How Wall Street Works
Dark Pools, HFT and Reg NMS: Modern Market Structure Explained
The U.S. equity market is dozens of venues held together by one rule about best price. What that rule created, who profits from the seams in it, and where the safety switches sit.
There is no such thing as the stock market. There are more than a dozen registered exchanges, a similar number of private venues and a small group of wholesale market makers, all trading the same securities at the same instant, in buildings a few miles apart in New Jersey. One rule about best price is what makes them behave as a single market, and everything strange about modern equity structure follows from that rule.
Lit and dark
A lit venue displays its order book. Anyone can see the resting bids and offers, with size, before trading against them. The registered exchanges are lit.
A dark venue accepts orders without displaying them. Matching happens internally, commonly at the midpoint of the public quote, and the trade prints to the consolidated tape afterwards. Nothing is hidden after the fact.
Dark venues exist because of information leakage. An institution that needs to sell 3 million shares of a stock that trades 800,000 shares a day cannot show that order, because the moment the size is visible everyone else reprices and the seller pays for its own announcement. Working the order in the dark, in pieces, reduces that cost. Whether it reduces it enough is a live argument among people who do this for a living.
| Venue | Pre-trade transparency | Typical print size | Who uses it |
|---|---|---|---|
| Registered exchange | Full order book displayed | Small, often 100 to 300 shares | Everyone |
| Dark pool | None | Larger blocks | Institutions working size |
| Wholesale market maker | None | Retail-sized | Retail flow routed by brokers |
Reg NMS built all of this
Regulation National Market System, adopted in 2005, turned competing venues into one linked market.
The national best bid and offer is the highest displayed bid and the lowest displayed offer across all exchanges at a given instant, and the order protection rule requires a venue to route away when another venue displays a better price. If exchange X is quoting 50.03 while exchange Y shows 50.01, X cannot fill your buy at 50.03. Y’s better offer is sitting there.
Follow that requirement to its conclusion. Every venue must know every other venue’s quote continuously, which means consuming market data feeds and reacting to them in microseconds, and that is how being marginally faster at seeing and acting became a business. The physical distance between data centers became a line item. A rule written to guarantee you the best displayed price is the reason firms pay for microwave towers.
What high-frequency actually covers
High-frequency trading is a speed category. Treating it as a strategy is where most public argument about it goes wrong. At least four distinct activities live inside the label.
Automated market making is the largest by volume: quoting both sides in thousands of names continuously, with the economics described in market makers and liquidity, and it is the same business as manual market making with the holding period cut to seconds and the inventory managed by machine.
Latency arbitrage trades the gap between a price changing at one venue and that change arriving elsewhere, so a firm that sees the move first trades against stale quotes still showing the old price. The counterparty being picked off is usually another professional.
Statistical arbitrage trades short-lived relationships between correlated instruments, such as an index ETF against the basket of stocks inside it.
Rebate capture works the fee schedule. Under maker-taker pricing an exchange pays a rebate of a fraction of a cent per share to orders that rest in the book and charges a fee to orders that remove liquidity, so a firm doing 100 million shares a day at a 0.20 cent rebate collects 100,000,000 x $0.0020 = $200,000 before any trading profit or loss.
One 200 share order, three routes
Take a retail buy of 200 shares. The NBBO is 49.99 by 50.01.
Routed to a wholesaler, it might fill at 50.005, saving you 200 x $0.005 = $1.00 against the public offer, and the wholesaler takes the rest of the spread on the other side of its book and pays your broker a fraction of a cent per share for the flow.
Routed to a lit exchange as a marketable order, it fills at 50.01 and the broker pays a taker fee of roughly 0.30 cents per share: 200 x $0.0030 = $0.60.
Routed as a resting limit at 49.99 that gets hit, the broker collects a maker rebate of roughly 0.20 cents per share: 200 x $0.0020 = $0.40, and you bought two cents under the offer.
The dollar amounts on 200 shares are trivial. That is the point. Multiply them by the tens of billions of shares that trade daily and you have the revenue of most of the firms mapped in what is Wall Street.
