Market Structure

What does market structure actually decide?

Market structure is the machinery that stands between a click and a fill. It decides which prices an order is allowed to trade at, which venue receives it, who is paid for supplying the quote it consumes, and how the resulting transaction becomes public information.

The United States does not run one stock market. It runs dozens of trading centers: registered exchanges, alternative trading systems, and off-exchange dealers, all trading the same securities at once. Nothing physically forces them to agree on a price. What links them is a rule set, a data pipeline, and a broker duty: Regulation NMS, the consolidated quote assembled by a securities information processor, and best execution under FINRA Rule 5310.

Every practical execution question resolves into those three pieces. An order fills at three prices because displayed size at each price is finite. It fills better than the screen showed because an off-exchange dealer competed for the flow. It sits unfilled while the price trades because a queue stood in front of it on that venue.

How does the quote that benchmarks everything get built?

The bid and ask identify the best displayed buying and selling prices. The bid is the highest displayed buying price, the ask is the lowest displayed selling price, and their average is the midpoint. The spread measures the distance between them. Depth reveals additional price levels, while hidden orders, midpoint matches, and price improvement add liquidity beyond the displayed top of book.

Because each venue keeps its own book, each venue has its own best bid and offer. The national best bid and offer consolidates them: the national best bid is the highest eligible displayed bid across venues, and the national best offer is the lowest eligible displayed offer. Those two prices routinely come from different exchanges, which is why an individual venue quote can be inferior to the national market.

That consolidation is a regulated process, not a vendor convenience. Plan processors collect required quote, trade, and status messages from participating exchanges and FINRA, validate and sequence them, then disseminate the national result. Consolidated Tape Association and Consolidated Quotation infrastructure covers Tape A, listed on NYSE, and Tape B, listed on other non-Nasdaq exchanges. Unlisted Trading Privileges infrastructure covers Tape C, the Nasdaq-listed names.

The NBBO is top of book only. Everything behind it lives in depth data. Level 2 market data extends the view to multiple participants, venues, or price levels, and market depth is the displayed quantity available at each of those prices. Traders summarize the shape of that ladder with order book imbalance, calculated as (Qb minus Qa) divided by (Qb plus Qa), which lands between negative one and positive one with balance at zero. Some depth screens report the bid-share form instead, Qb divided by (Qb plus Qa), which lands between zero and one with balance at 0.50, so check which scale a screen uses before you read it. Hidden orders, iceberg reserves and off-exchange interest sit outside all of these views, so displayed liquidity is the floor on what an order can find.

What does Regulation NMS actually require?

The Commission voted Regulation NMS through on April 6, 2005, and it took effect on August 29, 2005. Four pieces matter. Rule 611 is the Order Protection Rule. Rule 610 governs access to quotations and caps the fee a venue may charge to reach a protected quote. Rule 612 sets minimum pricing increments. The market data amendments define how consolidated quotes and last sales reach the public.

Rule 611 requires trading centers to maintain written policies and procedures reasonably designed to prevent trade-throughs, meaning buys printed above a protected offer or sells printed below a protected bid. The word protected is narrow. Only the automated, immediately and automatically accessible best bid or best offer of each exchange or national securities association is protected. Displayed depth behind the top of book is not protected. Hidden and reserve interest is not protected. A large order can legally execute through both.

Rule 611(b) supplies the exceptions that make this workable. The most used is the intermarket sweep order, which lets a destination venue execute immediately at its limit price or better provided the sender simultaneously routes sweep orders sized to take the full displayed size of every better-priced protected quote elsewhere. Other exceptions cover single-priced openings, reopenings and closings, crossed markets, flickering quotes, benchmark orders such as volume-weighted average price trades, and stopped orders.

The parameters are moving. Rule 612 originally set a $0.01 minimum quoting increment for NMS stocks at or above $1.00 and $0.0001 below it, and Rule 610 capped access fees for protected quotations at $0.003 per share. Amendments adopted on September 18, 2024 add a $0.005 increment for tight-spread stocks and cut the access fee cap to $0.001 per share; neither was implemented by June 2026. On June 11, 2026 the Commission proposed rescinding Rule 611 and the Rule 610(e) locked and crossed restrictions, with a 60-day comment period. Until the Commission adopts a final rule, Rule 611 stays in force.

