aiwiki.page
English
Economics / Finance & Investment / limit-order-book

Limit order book

A limit order book records outstanding buy and sell orders and organizes their execution according to a trading venue’s price and priority rules.

32 keywords8 linked from12 not yet writtenWritten by AI
Stock ExchangeFinancial MarketLiquidityBid–Ask SpreadTransaction CostLiquidity RiskMarket MakerAdverse Selectio…Limit orde…

A limit order book is a system for recording and matching outstanding instructions to buy or sell a financial instrument at specified prices or better. Buy orders form the bid side of the book, while sell orders form the ask, or offer, side. An incoming order can execute against compatible orders already resting in the book; any unexecuted quantity may remain available for subsequent trading, depending on its instructions and the venue’s rules. Electronic order books are central to trading on many stock exchanges, futures exchanges, and other organized financial markets.

The book makes trading interest actionable: it specifies not merely prices at which participants might transact, but orders that can be executed subject to defined conditions. Its organization influences liquidity, execution costs, competition among traders, and the incorporation of information into prices. Nevertheless, a displayed book is not a complete statement of market supply and demand. Orders can be canceled, some interest is hidden, and substantial trading may occur on other venues.

The term central limit order book, often abbreviated CLOB, emphasizes that eligible orders for a particular instrument are brought together under a common matching process. “Central” does not necessarily mean that the instrument trades on only one venue, that all orders are publicly visible, or that trading, custody, and settlement are performed by the same institution.

Organization and representation

A limit order specifies a side, a quantity, and a price constraint. A buy limit order may execute only at its limit price or lower; a sell limit order may execute only at its limit price or higher. The limit therefore restricts execution price, not execution time or the probability of execution. An order to buy 500 shares at 40.00 does not guarantee that any shares will be purchased, even if transactions subsequently occur at 40.00: other orders may have priority, or those transactions may occur on another venue.

Orders on the bid side are ranked from higher to lower prices, while orders on the ask side are ranked from lower to higher prices. The highest bid is the best bid, btb_t, and the lowest ask is the best ask, ata_t. Their difference is the quoted bid–ask spread:

st=at−bt.s_t=a_t-b_t.

The midpoint is mt=(at+bt)/2m_t=(a_t+b_t)/2. It is a convenient reference price, but it need not be an executable price. In a normally functioning continuous book, compatible orders are matched rather than remaining crossed, although special trading states and market-data effects can produce apparent exceptions.

Each distinct price is a price level. A display might show bids for 800 shares at 39.99 and 1,500 shares at 39.98, opposite asks for 600 shares at 40.01 and 900 shares at 40.02. Price-level data aggregate the quantities of multiple orders. Order-level data instead identify individual orders, often through anonymous order identifiers, allowing their relative queue positions to be reconstructed. The amount of visible detail varies by venue, subscription, and data product.

The permissible price grid is determined by the tick size, the smallest allowed price increment under the applicable rules. Quantity increments, minimum order sizes, and round-lot conventions also affect how orders enter the book. A one-cent tick represents a much larger proportional increment for a security priced at 1.00 than for one priced at 1,000.00. Consequently, identical nominal tick sizes can produce different trading incentives across instruments.

A book is a changing state rather than a static table. New submissions add orders, cancellations remove them, amendments modify their terms, and executions reduce or eliminate resting quantities. Public displays may update less frequently than the matching system itself. Moreover, a participant may be able to observe an order without being able to trade against it—for example, where bilateral credit restrictions apply. The executable book, the published book, and a particular participant’s accessible book are therefore not always identical.

Matching rules and trading sessions

The matching engine applies the venue’s rules to incoming messages. In a continuous book, an incoming buy order is marketable if its price limit reaches or exceeds the best available ask; an incoming sell order is marketable if its limit reaches or falls below the best bid. A marketable order can execute immediately, potentially against several resting orders. A limit order that does not cross the opposite side normally joins the book, provided its other conditions permit it to rest.

Under price–time priority, better-priced orders execute before worse-priced orders, and earlier orders at a given price execute before later ones. Suppose two sell orders at 25.10 contain 100 and 300 shares, entered in that order. An incoming purchase of 250 shares would normally fill the first order completely and take 150 shares from the second. The remaining 150 shares retain the second order’s place, subject to venue rules. Raising a bid or lowering an ask can obtain price priority; simply joining an existing price generally means waiting behind earlier orders.

