Part II · Markets · Chapter 2B

Where your order goes

Exchanges, wholesalers and the best bid and offer

Most retail traders see "buy" and "sell" as primitive actions. They are not. Between your click and the actual counterparty sits a routing decision that determines who fills you, at what price and on what venue. Understanding the routing layer is the difference between believing that "the market" took the other side of your trade and knowing that a specific wholesaler with a specific business model did.

1 · The NBBO: the rule everyone is supposedly following

Since the Securities and Exchange Commission (SEC) adopted Regulation NMS in 2005, American equities and single-name options have been required to execute at or better than the National Best Bid and Offer (NBBO). The NBBO is the highest bid and the lowest offer displayed across the registered exchanges at the moment of execution. It is computed continuously by the Securities Information Processor (SIP), a regulated infrastructure that aggregates quotes from every venue and publishes a consolidated tape of them.

When you submit a market order, your broker must route it to a venue that can fill you at or inside the SIP NBBO at that microsecond. This sounds like a hard constraint. In practice it is a soft one, and the reality of execution is more interesting than the regulation suggests.

2 · Three kinds of venue

American equity orders can be routed to three types of venue: lit exchanges, dark pools and wholesalers.

Lit exchanges

Lit exchanges display their order books publicly. As of 2025 there are 16 registered American equity exchanges. The major ones are NYSE, Nasdaq, Cboe BZX, Cboe BYX, Cboe EDGX, Cboe EDGA, IEX (the "speed-bump" exchange), MEMX, MIAX Pearl, NYSE Arca, NYSE American and several others. Each one has its own fee schedule, its own latency profile (how fast it handles an order) and its own matrix of order types. Institutional orders typically route across several lit exchanges to minimize their impact on the price.

Dark pools

Dark pools, formally called alternative trading systems (ATSs), are non-displayed order books run by banks and brokers. Examples are UBS ATS, Goldman Sigma X, JP Morgan JPM-X, Credit Suisse CrossFinder, Barclays LX, Liquidnet and many others. Orders rest in a dark pool without being publicly displayed, and they fill against incoming orders at the prevailing NBBO mid or at negotiated levels of price improvement. Because they display nothing, dark pools trade at or inside the NBBO rather than setting it. Dark pools handle roughly 12% to 15% of American equity volume on a typical day. On stress days the share can rise, as institutional sell flow seeks anonymity.

Wholesalers

Wholesalers are off-exchange principal liquidity providers. They pay broker-dealers for the right to internalize retail order flow against their own principal books, rather than route the orders to exchanges. Trading as principal means that the wholesaler is itself your counterparty, filling your buy from its own inventory. The American wholesaler hierarchy in retail equities, as of late 2025 and early 2026 data, runs as follows:

• Citadel Securities: approximately 35% of all American-listed retail equities and options volume, and about 25% of total equities trading across all customer types.
• Virtu Financial: approximately 15% to 20%, the historical second place.
• Hudson River Trading (HRT): approximately 20% of American cash equities, institutional and retail together. It launched in retail wholesaling in 2022.
• Jane Street: approximately 10% of American cash equities, dominant in ETF and institutional trading. It launched retail wholesaling in 2019.
• G1 Execution Services (Susquehanna): a smaller share, historically dominant in options.
• Wolverine Trading, IMC, DRW and Optiver: institutional and option-specific franchises.

Read these shares with their denominators. Citadel's 35% is of retail volume, while HRT's 20% and Jane Street's 10% are of all cash equities, institutional flow included. The percentages describe different populations, so they cannot be compared directly or added up.

The numbers shift by product. Citadel Securities is overwhelmingly dominant in retail equities. Jane Street has emerged as the dominant ETF market maker. Susquehanna and Wolverine retain large institutional options franchises. Concentration in retail flow has increased since 2020 despite the entry of HRT and Jane Street, because the franchise economics favor scale.

For American-listed options specifically, the wholesaler share is even higher than for equities. Citadel Securities, Susquehanna and Wolverine together execute roughly 85% of retail options order flow. Most retail options orders never touch a public order book.

Jane Street disclosed revenue of $20.5 billion in 2024 and $39.6 billion in 2025 from its market-making operations. The rough size of those numbers tells you how thick the layer of intermediation is between you and any retail counterparty with an actual directional view.

3 · Payment for order flow

Payment for order flow (PFOF) is the payment a wholesaler sends back to the broker for the right to fill that broker's order flow. The wholesaler earns the bid-ask spread on each trade, less any price improvement it chooses to give you. It shares a small portion of what it earns with the broker, as a per-contract or per-share kickback.

For American-listed equities, PFOF runs in the range of $0.001 to $0.003 per share, roughly 0.1 to 0.3 basis points on a $100 stock, which is small. For American-listed options, PFOF is much larger relative to position size: $0.40 to $0.65 per contract per side for retail SPY and QQQ options. On a 1-contract retail SPY options trade, the broker is paid roughly $0.50 to internalize it through a wholesaler. The wholesaler earns multiple dollars of edge on the trade, as compensation for taking the position and the hedge obligation that comes with it.