Rule 611 is a worst-price boundary, not a duty to the customer. That duty is separate.

Who is on the other side: exchange, internalizer, or dark pool?

A broker holding a marketable order has three broad destinations.

An exchange book is the lit path. The order interacts with displayed quotes posted by a market maker, by other dealers, or by ordinary customer limit orders. Exchange market makers register or receive appointments in specific securities or option classes, and their quoting obligations come from each venue’s rules rather than from a blanket promise to quote everything.

An internalizer fills customer orders as principal. Retail equity brokers route orders to wholesale market makers that supply price improvement and additional size. Measure price improvement against the relevant NBBO at order receipt: the offer for a buy and the bid for a sell. Size improvement measures shares filled beyond the quantity displayed at that price. Fractional-cent execution prices let wholesalers improve on a displayed quote by less than one cent.

A dark pool is an alternative trading system that matches interest without displaying its book. Many pools match at a reference price derived from the public quote, most commonly the NBBO midpoint, which splits the displayed spread between the two sides. Non-display is a pre-trade property only. Once a match occurs the print reports to the consolidated tape and to FINRA like any other off-exchange execution.

The two are commonly conflated. A dark pool matches counterparties who both sent orders in; an internalizer becomes the counterparty. Both print off-exchange, which is why the tape alone cannot tell them apart. See dark pool versus alternative trading system.

Who gets paid, and how does that shape routing?

Two payment models sit under almost every execution.

On exchanges, maker-taker pricing pays a rebate to the order that rested on the book and charges a fee to the order that executed against it. Classification happens at execution, not at entry: a limit order priced through the market becomes a taker instantly, while a limit order priced behind the market becomes a maker only if someone eventually trades into it. Rates are per share in equities and per contract in options. Under the Cboe BZX Options fee schedule effective October 1, 2026, the base customer rates for Penny Program classes are a $0.25-per-contract rebate for adding liquidity and a $0.48 fee for removing it. Qualifying tiers can change those rates. Inverted venues run the model backwards. Cboe EDGX Options rebates customers that remove liquidity and charges market makers that add it, which shortens queues and attracts marketable flow.

Payment for order flow compensates a broker for routing orders to a specified destination. In equities, wholesalers pay for flow they execute as principal. In listed options, routing payments also support orders executed through exchange books and auctions. Follow the payment and the execution separately: the routing arrangement identifies the incentive, while price improvement, effective spread, fill rate, and speed measure the result. Compare PFOF with maker-taker pricing to distinguish broker routing payments from exchange fees and rebates.

That conflict is what best execution addresses. FINRA Rule 5310 requires reasonable diligence to ascertain the best market and trade there so the resulting customer price is as favorable as possible under prevailing conditions. A firm that does not make an order-by-order determination must conduct regular and rigorous reviews of execution quality across destinations. Rule 606 requires brokers to disclose where orders were routed and on what terms, and Rule 605 requires standardized execution-quality statistics from market centers and from covered broker-dealers carrying or introducing at least 100,000 customer accounts used to transact NMS stocks. The SEC’s April 1, 2026 order confirms August 1, 2026 as the compliance date for the expanded Rule 605 reporting.

Inside a venue, price-time priority decides who trades first: best price ranks first, and among orders at the same price the earliest venue timestamp ranks first. A cancel and replace generally resets that timestamp and sends the order to the back of the line. Not every venue uses it. Cboe lists BZX Options as price-time, C2 Options as pro-rata, and EDGX Options as classic pro-rata with customer priority and a designated market maker step, so a small early quote is competitive on one book and nearly irrelevant on another.

Worked example: one order, three routes

A trader wants 1,000 shares. The national quote is $50.00 bid by $50.03 offer, so the quoted spread is $0.03 and the arrival midpoint is ($50.00 + $50.03) / 2 = $50.015. Protected offers sit at 300 shares at $50.03 on Exchange A, 400 at $50.04 on Exchange B, and 1,500 at $50.05 on Exchange C.