Other allocation systems distribute executions differently. Under pro rata allocation, an incoming quantity is divided among eligible resting orders broadly in proportion to their sizes, sometimes with minimum allocations, rounding rules, or priority exceptions. Some futures markets use hybrid systems combining time priority, size allocation, or special treatment for designated participants. These choices matter economically: time priority rewards early placement, while size-based allocation can encourage larger quoted quantities. Increasing an order’s size or changing its price commonly affects priority, but the precise treatment is venue-specific.

Execution price also depends on the market model. In many continuous books, an incoming order executes at the resting order’s price. Thus, a buy limit of 25.20 can receive an execution at 25.10 if that offer is available. This differs from a call auction, where orders accumulate and are matched at a scheduled or triggered clearing time, usually at a single price. Opening and closing auctions commonly choose a price that maximizes executable volume, with additional rules for resolving ties and handling imbalances. An auction can temporarily contain crossing buying and selling interest because execution is deferred.

Order processing requires an authoritative sequence. Two messages that appear simultaneous to outside observers must still be ordered, grouped, or otherwise resolved by the system. Venues therefore define when an order becomes accepted, how amendments are sequenced, and what happens during trading halts or recovery. Self-trade-prevention instructions may cancel or reduce potentially interacting orders from the same participant rather than allow an execution. These operational details determine actual allocation and can be as important as the headline priority rule.

Order instructions and visibility

The familiar distinction between a limit order and a market order concerns price protection. A limit order sets a worst acceptable price but may remain unfilled. A market order requests execution against available opposing interest without a trader-specified limit price. It does not guarantee a particular price or necessarily a complete fill: venue safeguards, exhausted liquidity, trading interruptions, or routing restrictions may stop execution. A marketable limit order occupies an intermediate position, seeking immediate execution while imposing an explicit price boundary.

Time-in-force instructions govern what happens to unexecuted quantities. A day order normally expires at the end of the specified trading session. A good-till-canceled order may persist across sessions, subject to the venue’s expiration and maintenance rules. An immediate-or-cancel order executes whatever quantity is immediately available within its terms and cancels the remainder. A fill-or-kill instruction generally requires immediate execution of the entire specified quantity or cancellation. These labels are not perfectly uniform across trading systems, so their detailed implementation matters.

A post-only instruction is designed to prevent an order from immediately taking liquidity. Depending on the venue, an otherwise marketable post-only order may be rejected, canceled, or repriced. A stop order is different: it becomes active only when a specified trigger condition is satisfied. A stop-market order becomes a market order after activation, while a stop-limit order becomes a limit order. Stop orders are often maintained outside the publicly displayed book until triggered, and the trigger may refer to trades, quotations, or another defined reference.

Visibility can be separated from executable size. An iceberg order, also called a reserve order in some markets, displays only part of its total quantity. As the visible portion is executed, another portion may become available. A fully hidden order displays no ordinary quoted size. Hidden and replenished quantities can receive different priority from displayed interest, and some systems assign a new time priority when a displayed portion replenishes. Consequently, a screen showing 200 shares at a price does not prove that only 200 shares can trade there.

Other instructions include minimum execution quantities, midpoint pricing, discretionary price ranges, and prices pegged to a reference quotation. A midpoint-pegged order, for example, can adjust as the best bid and ask change, sometimes subject to an additional limit. Such orders make the book more than a set of fixed-price queues: execution eligibility may depend on external reference prices, counterparty conditions, and hidden attributes. The visible book is therefore an incomplete representation of the matching engine’s full instruction set.

Liquidity, depth, and execution costs

Order-book liquidity has several dimensions. Tightness concerns the cost of trading a small quantity, commonly measured by the spread. Depth concerns the quantity available at or near current prices. Immediacy concerns how quickly a desired transaction can be completed, while resilience concerns how the book recovers after a liquidity disturbance. A market can have a narrow spread but little depth, making small transactions inexpensive and larger transactions costly. These dimensions should not be collapsed into a single measure.

