The Order Book Is Not the Market Anymore

01 — What the Order Book Was Supposed to Be

The exchange order book is elegant in theory. Buyers post bids; sellers post offers; when they overlap, a trade prints; the price of that trade becomes the market's signal. The limit order book is a continuous double auction, and for most of modern financial history it was the primary mechanism by which supply and demand discovered a clearing price in real time. The specialist system on the NYSE, the market makers on NASDAQ — these were mechanisms to ensure the book stayed liquid, but the central idea held: price was formed in a public, transparent, centralised venue.

Regulation drove this model. Consolidated tape, best execution obligations, and the fragmentation reforms of the early 2000s — Reg NMS in the United States (SEC, 2005), MiFID in Europe — were built on the assumption that competition between lit venues would sharpen price discovery. More exchanges, more competition, tighter spreads, better prices. The order book was the temple. Regulation built the pews.

Then the algorithms arrived. And they did not worship there.

02 — How Algorithms Colonised, Then Abandoned, the Book

The first generation of algorithmic trading did engage with the order book directly — indeed, the whole point was to pick it apart. Statistical arbitrage strategies searched for cross-asset mispricings. Market-making algorithms replaced human specialists, posting quotes in milliseconds, harvesting the spread. Execution algorithms — VWAP, TWAP, implementation shortfall — broke large orders into fragments and fed them into the book in a way that minimised market impact.

But as algorithms multiplied and competed, the order book itself became a hostile environment for genuine price discovery. The book is observable. Everything posted to a lit exchange is visible to any market participant with co-location access and a fast enough data feed. This created an adversarial dynamic: any resting order in the book became information that other algorithms could trade against. A large buy order posted to the book is, in effect, a signal that demand exists — and sophisticated high-frequency strategies could detect and front-run that signal before the order was filled.

The order book became a strategic environment — not a neutral marketplace. Price formation migrated to wherever it was hardest to detect and exploit.

The rational response, for any large order, was to hide. And an entire infrastructure was built to accommodate exactly that.

03 — Where Price Is Actually Formed Now

Dark Pools and Alternative Trading Systems

Dark pools — Alternative Trading Systems (ATS) in US regulatory parlance — are private matching venues that execute trades without pre-trade transparency. There is no visible order book. Orders are posted and matched internally, with prices typically referenced to the prevailing National Best Bid and Offer (NBBO) from the lit market. The irony is profound: dark pool prices are derived from the lit market's order book, yet the dark pool has become, by volume, a major share of all activity. By 2012, dark pools and internalizers already accounted for some 40% of US equity volume (Wikipedia — Dark pool, 2024); recent estimates put total off-exchange execution — dark pools combined with internalised retail flow — at roughly 35–45% of shares traded. (HeyGoTrade, 2024) The parasite has come to rival the host.

Internalisation and the Wholesaler Model

For retail order flow, the mechanics are even more removed from the traditional order book. Payment for order flow (PFOF) directs retail orders from brokers to market makers — wholesalers such as Citadel Securities, Virtu, or Jane Street — who execute those orders off-exchange, internalising them at prices marginally better than the NBBO. (SEC — Gensler, 2024) The exchange sees none of this flow. The trades print on the tape, but they never interact with the order book.

This has profound implications. Retail order flow — historically understood to be relatively uninformed, providing liquidity to the market — is now siphoned away from exchanges entirely. What remains on the lit book is increasingly the province of professional algorithmic strategies trading against each other: HFT firms, arbitrageurs, institutional algorithms. The composition of exchange volume has shifted dramatically toward sophisticated, fast, and information-rich participants. The book is a sparring ring for professionals, not a price-discovery venue for the market at large.

Derivatives and the Tail Wagging the Dog

For many of the most economically significant assets, price discovery has migrated not just off-exchange but off the cash market entirely. Equity index futures — S&P 500 futures on the CME, for instance — trade nearly around the clock and with vastly more leverage per dollar than the underlying stocks. When macro news breaks at 3am, the S&P futures move first. The cash equity market, when it opens, adjusts to that signal. The tail is wagging the dog.

Options markets present a similar dynamic. The notional size of the options market on major US equities and indices now dwarfs the cash market on any given day. Because options dealers must dynamically hedge their books, their hedging activity — the buying and selling of underlying equities to remain delta-neutral — transmits options market structure directly into equity prices.

The 0DTE effect. Zero-days-to-expiry options on the S&P 500 accounted for 57% of all SPX options volume in Q3 2025, according to Cboe data. (Cboe, Q3 2025) Because these options expire the same day, their gamma — the rate of change of delta — is enormous near the strike price. Options dealers who have sold these contracts are forced to hedge aggressively as the underlying moves, amplifying price swings rather than dampening them. This is a feedback loop with no historical precedent in equity microstructure. Price formation in this environment is not supply and demand meeting in a book. It is a mechanical, model-driven consequence of dealer positioning — a structural artifact of derivatives market architecture rather than any expression of fundamental value.

ETF Arbitrage and the Creation/Redemption Mechanism

The explosive growth of exchange-traded funds introduces yet another layer of price formation complexity. ETFs trade on exchanges like stocks, but their prices are kept roughly in line with their net asset values through an arbitrage mechanism involving authorised participants (APs). When an ETF trades at a premium, APs create new shares by delivering the underlying basket; when it trades at a discount, they redeem shares for the underlying. This continuous arbitrage links ETF prices to basket prices, and basket prices to individual component prices.