To compare the two, divide the payment by the value of what was traded and express it in basis points:

On a $100 stock, $0.002 a share is $0.002/100 \times 10{,}000 = 0.2$ basis points. On a hypothetical SPY option priced at $2.00, which costs $200 a contract, a $0.50 payment is $0.50/200 \times 10{,}000 = 25$ basis points of the premium, over a hundred times as much in proportion.

PFOF is controversial because the broker's incentive is to route to the highest-paying venue, which is not necessarily the venue with the best execution. The historical academic debate has been whether retail execution under PFOF is genuinely worse than execution on lit exchanges, after accounting for the price improvement that wholesalers do typically provide. The empirical evidence is mixed and depends on the metric used.

Rule 615, proposed and withdrawn

In December 2022, the SEC under then-Chair Gary Gensler proposed Rule 615, the "Order Competition Rule". It would have required certain retail marketable equity orders that were going to be internalized to be exposed instead to a public auction, to potentially capture price improvement competitively. A marketable order is one priced to trade at once against the NBBO (Chapter 2C). The Commission voted 3-2 to propose the rule, but it was never finalized.

On June 17, 2025, the SEC under the new administration formally withdrew Rule 615, along with twelve other Gensler-era proposed rules. After that withdrawal, PFOF remains legal in the United States. The UK and EU regulators have banned PFOF outright. The United States allows it with disclosure requirements, the Rule 605 and Rule 606 reports, but with no structural intervention against the wholesaler-internalization business model.

For the foreseeable future, the wholesaler model is the operational reality of American retail trading.

4 · Rules 605 and 606: the only public lens on execution quality

Two disclosure rules under Regulation NMS let you see, on average, what happens to retail orders.

Rule 605, in force since 2001 and amended substantially in 2024, requires each market center to publish monthly reports on execution quality at the order-type level. A market center is an exchange, an ATS or a wholesaler that internalizes. The mandatory metrics include the realized spread, the effective spread, the price-improvement rate and the average response time. Since the 2024 amendments, the reports also break these down by order size and by time-of-day bucket. They are downloadable as plain text from each market center's website.

Rule 606, in force since 2001 and amended in 2018 and 2020, requires each broker-dealer to publish quarterly reports on where it routed retail and institutional non-directed orders. A non-directed order is one whose customer did not choose the venue. The report gives the proportion of total flow sent to each market center. Since 2018 it also includes the dollar amount of PFOF received from each routing venue, broken down by S&P 500 against non-S&P stocks, by market against marketable limit orders, and by other categories.

Reading the spread metrics

Two of the Rule 605 metrics measure what an order paid. Write $P$ for the fill price, $M_0$ for the NBBO mid when the order arrived and $M_1$ for the mid a short interval later. Let $d = +1$ for a buy and $d = -1$ for a sell. Then

The effective spread is what you paid against the mid, doubled so that it compares with a quoted spread: a buy at the ask of a 2-cent market has an effective spread of 2 cents. The realized spread is what the firm that filled you kept, once the price had moved on. The difference between them, $2d\,(M_1 - M_0)$, is the price impact: how far the mid moved in the direction of your trade. Price improvement is measured from the NBBO on your side: for a buy, the ask minus your fill, and for a sell, your fill minus the bid.

The figure puts that fill on a price axis. Drag the later mid: the effective spread stays at 1.0¢ while the realized spread and the impact shift against each other.

order buy sell

fill P

$100.035

later mid M₁

$100.040

effective = +1.0¢

realized = −1.0¢

impact = +2.0¢

improvement = +0.5¢ a share

A buy at $100.035. Price improvement is 0.5¢ a share, $0.50 on 100 shares. The effective spread is +1.0¢, the realized spread −1.0¢ and the impact +2.0¢: the firm that filled you lost 1.0¢ a share against the later mid.

What the reports show, and what they do not

The Rule 605 reports show averages of execution quality. An illustrative line might say: "Citadel Securities provided $0.0017 per share of price improvement on average for SPY market orders sized 100-499 shares in October." This is informative but coarse. Two important things the reports do not show:

1. Order-level execution quality: you cannot reconstruct your own trades from them.
2. How price improvement was distributed across orders: a wholesaler can boost the average by improving large favorable trades while leaving small marginal trades at the NBBO.

Suppose a wholesaler fills one 1,000-share order with half a cent of improvement and ten 100-share orders at the NBBO. Weighted by shares, the average improvement is $1{,}000 \times 0.005/2{,}000 = \$0.0025$, a quarter of a cent a share, although ten of the eleven orders got nothing.

For options, the equivalent disclosures are weaker still. The Rule 605 framework was built around NMS stocks, and option-specific reporting standards are still evolving.

In practice, to compare brokers you read two sets of reports together. Your broker's Rule 606 reports answer where it sends your orders. The Rule 605 reports of those venues answer what execution quality they deliver, on average. The combination gives a defensible, like-for-like comparison across brokers. It is the single most underused public dataset in retail execution analysis. Most retail traders have never opened either report, and both are public.