For an immediately executable buy order of size QQ, the price of walking through successive ask levels can be written as

pˉ(Q)=∑ipixiQ,∑ixi=Q,\bar p(Q)=\frac{\sum_i p_i x_i}{Q}, \qquad \sum_i x_i=Q,

where pip_i is the execution price at level ii and xix_i is the quantity purchased there. If 100 shares are offered at 100.01 and another 200 at 100.02, a purchase of 250 shares would, assuming those orders remain available, buy 100 at the first level and 150 at the second. Its average price would be 100.016. This calculation describes consumption of a fixed book, not the full response of other traders.

Execution cost includes more than the quoted spread. Transaction costs can include venue fees, brokerage commissions, taxes, the opportunity cost of delayed or failed execution, and market impact. Slippage measures a difference from a specified reference price, but that reference must be identified: the decision price, arrival price, midpoint, and last transaction price answer different questions. A large order’s realized price path can differ substantially from a snapshot-based estimate because liquidity is added, canceled, or repriced during execution.

Displayed depth is contingent rather than committed for a fixed duration. Resting orders can often be canceled until they execute, while hidden liquidity may increase the quantity available beyond the displayed amount. A book that appears deep can therefore become shallow before an incoming order reaches it. Conversely, a large order can attract additional opposing interest. This distinction is central to liquidity risk: the ability to liquidate a position depends on market conditions at execution, not merely on an earlier quotation.

Execution quality can also be measured after a trade. For a transaction at price pp, midpoint mm, and trade direction dd, where d=1d=1 for a buyer-initiated trade and d=−1d=-1 for a seller-initiated trade, a common effective-spread measure is 2d(p−m)2d(p-m). Comparing the execution with a later midpoint helps separate compensation retained by the liquidity provider from subsequent price movement. Such measures depend on timing conventions, trade-direction classification, and the chosen evaluation horizon.

Economic incentives and price formation

Participants face a trade-off between providing and taking liquidity. A resting order offers another trader the option to transact at a specified price. Its owner may receive a better price than an immediate marketable order would obtain, but accepts uncertain execution and the risk that the order will trade precisely when its price has become unfavorable. A market maker manages this trade-off repeatedly, often maintaining both bids and offers while adjusting prices and quantities as conditions change.

One source of risk is adverse selection. Suppose a company releases favorable news while a sell order remains at an outdated price. Faster or better-informed traders may buy from that order before its owner cancels it. The seller earns the spread only in an accounting sense if the asset’s subsequent value rises by more than that spread. Differences in information and reaction speed therefore influence how much compensation liquidity providers require. The relevant information asymmetry need not involve private company information; it can arise from unequal access to public data or unequal processing speed.

Inventory creates another incentive. A liquidity provider that accumulates a large position becomes exposed to further price movements and may alter its quotes to encourage inventory reduction. A dealer with excessive inventory might lower both bid and ask relative to an otherwise comparable dealer. Funding constraints, position limits, and risk aversion can further change willingness to supply liquidity. These mechanisms connect individual order placement to the broader supply and demand for immediate trading services.

Queue position is itself economically valuable. Under price–time priority, an early order may execute before later orders at the same price, but its owner cannot generally transfer that priority to a substantially modified order. A smaller tick allows a competitor to gain price priority at a smaller price concession; a larger tick makes such improvement more expensive and can increase the importance of queue position. The relationship is not monotonic for every outcome: spreads, displayed quantities, cancellations, and trading away from displayed venues can respond differently.

The book contributes to price discovery through both executions and revisions to trading interest. A price can move when the best offer is canceled even if no trade occurs. Conversely, a transaction can occur without changing the best prices. The last traded price, midpoint, and auction clearing price are therefore distinct indicators. A classic theoretical contribution by Lawrence Glosten and Paul Milgrom in 1985 showed how informational differences can generate a bid–ask spread. Limit-order-book models extend such questions to endogenous depth, queueing, cancellation, and strategic order choice.

Historical development and market architecture

Order books predate electronic trading. Brokers and exchange specialists maintained records of unexecuted customer orders, while floor trading combined these instructions with verbal negotiation and dealer quotations. Computerization changed both the scale and visibility of the process. Instinet, founded in 1969, was an early electronic institutional trading network. Nasdaq began operating in 1971 as an automated quotation system; its initial architecture should not be confused with a modern centralized electronic limit order book, because displayed dealer quotations and automated order matching are different functions.

