Market microstructure

Retail traders in the 1990s called a broker, who routed the order to the NYSE floor, where a specialist matched it with someone else's order. The retail trader paid a commission of $20-50. The specialist took a small spread. The trade settled in days.

Retail traders in 2025 tap "buy" in Robinhood, pay zero commission, see a fill in microseconds, and have no idea what happened behind the scenes. The system that produced "free" trading is, on the surface, a triumph of competition. Underneath, it is a transfer system whose winners and losers are not who they appear to be.

The actors

Retail brokers (Robinhood, Schwab, Fidelity, etc.). The customer-facing app. They no longer make money on commissions; they make it on order flow.

Market makers / wholesalers (Citadel Securities, Virtu, Susquehanna, Jane Street, Two Sigma Securities). The firms that actually take the other side of retail trades. They make a tiny spread on each transaction and pay the broker for the privilege of being the first counterparty (payment for order flow, "PFOF").

Public exchanges (NYSE, Nasdaq, Cboe, IEX). The visible "stock market" you see on the news. They host the public order book and produce the price you see quoted — but only a fraction of total volume actually trades through them anymore.

Dark pools / alternative trading systems (ATS). Private trading venues, often owned by big banks (Credit Suisse's Crossfinder used to be the largest; UBS, JP Morgan, and others run their own). They match orders without showing them publicly. Roughly 40% of US equity volume trades off the public exchanges, much of it in dark pools.

High-frequency trading firms (Citadel, Virtu, Jump Trading, Tower, Hudson River, Optiver). Firms that trade in microseconds, profiting from arbitrage between venues, market making, and reading order flow patterns. They co-locate their servers next to exchange data centers and pay millions for microseconds of latency advantage.

Prime brokers (Goldman, Morgan Stanley, JPM). Banks that lend stock for shorting, finance hedge fund positions, and serve as the back-office for hedge funds. They see flow that nobody else sees.

Where the money actually moves

The economic flow of a typical retail trade in 2025:

You pay $0 commission. Citadel Securities pays your broker (Robinhood) ~$0.20 per options contract, or a fraction of a cent per share, for the right to fill your order. Citadel's market-making algorithm takes the other side at a price slightly better than the National Best Bid and Offer (NBBO), and pockets the spread — which, on liquid stocks, might be $0.001 per share, but multiplied across hundreds of millions of trades per day, is real money. The broker passes you a slightly better fill than you'd have gotten on the public exchange. Everyone is technically better off than under the old commission model.

But — and this is the case the critics make — the public quote you saw was set by traders who weren't getting a chance to interact with your order. The "price improvement" you got is measured against the NBBO, which itself is degraded when 40%+ of volume happens off-exchange. The spread you save on your trade is funded by adverse selection imposed on the public order book. Whether this is net good or net bad for retail traders is contested; whether it makes Citadel Securities one of the most profitable trading firms in human history is not.

The HFT layer

High-frequency trading is the most caricatured corner of finance and the most worth understanding precisely. HFT firms compete on speed because tiny price discrepancies between venues exist for microseconds before they're closed. The race produces:

Co-location. Exchanges sell space inside their data centers so HFT firms can put servers next to the matching engine. The exchanges make millions in colo revenue; the firms get a few microseconds of edge.

Microwave networks. Light through fiber is slower than radio through air. HFT firms have built microwave relay networks between Chicago (futures) and New Jersey (equities) to shave milliseconds. Some have laid private dark fiber lines between data centers. McKay Brothers and others sell this infrastructure as a service.

Specialized hardware. Field-programmable gate arrays (FPGAs) and ASICs do trade decisions in hardware rather than software, shaving microseconds further.

This is, in Michael Lewis's term, the "Flash Boys" problem: a tax on every trade, paid to whoever owns the fastest pipes. The defenders' counter is that HFT has compressed spreads from cents to fractions of a cent and lowered transaction costs across the board. Both can be true; both are.

Dark pools and the visibility problem

The public stock market — the prices you see scrolling on CNBC — increasingly is not where the trading happens. Dark pools were originally a service for institutional investors who needed to move large blocks without moving the public market (a fund trying to buy 10 million shares of XYZ doesn't want every HFT firm front-running them). The pools matched these large orders privately.

That basic logic still exists, but dark pools have expanded to handle huge slices of retail and mid-size order flow too. Their critics argue this fragments the market and degrades the quality of public price discovery. Their defenders argue they reduce market impact and provide better fills. Both arguments have empirical support.

The structural concern: the price you see on screen is increasingly an inference about what trades would happen on the public exchange if all the trades happening privately were forced into the open. We don't know what the "true" price is anymore in any of the old senses.

Index fund volume and the price discovery problem

One last subtle point. Index funds, by their nature, don't do price discovery. They buy stocks because the stocks are in an index, not because they've decided the stocks are cheap. As index fund volume has grown to 30%+ of US equity trading (closer to 50% if you include other passive strategies), the share of trading actually trying to figure out what stocks are worth has shrunk.

The implication: price discovery now happens on smaller and smaller margins of active capital. Mispricings can persist longer than they used to. Volatility can be sharper. And when index funds rebalance (quarterly, or when the underlying index changes), they create predictable demand pressure that active traders front-run. This is one of the cleanest documented edges in modern markets.