5 · The franchise: one firm on both sides

Citadel Securities, Susquehanna International Group and Wolverine are not just options market makers. They are also the dominant retail equity wholesalers. The same firm that takes the other side of your SPY put option also takes the other side of your SPY share purchase. The two desks share infrastructure, capital and inventory, and they hedge across each other. When Citadel sells you a SPY put, the SPY shares it sells as the hedge may go to another piece of internal Citadel flow, crossing the book entirely without touching a lit exchange.

This is the internalization model in its full form. Retail buy-side flow and retail sell-side flow are netted against each other inside the wholesaler's book. Only the residual, the net of all retail buys against retail sells, actually needs to be hedged in the external market. Retail flow is roughly balanced on most days, with slightly more buying than selling on average but only a small imbalance. So the gross flow the wholesaler sees is far larger than the net hedging flow it actually generates.

No wholesaler for SPX

For SPX, XSP and other index options, the routing is simpler. They trade only on Cboe, and most retail brokers route them directly to the Cboe order book. There is no wholesaler bypass for SPX. This is one of several reasons SPX and XSP spreads are somewhat wider than SPY's: the wholesaler price-improvement machinery does not apply.

It is also one of the reasons that gamma exposure (GEX) models computed from SPY open interest can be misleading, a point Chapter 8B develops. Open interest (OI) is the number of contracts outstanding at a strike. SPY open interest is gross, but the hedge flow that SPY options actually generate is much smaller than that gross, because wholesalers internalize so much of it. SPX open interest, by contrast, represents flow that all sits on the Cboe book. Wholesalers cannot internalize it because they are not the counterparty. So the dealer-hedging implications of SPX open interest are larger, per contract of open interest, than those of the same open interest in SPY.

6 · FLEX options: the institutional sidewalk

A growing share of institutional options trading happens through FLEX options: exchange-listed contracts with customizable strikes, expirations and exercise styles, traded primarily on Cboe (Chapter 4A). FLEX volumes have exploded:

• In 2012, about 12 million FLEX contracts traded in the year.
• In 2019, about 125,000 contracts traded a day, with total open interest of 2 million.
• In 2024, about 200 million FLEX contracts traded in the year.
• In 2025, average daily volume reached 1.3 to 1.4 million contracts a day, up 65% on the year before. By type, single-stock FLEX averaged about 958,000 a day, ETF FLEX about 301,000 and index FLEX about 29,000. Open interest reached 35 million contracts, 17.5 times the 2019 level.

FLEX growth has tracked institutional demand for tailored hedges. A pension fund may want a put struck exactly at its actuarial liability level, not at the nearest 25-point strike on the listed grid. The issuer of an autocallable, a structured product whose payoff switches at preset index levels, may need to hedge a knock-in level at a specific spot price. A structured-product desk may need a non-standard expiration to match an internal liability schedule. All of them use FLEX. The dealer on the other side of a FLEX position hedges it alongside its listed book. The resulting hedge flows show up in the lit markets without any FLEX identifier on the trade tape.

This matters for GEX modeling. FLEX positions are visible to Cboe, which records the trades. They are not visible on the public OPRA feed, the consolidated feed of options trades, in a way that attributes them to specific strikes for a GEX computation. The major published GEX models (SqueezeMetrics, SpotGamma, Tier1Alpha) all see FLEX volume in aggregate but cannot decompose it cleanly by strike. The result: published GEX models systematically under-count the gamma exposure dealers actually have. This effect is small on a typical day and large during stress episodes, when FLEX-based hedging activity spikes. Chapter 8B returns to it in detail.

7 · What this means for dealers

Dealer hedging, the subject of every chapter from Part VI on, is executed primarily in two places. Delta hedges go to shares on the lit exchanges, SPY or the SPX component stocks, and larger or more cost-sensitive hedges go to ES futures on the CME. Chapter 2A gives the reasons an SPX dealer leans on ES and uses shares only at the margin. The wholesaler ecosystem described in this chapter is one piece. The dealer-hedging ecosystem on top of it is another.

Baltussen, Da, Lammers and Martens (2021) found that the last 30 minutes of trading concentrate intraday momentum, and Chapter 6C examines their paper. The finding is partly mechanical, because dealers' end-of-day rebalancing flows through the closing auction, the matching of orders that sets the official close (Chapter 2C). There it interacts with mandatory index-fund rebalancing and with the day's accumulated charm and vanna decay from the options book. Charm and vanna are the rates at which an option's delta drifts with time and with implied volatility (Chapter 5D). The auction is where the day's options-related hedge obligations get settled into the underlying.

One subtle implication of wholesaler concentration is that SPY options flow is largely internalized. The wholesaler sees both retail buys and retail sells and crosses them internally, so the actual dealer delta hedge it needs is smaller than the gross retail flow suggests. SPX options flow, by contrast, all hits the Cboe order book, and the institutional dealer there has to hedge the full gross position, not the net. This is one reason why "SPX dealer gamma" and "SPY dealer gamma" are not literally the same concept. It is also why SqueezeMetrics-style GEX calculations on SPY tend to overestimate the actual hedge flow.

8 · Check yourself