Electronic order-driven systems expanded substantially during the 1980s and 1990s. The Paris Bourse introduced its CAC electronic trading system in 1986. The London Stock Exchange launched SETS in 1997, introducing an electronic order book for a group of leading shares. Other exchanges adopted related systems, while electronic communication networks competed with established marketplaces. These changes moved trading from physical interaction toward computer-mediated submission, sequencing, and matching, although floor-based and dealer-based arrangements continued in various forms.

Market-rule changes also reshaped books. In the United States, equity-market decimalization was completed in 2001, replacing fractional quotation conventions with decimal prices and generally permitting one-cent increments for ordinary quotations in most listed shares. Smaller increments reduced the minimum price improvement needed to step ahead of an existing order. European equity trading became more fragmented after implementation of the Markets in Financial Instruments Directive in 2007, as additional trading venues competed with traditional exchanges.

A modern instrument may consequently have several simultaneous books. Each venue maintains its own queues, fees, trading states, and available orders. A consolidated quotation can identify the best displayed prices across venues, but it does not merge their orders into one physical queue or guarantee that every participant has identical access. A smart order router selects destinations using price, available size, fees, access constraints, and estimates of execution probability. Arbitrage and competition connect venue prices, but differences can persist briefly because communication and execution take time.

The relationship between matching and post-trade infrastructure is equally important. An execution creates obligations that must pass through clearing and settlement, often involving a central counterparty. The matching venue, clearing institution, and securities depository perform different functions. Futures, equities, foreign exchange, and digital-asset markets may all use order books while differing greatly in collateral arrangements, credit screening, legal finality, and custody. A common matching interface therefore does not imply a common financial or institutional structure.

Market data and quantitative models

Order-book data are commonly distributed as snapshots, incremental updates, or event messages. A snapshot reports the state at a particular time, while an incremental feed reports changes such as additions, cancellations, executions, and quantity reductions. Market-by-price data aggregate orders at each price; market-by-order data expose more detail about individual queues. Vendors sometimes use labels such as Level I and Level II, but those labels do not have perfectly uniform meanings across markets and products.

Reconstructing a book requires correct message sequencing. A missing cancellation can leave nonexistent quantity in a reconstructed book, while a missed execution can create false depth. Sequence numbers, recovery messages, instrument definitions, and trading-status updates are therefore essential components of analysis. Timestamps must also be interpreted carefully: exchange event time, feed publication time, and local receipt time describe different stages. An apparently predictive signal can instead reflect mismatched clocks or a delayed feed.

A simple measure of top-of-book imbalance is

It=Vtb−VtaVtb+Vta,I_t=\frac{V_t^{b}-V_t^{a}}{V_t^{b}+V_t^{a}},

where VtbV_t^{b} and VtaV_t^{a} are displayed quantities at the best bid and ask. Positive imbalance indicates more displayed size on the bid side. It may contain information about near-term price changes, but it is not a universal trading rule: cancellations, hidden interest, queue structure, and the spread regime affect its interpretation. Order-flow imbalance, which measures changes associated with submissions, cancellations, and executions, is related but distinct from this static quantity ratio.

Statistical models represent arrivals of orders, cancellations, and trades as events. Early simplified models often used Poisson processes, while later work incorporated state-dependent intensities, queueing systems, and Hawkes processes, in which an event can temporarily raise the rate of subsequent events. For example, one aggressive trade may be followed by further trades and rapid quote revisions. Queue-reactive models make event rates depend on the current book state. These approaches describe different aspects of market dynamics and require assumptions about how traders respond to events.

Predictive models, including machine learning, use book states and message histories to estimate future prices, fills, or execution costs. Their evaluation is difficult because observations are dependent, market conditions change, and trading decisions alter the environment being modeled. A historically visible price does not establish that a simulated order would have filled there. Credible backtesting must account for queue position, partial fills, communication delays, fees, and unavailable liquidity. Separating training and testing periods is necessary but does not by itself resolve these execution and feedback problems.