Effective spread below uses the signed, direction-adjusted form that SEC Rule 600(b) codifies for the average effective spread published under Rule 605: 2 × (average fill minus the arrival midpoint) for a buy, and 2 × (arrival midpoint minus average fill) for a sell. Doubling the half-spread makes the number comparable to the $0.03 quoted width, and the sign carries the direction: a fill better than the midpoint returns a negative value. Keep the sign. An absolute-value version reports Route 3's captured half-spread as positive $0.03, the same figure a fill 1.5 cents worse than the midpoint produces.

Route 1, internalized. The wholesaler fills all 1,000 shares at $50.0285.

• Price improvement versus the $50.03 offer: ($50.03 − $50.0285) × 1,000 = $1.50
• Cost versus the arrival midpoint: ($50.0285 − $50.015) × 1,000 = $13.50
• Effective spread: 2 × ($50.0285 − $50.015) = $0.027 per share
• Total consideration: 1,000 × $50.0285 = $50,028.50

Route 2, intermarket sweep across the lit book. The router takes every better-priced protected quote and finishes at Exchange C.

• 300 × $50.03 = $15,009.00
• 400 × $50.04 = $20,016.00
• 300 × $50.05 = $15,015.00
• Total: $50,040.00, an average of $50.04 per share
• Slippage versus the $50.03 offer: $0.01 per share, or $10.00
• Cost versus the arrival midpoint: ($50.04 − $50.015) × 1,000 = $25.00
• Effective spread: 2 × ($50.04 − $50.015) = $0.05 per share
• Access fees at the $0.003 per share cap in force: up to $3.00

Route 3, passive limit joining the bid. The order rests at $50.00 behind 4,000 shares already displayed on that venue.

• If it fills: average $50.00, which is $0.03 per share better than the offer, or $30.00
• Cost versus the arrival midpoint: ($50.00 − $50.015) × 1,000 = negative $15.00, meaning half the spread was captured rather than paid
• Effective spread: 2 × ($50.00 − $50.015) = negative $0.03 per share
• Queue position: 4,000 shares are ahead at $50.00. Executions and cancellations reduce the quantity ahead under price-time priority.

The passive route wins on every price column and can lose the trade entirely. That is the tradeoff of market structure in one table: price control and completion certainty are purchased from each other.

The same arithmetic scales in options through the 100-share multiplier. A 20-contract call quoted $1.20 bid by $1.25 offer has a $1.225 midpoint. Paying the offer costs $1.25 × 20 × 100 = $2,500. A Cboe Automated Improvement Mechanism auction fill at $1.22 costs $1.22 × 20 × 100 = $2,440, a $60 saving and an effective spread of negative $0.01 per share. On the same 20 contracts, the maker-to-taker fee swing at the October 1, 2026 BZX base customer Penny rates is (20 × $0.48) + (20 × $0.25) = $14.60, less than the $20.00 that a single penny of option price is worth. Fee optimization rarely beats one tick.

Which order type controls which risk?

A market order sets no price boundary. A limit order sets the boundary and accepts nonexecution risk. A marketable limit crosses the current quote while capping the worst acceptable price, the standard compromise for urgent trades in wide markets. A stop order is inactive until its trigger occurs, then becomes a market order and inherits every market order risk: the stop price is an activation level, not an execution guarantee. Compare market order versus limit order and stop order versus stop-limit order.

Where does slippage actually come from?

Slippage is the difference between a stated benchmark and the volume-weighted average price actually received. Under the standard sign convention, a buy filled above the benchmark and a sell filled below it both count as adverse. The benchmark has to be named, because a buy filled exactly at the displayed ask has zero slippage against that ask and still paid half the spread against the midpoint.

Four separable sources produce it. Half the quoted spread is paid the moment an order crosses from the midpoint to the opposite quote, which is why the effective spread doubles that distance before comparing it with the full quoted width. Market impact is the price movement the order causes by exhausting displayed size and reaching deeper levels. Across large samples it grows with the square root of order size relative to normal daily volume: double the order and the impact rises by less than double. Drift is the market moving for unrelated reasons while the order is worked. Latency is the rest: quotes at national securities exchanges update on microsecond timescales, and displayed size can be repriced while an order is in flight.

