Algorithmic trading

Algorithmic trading (often shortened to “algo trading”) refers to the use of computer‑based models that automatically generate, place, modify or cancel orders without human intervention. The logic behind the algorithm is derived from quantitative analysis, statistical arbitrage, or pre‑set execution strategies, and it runs on a high‑speed server that connects directly to the exchange’s order‑matching engine.

In the Indian context, SEBI defines an algorithm as “a set of instructions that, when executed, result in a trade or a series of trades”. The primary purpose is to achieve better price discovery, lower market impact, and consistent execution quality, especially for large institutional orders in equity derivatives.

For the NISM exam, you will be asked to identify characteristics of algorithmic trading, recognise common algorithm types, and apply simple calculations such as VWAP or implementation shortfall. Remember that many exam questions test your ability to link a concept (e.g., “minimise market impact”) with the appropriate algorithmic approach.

The first component is market data feed, which supplies real‑time price, volume and order‑book information. Accurate and low‑latency data is essential because the algorithm’s decision logic depends on the latest market state.

The second component is the strategy engine. This is where the mathematical model lives – it could be a VWAP schedule, a statistical arbitrage rule, or a risk‑limit checker. The engine receives data, evaluates the rule set, and generates order instructions.

The third component is the execution gateway. It translates the algorithm’s instructions into exchange‑compatible order messages (e.g., limit, market, iceberg). It also handles acknowledgements, cancellations, and post‑trade reporting. Finally, a risk‑monitoring module enforces SEBI‑mandated pre‑trade checks such as order‑to‑trade ratio, price‑band limits and maximum exposure.

VWAP (Volume Weighted Average Price) aims to execute an order such that the average price matches the market’s volume‑weighted average over a specified horizon. It is popular for large institutional orders because it spreads the trade proportionally to market volume, reducing impact.

TWAP (Time Weighted Average Price) distributes the order evenly across time intervals, regardless of volume. It is useful when volume patterns are unpredictable or when a trader wants a simple, time‑based execution schedule.

Implementation Shortfall (also called “arrival price” algorithm) seeks to minimise the difference between the decision price and the actual execution price, factoring in both market impact and opportunity cost. It is often chosen when the trader’s primary concern is the total cost of execution.

Percentage of Volume (POV) submits orders as a fixed percentage of the prevailing market volume, automatically adjusting to market liquidity. Iceberg hides the true order size by displaying only a small visible portion, protecting large orders from signalling. Sniper or “aggressive” algorithms target specific price levels and execute quickly when those levels are reached.

The VWAP formula aggregates price and volume across the chosen horizon, delivering a single benchmark price. In practice, brokers compare the actual execution price of each slice to the VWAP; a “VWAP‑slippage” of less than 5 bps is often considered acceptable for large equity‑derivative orders.

For the exam, remember that VWAP is a *volume‑weighted* average – the denominator is total volume, not the number of intervals. A common mistake is to average the prices first and then weight them, which yields an incorrect result.

SEBI’s algorithm‑trading guidelines require that the algorithm disclose its VWAP schedule to the exchange and maintain an audit trail of the calculated VWAP versus actual execution. Candidates may be asked to identify which compliance document (e.g., “Algo‑Trading Policy”) should contain this information.

Implementation shortfall captures both *explicit* costs (commissions, taxes) and *implicit* costs (market impact, opportunity cost). The algorithm attempts to minimise this metric by balancing urgency against price impact, often using a dynamic schedule that slows down when the market moves unfavourably.

In NISM questions, you may be given a decision price and a series of execution prices, then asked to compute the total shortfall. Remember to multiply the price difference by the total quantity, not by each slice individually unless the question explicitly asks for per‑slice shortfall.

SEBI requires that firms disclose the methodology used to calculate implementation shortfall in their risk‑management framework. This ensures transparency for investors and regulators alike.

SEBI’s “Guidelines on Algorithmic Trading” (issued in 2015 and updated periodically) mandate that every algorithmic strategy be approved by the exchange’s Surveillance Committee and that a pre‑trade risk check be performed for each order. The checks include order‑size limits, price‑band restrictions, and order‑to‑trade ratio caps (typically 10:1).

Broker‑dealing members must maintain an audit trail for at least five years, capturing order submission time, parameters, execution details, and any manual overrides. The audit logs must be made available to SEBI on request.

For the exam, you may be asked which of the following is *not* a SEBI requirement for algo trading. Common distractors include “mandatory co‑location” (which is a market‑practice, not a regulation) and “minimum daily trade volume” (no such threshold exists).

Latency is the time delay between the moment an algorithm decides to act and the moment the exchange receives the order. In Indian equity‑derivative markets, latency is measured in microseconds (µs) to milliseconds (ms). Co‑location services offered by NSE and BSE reduce latency to sub‑millisecond levels.

Market microstructure – the way the order book is organised, tick size, and matching algorithm – influences how an algo performs. For example, a VWAP algorithm may suffer higher slippage in a thin order‑book because each slice moves the price more.

Exam questions often present a scenario where a trader chooses between a co‑located server (0.5 ms latency) and a remote server (5 ms latency). The correct answer hinges on the trade‑off between cost (co‑location fees) and execution quality, especially for high‑frequency strategies.

Before deployment, algorithms undergo back‑testing using historical tick data to verify that the strategy would have met its performance targets. A robust back‑test includes transaction cost modelling, slippage assumptions and out‑of‑sample validation (walk‑forward analysis).

Live monitoring involves real‑time dashboards that track execution quality metrics such as average execution price vs. benchmark, fill rate, and latency per slice. Alerts are triggered if any metric deviates beyond pre‑set thresholds, prompting a manual intervention.

SEBI expects firms to document both the back‑testing methodology and the live‑monitoring framework. In the exam, you may be asked to select the correct sequence: back‑test → simulation → live‑monitor → periodic audit.