Limit up limit down
Single stock volatility controls came out of the May 2010 flash crash, when some stocks printed at absurd prices for seconds before recovering.
Each stock now carries a price band around a rolling five minute average of its own trades. Trades cannot print outside the band. If the stock sits at a band edge for fifteen seconds without moving back inside, it pauses for five minutes and reopens with an auction.
Band widths depend on the stock’s price and reference tier, and they double in the opening and closing periods when volatility is normally higher. The mechanism removes the worst prints from a disorderly minute. It also freezes your position in the five minutes you most want to act on it.
Market-wide circuit breakers
Above the single stock layer sits a market-wide one, keyed to the S&P 500 against the previous session’s close.
| Level | S&P 500 decline | Action |
|---|---|---|
| 1 | 7 percent | 15 minute halt, only before 3:25 p.m. Eastern |
| 2 | 13 percent | 15 minute halt, only before 3:25 p.m. Eastern |
| 3 | 20 percent | Trading stops for the remainder of the day |
Levels 1 and 2 trigger once each per day. After 3:25 p.m. Eastern only the 20 percent level applies, so a market down 8 percent at 3:40 p.m. keeps trading into the close.
What follows from all of it for you
Displayed size understates real liquidity in large stocks, because much of the available interest is hidden in dark venues or in reserve orders, and it overstates liquidity in small stocks, where the same order can appear on several venues at once and get counted twice by anyone reading a depth screen.
Speed competition is over. You were never in it. Any strategy whose edge depends on reacting to a public data release faster than other participants is competing with dedicated microwave links and colocated servers, so pick a horizon where that does not decide the outcome.
The mechanisms that protect you are the same ones that constrain you. Reg NMS gets your 200 share order the best displayed price. Limit up limit down and the circuit breakers stop the market at precisely the moments getting out matters most. Both are true at once. Any account of market structure that tells you only one of them is selling something.
Market microstructure goes deeper into order book dynamics and adverse selection, and stock market hours covers how these venues behave when the session ends. For the ground floor, read how the stock market works and order types explained. The market microstructure quiz tests this material, and the market breadth tool shows what the tape looked like on days when these mechanisms were tested.
Frequently asked questions
What is a dark pool?
A dark pool is a trading venue that does not display its orders publicly before they execute. Institutions use them to work large positions without revealing size to the market, which would move the price against them. Trades still print to the consolidated tape after execution, so the transaction is visible once it is done.
What is Reg NMS and the NBBO?
Regulation National Market System is the set of SEC rules governing how U.S. equity venues interact. The national best bid and offer is the highest bid and lowest offer displayed across all exchanges at any moment. The order protection rule requires venues to avoid executing at prices worse than the NBBO, which keeps quotes across dozens of venues aligned.
Is high-frequency trading harmful to retail investors?
The evidence is mixed and depends on which activity you mean. Automated market making has coincided with much narrower spreads than the era of manual quoting, which benefits small orders. Latency arbitrage strategies extract value from slower participants, mostly institutional ones. For a 200 share order the net effect is likely positive, and for a large institutional order it is contested.
What are limit up limit down bands?
A single stock mechanism that prevents trades from printing outside a percentage band around a rolling average of recent prices. If the price sits at the band edge for fifteen seconds without moving back inside, the stock pauses for five minutes. The bands are wider for lower priced stocks and near the open and close.
When does the whole market halt?
Market-wide circuit breakers are keyed to the S&P 500 falling against the previous close. A 7 percent decline triggers a 15 minute halt, a 13 percent decline triggers a second 15 minute halt, and a 20 percent decline closes the market for the remainder of the day. The first two levels only apply before 3:25 p.m. Eastern.
What is maker-taker pricing?
An exchange fee model that pays a rebate to orders that rest in the book and provide liquidity, and charges a fee to orders that remove it. The rebate is a fraction of a cent per share. It shapes where brokers route orders, which is why execution quality reports and routing disclosures exist.