The consequence is that large ETF flows generate mechanical buying and selling of constituent securities — not driven by any view on those securities' individual merits, but by the arithmetic of the creation/redemption basket. In a market where passive ETF assets rival or exceed active ones, this mechanical flow is substantial. Price formation in individual equities is increasingly contaminated by these portfolio-level flows, which have nothing to do with the order book dynamics of those individual names.

04 — Algorithmic Trading as Price Formation Infrastructure

It would be a mistake to conclude from all of this that price discovery has collapsed or that prices have become arbitrary. Algorithms are, in many contexts, extraordinarily good at processing information and incorporating it into prices rapidly. The question is not whether prices are informationally efficient — they may well be more efficient, in some narrow technical sense, than they have ever been — but rather what the mechanism of price formation actually looks like, and what it means.

The dominant algorithmic strategies operating across these venues can be classified roughly into three categories with respect to price formation.

• Market-making algorithms provide continuous two-sided quotes and are the primary mechanism by which liquidity is maintained in fragmented markets. HFT market makers now contribute more than half of displayed depth on leading equity venues. (Mordor Intelligence, 2026)
• Arbitrage algorithms — statistical arbitrage, ETF arbitrage, cross-asset — continuously enforce relationships between related instruments, transmitting price signals across markets at near-zero latency.
• Directional algorithms, including trend-following and momentum strategies, trade on signals derived from price itself, creating the potential for feedback loops and momentum cascades.

What these strategies share is that their behaviour is determined by models, parameters, and market conditions — not by fundamental views on intrinsic value. Price formation in heavily algorithmic markets is increasingly the output of a complex adaptive system of interacting models, each responding to the outputs of the others. The order book is one input among many, and often not the most important one.

05 — The Implication No One Wants to Confront

If price formation has migrated away from the exchange order book, then the entire regulatory and conceptual apparatus built around the order book is addressing the wrong thing. Best execution obligations calibrated to the NBBO are measuring a derived, reference price — not the price at which most actual economic interest is transacted. Transparency requirements for lit markets are providing visibility into a venue that represents a shrinking fraction of genuine price discovery. Systemic risk monitoring focused on exchange activity may be missing where the real leverage, interconnection, and fragility actually lives.

Regulators built cathedrals to centralised price discovery. The congregation moved to the basement, the alley, and the derivatives market — and no one told them.

There is also a deeper epistemological problem. The canonical model of market microstructure — the Glosten-Milgrom model of bid-ask spreads under adverse selection (Glosten & Milgrom, 1985) and Kyle's model of sequential auctions with an informed insider (Kyle, 1985) — assumes that price formation is a process by which informed traders gradually reveal private information through their order flow, and market makers update their quotes accordingly. This model made sense when humans with fundamental views were the primary participants. In a market dominated by algorithmic strategies that have no fundamental view — that are trading on signals derived from price, volatility, positioning, and flow — the information content of order flow is radically different. The price formed may be technically efficient without being economically meaningful in the way the theoretical framework assumes.

06 — The Order Book as Echo Chamber

What remains in the lit order book of major exchanges is, in many ways, an echo of a market that has largely moved elsewhere. HFT market makers post quotes as obligations of their strategy, not as expressions of genuine value opinion. Institutional execution algorithms post and cancel orders as part of a tactical game of information concealment. The visible order book is, simultaneously, a venue, a signal, and a trap — and every sophisticated participant treats it as all three.

The spread quoted in the book is narrow. The depth is often illusory — available in microseconds, vanishing the moment a real order appears. The price printed on the tape may be the NBBO at the instant of execution, but that NBBO is itself a construction of algorithmic market-making models across fragmented venues, not an organic clearing of supply and demand.

This does not mean markets are broken. Liquidity in major instruments is, by most measures, abundant. Transaction costs have fallen dramatically over the past two decades. Information is incorporated into prices quickly. But the mechanism by which all this happens bears almost no resemblance to the order book model that still dominates how markets are taught, regulated, and discussed.

— Where Does This Leave Us?

Understanding price formation in contemporary markets requires abandoning the order book as the organising metaphor and replacing it with something more honest: a distributed, multi-venue, multi-asset, algorithmic system in which price is the emergent output of interacting models operating across lit markets, dark pools, derivatives markets, and bilateral trades, at timescales ranging from microseconds to days.

For practitioners, this means that the order book is an input, not a destination — and for large orders, it is often the last place you want to be. Understanding dealer positioning in options markets, ETF creation/redemption flows, futures-cash basis dynamics, and the inventory constraints of major market-making firms is at least as important as reading the equity order book. The information that matters is no longer centrally located or publicly visible.

For market structure researchers, the challenge is to develop models of price formation that accurately describe the system as it actually operates — multi-agent, multi-venue, feedback-driven — rather than optimising theories built for a world that no longer exists.

For regulators, the challenge is frankest of all: the rules were written for a market where the order book was the locus of price discovery. In a market where more than a third of equity volume never touches an exchange order book (HeyGoTrade, 2024), where derivatives dwarf the cash market in notional terms, and where the primary liquidity providers are high-frequency algorithms with no long-term view, the rules need to catch up to the reality. The order book is not the market anymore. The sooner that is acknowledged, the sooner the right questions can be asked.