That last source is where high-frequency trading enters. In U.S. equity-market usage, SEC staff describe HFT through characteristics rather than a single registration category: extremely high-speed order generation and routing, colocation and proprietary data feeds, very short holding periods, heavy cancellation, and flat overnight books. The market-quality evidence splits along the passive and aggressive divide. The SEC staff’s March 2014 literature review reports a historical Dutch study in which entry by a large, primarily passive firm coincided with 15 percent lower effective spreads and 23 percent lower adverse selection costs (page 22). Other studies in that review distinguish aggressive strategies that improved very short-horizon price efficiency while increasing adverse selection costs for slower passive traders. For an order at a keyboard the consequence is narrow: quotes reprice faster than a human reaction and displayed size understates the true book. Compare market maker versus high-frequency trading.

How does all of this read on a live tape?

Time and sales is where market structure becomes observable. Each row reports execution time, price, size, venue or reporting source, and sale conditions. Read against the contemporaneous quote, the tape shows which of the mechanics above just fired.

A burst of small prints across several exchanges inside a few milliseconds is the signature of a router respecting protected quotes or sweeping to reach them. That is the shape Rule 611 imposes on aggressive institutional activity, and it is the pattern sweep detection keys on.

A single large print carrying an off-exchange identifier can be an internalized fill, a dark pool cross, or a negotiated block reported through a trade reporting facility; the venue field says only that it happened away from an exchange. A print at the midpoint of a wide quote is more consistent with a pool cross or an auction than with a lift of the offer. A print larger than the displayed offer size at one venue means the order reached size the quote never showed, which can be hidden or reserve interest at that venue, a wholesaler improving size beyond the displayed quantity, a sweep across several venues, or a negotiated block.

Depth behavior supplies the counter-evidence. A large displayed bid that vanishes before any trade says nothing about demand. A bid that repeatedly absorbs sellers and replenishes says something real. Bid and ask coloring on the tape comes from matching each trade to the quote at that moment. In fast or wide markets, check the quote behind a print before trusting its side, and read a complex option order as one package, because its legs can print outside their own quotes.

What do traders get wrong about market structure?

Treating the displayed NBBO size as the whole market. Size the order against depth, hidden interest, and size improvement as well as the top of book. The NBBO identifies the best displayed prices and their quoted quantities. In the worked example, the wholesaler fills all 1,000 shares at one improved price even though the best displayed offer contains only 300.

Reading a large displayed bid as support. Lean on a bid only after it trades and holds. Displayed interest is revocable and can be canceled in microseconds. Layering size to fake pressure and pulling it before execution is spoofing, which is prohibited and prosecuted, and an imbalance reading counts that decoy size like any other.

Assuming price improvement proves a good execution. Judge the full execution: effective spread, fill rate, and speed alongside the improvement. Improvement is measured against the quote, so a wide, stale, or thin reference inflates it. A route that improves a fraction of a cent on 10 percent of an urgent order can deliver a worse result than one that completes immediately at the offer.

Treating an off-exchange print as suspicious. Read it against the quote at that moment. A retail marketable order that fills a few tenths of a cent better than the offer, in full, with no matching exchange print behaved exactly as designed.

Assuming Rule 611 covers options. Read the options rules instead. Listed options operate under a separate linkage plan and their own exchange allocation and auction rules.

Confusing a wide spread with a directional signal. Read a wide spread as a condition: liquidity providers perceive volatility, inventory risk, or an approaching catalyst. The same widening precedes sharp moves either way, so take direction from price and flow.

Mistaking volume for liquidity. Read depth at the quote and the speed at which the book refills, because they describe what can trade now. Volume records what already traded.

What to learn next

Start with the quote. Bid, ask, and midpoint and spread are the units every other measurement is denominated in. From there, the NBBO and the SIP explain where the consolidated number comes from and what it leaves out.

Then the routing layer: Regulation NMS for the worst-price boundary, best execution for the customer duty, internalization and payment for order flow for where most retail equity flow goes, and maker-taker pricing plus price-time priority for the incentives and the queue behind every resting quote.

For execution work, pair market orders, limit orders, and stop orders with the costs they produce: slippage, market impact, and price improvement. For reading the tape, combine Level 2, market depth, order book imbalance, and time and sales, then add the venue vocabulary from dark pool, market maker, liquidity, and high-frequency trading.