one logical market.

Hasbrouck (Ch 2) gives the taxonomy that has survived three decades of electronification:

• Continuous limit-order market - a public book continuously matched by price-time priority. The workhorse of equities and futures.
• Dealer / quote-driven market - an intermediary quotes bid and ask and trades against its own inventory. Dominant in FX, corporate bonds, and swaps.
• Floor / open-outcry - bilateral negotiation by members (largely historical; CBOT/NYMEX survivors).
• Auction - a periodic single-price match (the fixings, opens, and closes) - the subject of 04 · Auctions & Continuous Trading.
• Crossing network / dark pool - off-quote matching, typically at the midpoint of the lit NBBO, with delayed or no trade reporting.
• OTC - bilateral, non-displayed, dealer-intermediated; no central matching engine.

Why fragmentation exists. Liquidity is a network externality (a deeper book attracts more traders), which favors consolidation - but retail vs institutional needs, venue innovation, and regulation that permits multiple venues pull the other way (Hasbrouck Ch 1). The result is a fragmented market where the best bid may sit on venue A and the best ask on venue B, forcing smart order routing (SOR) to consolidate them - and forcing this page to define the NBBO.

The one-sentence essence. “A venue is a matching rule plus a fee schedule plus a display policy: lit exchanges show quotes and charge maker-taker fees, dark pools hide them and match at the midpoint, OTC dealers quote bilaterally - and the NBBO is the synthetic best price across all of them.”

---

2. Mathematical Ground Truth & Derivations

Consolidation and the NBBO. With venues $v=1,\dots ,V$ each quoting $(b_v,a_v)$, the consolidated National Best Bid and Offer is

$$b^{\text{NBBO}}=\underset{v}{\max }{b}_v,a^{\text{NBBO}}=\underset{v}{\min }{a}_v,m^{\text{NBBO}}=\frac{1}{2}(b^{\text{NBBO}}+a^{\text{NBBO}}).$$

A market is locked if $b^{\text{NBBO}}\ge a^{\text{NBBO}}$ and crossed if strict - both are transient arbitrage states that SOR and latency arbitrage exist to exploit or prevent.

Smart order routing as an allocation problem. To buy $Q$ at minimum cost across venues with ask schedules $\{(p_{v,j}^a,q_{v,j})\}$, SOR solves

$$\underset{\{x_{v,j}\ge 0\}}{\min }\sum _{v,j}{x}_{v,j}{p}_{v,j}^a\text{s.t.}\sum _{v,j}{x}_{v,j}=Q,$$

a simple greedy fill of the globally cheapest levels (a concave/linear program with a trivial level-merging solution) - but one that ignores queue position, fees, and information leakage, which is why real SOR is a heuristic, not this optimum.

Fees and the all-in price. For $N$ shares at price $p$ on a maker-taker venue with maker rebate $r_m$ and taker fee $t_a$:

$$\text{cost}=\underset{\text{principal}}{\underbrace{Np}}+\underset{\text{access}}{\underbrace{Nf}},f=\{\begin{array}{ll}-r_m& \text{you were the maker}\\ +t_a& \text{you were the taker}\end{array}.$$

On an inverted venue the roles reverse (makers pay, takers earn). The effective spread you capture is therefore the quoted spread plus the fee differential:

$$S^{\text{eff}}=S^{\text{quoted}}+(f_{\text{taker}}-f_{\text{maker}}).$$

Dark-pool midpoint economics. A dark order that executes at $m^{\text{NBBO}}$ avoids crossing the spread: versus a taker paying $a^{\text{NBBO}}$, the saving is

$$a^{\text{NBBO}}-m^{\text{NBBO}}=\frac{1}{2}(a^{\text{NBBO}}-b^{\text{NBBO}})=\frac{1}{2}{S}^{\text{NBBO}}.$$

The saving is exactly half the NBBO spread - the entire allure of dark trading, offset by a lower fill probability and by the adverse selection of whoever chooses to trade dark.

---

3. Computational Implementation - fees, NBBO consolidation, dark midpoint

Price the same 1000-share trade across fee models and consolidate a fragmented NBBO. Standard library only; re-executed and reproduced.

Read the numbers. The **maker-taker swing is $5.00per1000shares\ast \ast -0.0050dollars/share-whichona1-cent-spreadnameis\ast halfthespread\ast ,sothechoiceofvenueandofpassive-vs-aggressivechangestheeconomicsofthetradeasmuchasthepricedoes.Onthefragmentedbook,boththebestbidandthebestasksitonvenueBinthisexample,butingeneraltheycansiton\ast different\ast venues:theNBBO($100.01/100.04$)isa\ast synthetic\ast pricethatnosingleexchangedisplays,andamidpointdarkmatchat$100.025$ captures 1.50 bps - one half of the 3-cent NBBO spread - without ever showing an order.

---

4. Failure Modes & First-Principles Breakdowns

1. Ignoring fees makes routing wrong. A venue with a 1-cent-better price but a 3-cent taker fee is worse than a 2-cent-worse price with a 2-cent rebate. SOR that optimizes quoted price alone systematically over-pays. The all-in spread $S^{\text{quoted}}+(f_{\text{taker}}-f_{\text{maker}})$ is the only correct objective.
2. Dark pools trade adverse selection, not just spread. The midpoint saving ($\frac{1}{2}{S}^{\text{NBBO}}$) is real, but dark fills are conditioned on a counterparty choosing to cross - precisely when informed flow wants the other side. The lower fill probability (see the 35% hit-rate in 05 · Failure Modes) plus the adverse selection can erase the half-spread saving.
3. Fragmentation mismanages the queue. Splitting an order across venues is not free: your passive order now sits behind other traders’ queues at each venue, and the consolidated fill probability is below the best single-venue queue. More venues is not more liquidity - it is more queue positions each of which may not fill.
4. Regulatory constraints bind routing. Order-protection rules (which force routing to the NBBO), best-execution obligations, and trade-reporting regimes change what an SOR is legally allowed to do - a routing decision is never purely a cost minimization.

---

5. References

• Hasbrouck, Joel - Empirical Market Microstructure (2007)
• Foucault, Pagano & Röell - Market Liquidity (2013)
• O’Hara & Ye - “Is market fragmentation harming market quality?” JFE 100(3), 2011. The empirical anchor on fragmentation and market quality.
• Degryse, de Jong & van Kervel - “The impact of dark trading and visible fragmentation on market quality,” Review of Finance 19(4), 2015. Dark vs lit fragmentation.
• Menkveld - “High-frequency trading and the new market makers,” JFM 16(4), 2013. *Why HFT venues concentrate across fragmented markets
• Johnson, Barry - Algorithmic Trading & DMA (2010). SOR and order-lifecycle mechanics in practice (with regulation: Reg NMS order protection, MiFID II / RTS 6 best execution and algorithmic-trading rules).

---

6. Connected Graph Bridges

• Back: 02 · Order Types · Index Hub
• Continue: 04 · Auctions & Continuous Trading · 06 · Advanced Extensions
• Rebates & market-maker economics: Market-Maker Economics & Rebates
• Fill dynamics across venues: Queue Position & Fill Probability
• Execution across the fragment: Execution Algorithms: VWAP, TWAP, POV