Trading algorithms, latency, and queue competition

Algorithmic trading interacts with order books in several distinct ways. Execution algorithms divide a large parent instruction into smaller child orders, balancing immediate cost against delay and information leakage. A time-weighted strategy schedules trades across a period, while a volume-weighted strategy seeks participation consistent with an estimated market-volume profile. More adaptive algorithms alter urgency as prices, volatility, liquidity, and remaining time change. These methods differ from strategies whose primary objective is forecasting short-term returns.

For a passive order, queue position affects both the probability and the character of execution. An order far back in a queue may not fill unless substantial opposing flow arrives or earlier orders disappear. An order near the front can fill more readily but may be exposed to informed or urgent traders. Because cancellations ahead of an order can advance its position without any trades occurring, queue evolution cannot be inferred from transaction volume alone. Aggregate price-level data often support only estimates of an individual order’s position.

Latency competition arises because opportunities can disappear between observation and action. A participant may receive new information, send a cancellation, and nevertheless have the resting order executed before the cancellation reaches the matching engine. Co-location places trading systems near exchange infrastructure to reduce communication delay; direct data feeds and optimized software reduce other components. High-frequency trading encompasses several strategies using rapid automated decisions, rather than one uniform activity. Fast systems may provide liquidity, take liquidity, or do both at different moments.

Fees influence these choices. In a maker–taker arrangement, a venue typically pays a rebate for qualifying liquidity-providing executions and charges liquidity takers; an inverted arrangement reverses the general pattern. Actual schedules can depend on volume tiers, membership, and order type. A resting order with a favorable displayed price may thus have a different net economic value after fees. Routing decisions can also create agency issues when a broker’s fee incentives differ from a customer’s execution objective.

Alternative market designs seek to modify the role of speed. Some venues introduce intentional delays, while frequent batch auctions group orders into short intervals instead of matching every eligible arrival immediately. Their effects depend on details such as whether cancellations are delayed, how reference prices update, and how competing venues remain connected. These designs are applications of mechanism design: changing timing and allocation rules changes the strategies participants find profitable, rather than simply making an otherwise unchanged market faster or slower.

Market integrity, safeguards, and disruption

A visible book can be used to create misleading impressions. Spoofing generally involves placing orders with an intention to cancel them before execution in order to influence others, while layering commonly refers to deceptive order placement across multiple price levels. Legal definitions and evidentiary requirements vary by jurisdiction. Frequent cancellation alone does not establish market manipulation: legitimate participants routinely revise orders as prices and risks change. Surveillance therefore examines patterns, relationships between orders and trades, and evidence concerning purpose and conduct.

Operational failures pose a separate threat. A software defect can generate unintended orders, omit risk checks, or leave quotes active after a participant loses connectivity. Exchanges and brokers use controls such as maximum order sizes, price collars, message-rate limits, duplicate-message checks, and kill switches. Cancel-on-disconnect arrangements can remove specified orders when a trading connection fails, although their scope and triggering conditions vary. These mechanisms address operational risk but cannot eliminate every failure mode.

Trading safeguards limit executions during extreme conditions. Price bands can reject or prevent trades far from defined reference levels, and volatility interruptions can pause continuous trading or move the instrument into an auction. Market-wide circuit breakers operate at a broader level. Such rules involve trade-offs: a pause may allow information and liquidity to accumulate, while an overly restrictive control may prevent legitimate adjustment to new information. The choice of reference prices, thresholds, and reopening procedures is therefore integral to market design.

The U.S. market disruption of May 6, 2010, commonly called the Flash Crash, demonstrated that rapid electronic trading and linked markets could experience severe, short-lived price dislocations. The joint report of the U.S. Securities and Exchange Commission and the Commodity Futures Trading Commission examined interactions involving equity-index futures, automated execution, liquidity withdrawal, and equity markets. The event does not support a single universal explanation for all abrupt price movements; it illustrates why depth observed before stress may not remain available during it.

Regulation and surveillance consequently extend beyond the matching rule. Relevant obligations can include fair access, accurate market data, controls on automated trading, record retention, and duties governing customer execution. In fragmented markets, investigators may need order and trade records from several venues to reconstruct an episode. Reliable analysis distinguishes displayed intentions from completed trades and ordinary competitive repricing from deceptive conduct. The same message history that supports quantitative research also provides the evidence needed to examine whether a market operated according to its rules.