1.4 The continuous limit order book · M01 Market structure

Motivation

Between auctions, NSE and BSE run a continuous limit order book (CLOB). Every intraday price you see is a trade in this book. Understanding it explains why "the price" is two prices, why a market order costs more than the spread when it is large, and why speed matters.

1.4.1Limit orders and priority#

Definitions

A limit order specifies side, quantity and a limit price. A buy limit at $p$ will trade at $p$ or lower; a sell limit at $p$ at $p$ or higher. If it cannot trade immediately, it rests in the book.
The bid is the highest resting buy price; the ask (offer) the lowest resting sell price. The spread is ask − bid. The mid is their average. The depth at a level is the total resting quantity at that price.
A marketable order is one that can trade on arrival: a buy limit at or above the ask, or a sell limit at or below the bid. A market order has no price limit and trades against the best available prices until filled.
The resting order is the maker (it supplied liquidity); the arriving order is the taker.
Price-time priority. Better prices trade first. At the same price, the earlier order trades first. The trade price is the maker's price.
Tick size: the minimum price increment. NSE sets it by price band through circulars; check the current circular before hard-coding it.

1.4.2Retail order types#

Order types available to an Indian retail trader (exchange features and common broker features; the details vary by broker):

Limit and market orders, as above.
Stop-loss limit (SL): a buy (sell) order that becomes a limit order only when the last traded price rises to (falls to) a trigger price. It protects a position without being visible as a marketable order beforehand, but it may not fill if price gaps through the limit.
Stop-loss market (SL-M): becomes a market order on trigger. It always fills if there is liquidity, but at an unknown price. Several brokers do not offer SL-M in F&O.
Validity: Day (expires at the end of the session) or IOC (immediate-or-cancel: fill what can be filled on arrival, cancel the rest).
Disclosed quantity / iceberg: only part of the order is displayed; the rest is hidden and refreshes after the visible part fills.
After-market order (AMO): entered outside market hours and sent at the next open.
GTT (good-till-triggered): a broker-side, not exchange-side, conditional order that sits on the broker's server, sometimes for up to a year, and sends a limit order to the exchange when the trigger is hit. The exchange never sees it until then, so it has no queue position.
Product codes (broker margin categories, not exchange order types): CNC (cash and carry, delivery), MIS (intraday; auto-squared off before the close), NRML (carry-forward F&O). Since the peak-margin rules (2021), brokers cannot give intraday leverage beyond exchange margins.
Algorithmic orders: SEBI's retail algo framework (circular SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/0000013 of 4 Feb 2025; applicable to all brokers from 1 Apr 2026 under circular SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/132 of 30 Sep 2025, para 8) makes the broker the principal and any algo provider its agent. Every algo order sent through a broker API carries an exchange-issued identifier; open APIs are not allowed (access only through a vendor-client-specific API key and a static IP whitelisted by the broker, OAuth and two-factor authentication); an algo you write yourself must be registered with the exchange through your broker once it exceeds the orders-per-second threshold set by the exchanges; and third-party providers must be empanelled with the exchanges. Any live automation you build after this course must comply (M14 · 14.10.2 lists the operational controls).

1.4.3Slippage and market impact#

Market order cost: the walk up the book Suppose asks rest at prices $a_1 < a_2 < \dots$ with depths $d_1, d_2, \dots$ . A market buy of $Q$ fills $\min(Q, d_1)$ at $a_1$ , then continues. Let $m$ be the mid before the order. The volume-weighted fill price and slippage in basis points are

\bar p(Q) = \frac{1}{Q}\sum_{i} a_i \min\Big(d_i, \big(Q - \textstyle\sum_{l<i} d_l\big)^+\Big), \qquad \text{slip}(Q) = 10^4\,\frac{\bar p(Q) - m}{m}.

Because each additional share fills at a price no better than the previous one, $Q\,\bar p(Q)$ is a convex, piecewise-linear function of $Q$ , and $\bar p(Q)$ is non-decreasing. This is the simplest version of market impact studied in M05 · 5.5.

Worked example

A ₹100 stock has bids at 99.95, 99.90, … and asks at 100.05, 100.10, 100.15, …, 100 shares at each level. A market buy of 250 fills 100 at 100.05, 100 at 100.10 and 50 at 100.15. Average price ₹100.09, slippage 9 bp against the mid of ₹100. A market sell of 100 fills at 99.95: 5 bp. These are the fixtures in the tests of Exercise m01_execution (in the app: Exercises → m01_execution).

Code Exercise m01_order_book (in the app: Exercises → m01_order_book) asks you to build the book with price-time priority, cancels and market orders. Its tests include a 3,000-order fuzz test that checks two invariants: quantity is conserved (submitted = filled + cancelled + resting) and the book is never crossed. Exercise m01_execution builds market_impact and round_trip_bps on top of it.

Practitioner notes

Queue position is an asset. A limit order placed early at the bid trades before later orders at the same price. Market makers invest heavily to be first in the queue. A retail limit order joins the back.
Displayed depth is not reliable depth. Orders are cancelled and replaced many times a second. SEBI's FY26 F&O study reports that about 99% of prop-trader and FPI gross profit came from algorithmic entities (1.11), which is a measure of who controls the queue.
A limit order is an option given away for free. A resting buy at ₹99.95 lets anyone sell to you at that price if news arrives. You are paid the spread for the risk of being picked off. M05 · 5.3 models this (Glosten–Milgrom).